M. Caroline Burton, MD

Immune checkpoint inhibitors (ICIs), a form of immunotherapy, have changed the management of cancer since their introduction in 2011.¹ They were initially tested on melanoma.² Their use in the advanced stages of the disease demonstrated a 2-year survival of 18% compared with 5% by using other therapies.³ Similar results were observed in nonsmall cell lung carcinoma (NSCLC); the overall survival benefit was 3 months with the use of ICIs compared with traditional chemotherapy (42% and 24% at 1 year, respectively).⁴ Antitumor activity has also been seen in the treatment of other malignancies, including renal cell carcinoma,⁵ bladder carcinoma,^6,7 head and neck carcinoma,⁸ colorectal cancer,⁹ Hodgkin lymphoma,¹⁰ and, more recently, hepatocellular carcinoma.¹¹ The use of ICIs has also been linked to serious complications.¹² Although the skin, kidneys, lungs, and endocrine and nervous systems may be affected, complications of the gastrointestinal (GI) tract are frequent and can be life-threatening.^12-16 We performed a thorough review of the literature to familiarize hospitalists with the mechanism of action and uses of ICIs, the clinical presentation of their GI toxicity, and the current recommendations regarding diagnosis and treatment.

CASE PRESENTATION

A 66-year-old man was admitted to our institution with a 1-week history of severe, diffuse abdominal pain and profuse watery diarrhea. He reported having more than 8 watery bowel movements per day and denied fever, recent travel, ill contacts, or ingestion of undercooked food. He had a history of metastatic melanoma and was undergoing treatment with both nivolumab and ipilimumab; the drugs were started 6 weeks prior to presentation. Physical examination revealed a heart rate of 110 beats/minute while supine and 123 beats/minute while standing, blood pressure of 112/69 mm Hg while supine and 92/62 mm Hg while standing, and a temperature of 37.2°C. He was in mild distress and had dry oral mucosa. Abdominal examination revealed hyperactive bowel sounds and mild diffuse abdominal tenderness with no guarding or rebound. His extremities were cool, but peripheral pulses were present. Initial laboratory results included a hemoglobin level of 15.3 g/dL (range 12.0-16.0 mg/dL), white blood cell count 14.2 × 10⁹/L (range 4.5-11.0 × 10⁹/L), and platelet count 236 × 10⁹/L (range 150-400 × 10⁹/L); other test results included a sodium level of 130 mmol/L (range 135-145 mmol/L), potassium 2.3 mmol/L (range 3.5-5.5 mmol/L), serum creatinine 2.2 mg/dL (range 0.8-1.3 mg/dL), blood urea nitrogen 72 mg/dL (range 8-21 mg/dL), and serum venous lactate 5.9 mmol/L (range 0.9-1.7 mmol/L).

MECHANISM OF ACTION AND USES OF ICIS

T-cell lymphocytes play a pivotal role in acquired immunity, but their function requires an appropriate balance between stimulatory and inhibitory signals to prevent autoimmunity.¹⁷ Immune checkpoint molecules are used by the immune system to assist with this balance.¹⁸ Although several of these molecules exist, the cytotoxic T-lymphocyte antigen-4 (CTLA-4) and programmed cell death-1 (PD-1) are among the most widely studied.¹²

Activation or inhibition of T cells depends on the interaction of their receptors with ligands located on the surface of other cells. Both CTLA-4 and PD-1 are receptors located on the surface of T-cell lymphocytes that inhibit the function of T cells after binding with their ligands.^19-21 Cancer cells often use this mechanism to avoid immune recognition and promote their survival.^18,21,22 Importantly, ligands that bind CTLA-4 are expressed by numerous tissues throughout the body, contrary to ligands that bind to PD-1 (PD-L1 or PD-L2), which are more specific to tumor cells (Figure).^21-23 ICIs are monoclonal antibodies that block these pathways and increase T-cell activity.¹⁸

Ipilimumab is a monoclonal antibody directed against CTLA-4.²⁴ After demonstrating survival benefits in patients with unresectable and metastatic melanoma, ipilimumab was the first ICI approved for use by the US Food and Drug Administration (FDA).^1,3 Another monoclonal antibody directed against CTLA-4, tremelimumab, is not currently approved for use by the FDA.

Pembrolizumab and nivolumab are monoclonal antibodies against PD-1. The FDA approved them for the treatment of advanced melanoma in 2014¹⁷ and metastatic NSCLC in 2015.¹² Nivolumab was also approved for the treatment of renal cell carcinoma and for advanced-stage melanoma in combination with ipilimumab.^12,17 Atezolizumab, avelumab, and durvalumab are PD-L1 inhibitors. All 3 are approved by the FDA for treatment of advanced urothelial carcinoma.^25-27 Atezolizumab is also approved for the treatment of metastatic NSCLC,²⁸ and avelumab is approved for treatment of metastatic Merkel cell carcinoma.²⁹ Table 1 summarizes the medications, their target, and FDA-approved indications.^{1,12,17,26,27,29,30}

 

TOXIC PROFILE

Because of the sustained T-cell activation, ICIs have been associated with autoimmune-like toxicities known as immune-related adverse events (irAEs).^19,31 Because the PD-1/PD-L1 pathway is more tumor-specific than the CTLA-4 pathway,^21-23 there is a higher incidence of serious irAEs seen with ipilimumab, reported to be around 27%.^18,22 Furthermore, the risk of developing irAEs is dose-dependent and can increase up to 55% when anti-CTLA-4 are used with other ICIs such as nivolumab.^13,32-34

The skin and GI tract are the most commonly involved organs.^14-16 Skin is affected in 50% of patients receiving ipilimumab and 40% of patients on nivolumab or pembrolizumab, often in the form of a rash or pruritus.^12,35-37 The rash is often described as faintly erythematous, reticular, and maculopapular and typically affects the trunk and extremities.³⁸ Importantly, these events usually occur within the first 2 weeks of treatment, and fewer than 5% are severe.^12,36,39 A higher percentage of severe adverse events occurs in the GI tract, with a reported incidence of 12%.^3,14,36,39

CLINICAL PRESENTATION

Although any portion of the GI tract can be affected by ICIs, the lower GI tract is most commonly involved. Clinical signs include watery diarrhea, colitis, and enteritis.^15,19 Less commonly, the upper GI tract is involved, and clinical manifestations include aphthous ulcers, esophagitis, and gastritis.^40,41 GI symptoms usually begin 6 weeks after the initial dose of ICIs and typically follow cutaneous manifestations.^15,20,36,37 However, they can occur as late as 4 months after the last dose.¹⁹ Watery, nonbloody diarrhea is the most common presentation of GI involvement, occurring in 19% of patients receiving anti-PD-1/PD-L1 and 33% receiving anti-CTLA-4 medications.^19,37 When patients receive both therapies, the incidence rate increases to 44%.³² The clinical severity of diarrhea can be graded on a scale of 1 to 5 according to the Common Terminology Criteria for Adverse Events (CTCAE) (Table 2).⁴² Though most patients have mild disease (grade 1 or 2), close to 3% develop severe diarrhea (grade 3 or higher) with electrolyte disturbances and weight loss.^14,20

Colitis, defined by either the presence of symptoms or radiologic findings suggestive of inflammation, occurs less often than diarrhea alone, with a reported incidence of 2.3%.^37,43 This incidence increases to almost 12% when anti-CTLA-4 and anti-PD-1/PD-L1 are combined.³² Colitis symptoms include abdominal pain (20%), nausea and vomiting (15%), fever (12%), and, less often, bloody diarrhea or rectal bleeding.^19,20 Colitis severity is graded according to the CTCAE (Table 2).⁴² Most patients have mild colitis (grade 1 or 2).¹⁹ The risk for developing severe colitis (grade 3 or higher) is almost 10 times higher with the use of anti-CTLA-4 compared with anti-PD-1/PD-L1 agents.⁴³ Patients with severe disease are at risk of developing life-threatening complications, such as ileus, toxic megacolon, bowel ischemia, necrosis, or even perforation, which has been reported in up to 5% of patients with colitis because of ipilimumab.^13,17

CASE APPROACH STRATEGY

Based on the patient’s symptoms, physical findings, and temporal relationship to ICI therapy, he was believed to have immune-mediated colitis. Stool studies, including those looking for ova and parasites, Clostridium difficile polymerase chain reaction (PCR), and stool cultures were negative.

DIAGNOSIS

In a patient undergoing ICI treatment who has diarrhea, the initial assessment should exclude C. difficile and Salmonella by stool culture, PCR, or pathogenic antigens.¹⁹ Cytomegalovirus reactivation should also be considered. Immune-mediated colitis and infection can coexist; thus, a positive infectious etiology does not rule out the presence of immune colitis or vice versa.⁴⁴ Fecal calprotectin, a marker of neutrophil-associated inflammation, is nonspecific for ICI-induced colitis; however, it may help to distinguish inflammatory from noninflammatory diarrhea.^33,45

No clear guideline exists for the use of abdominal imaging. Some experts suggest using computed tomography in patients with severe, persistent, or progressive symptoms in order to exclude bowel obstruction, toxic megacolon, or perforation.^19,46

In patients with typical symptoms, and after infectious etiologies are ruled out, empiric use of corticosteroids can be initiated without an endoscopic evaluation, which is not necessary to establish a diagnosis and rarely changes management.^12,37,47 In patients with atypical presentations or for whom the diagnosis remains in question, endoscopic evaluation with biopsies may be required. Macroscopic findings may be similar to those seen with inflammatory bowel disease (IBD), including erythema, edema, ulceration, granularity, or loss of vascular pattern. Although immune-mediated colitis affects the descending colon more often than IBD, this feature and any macroscopic findings are insufficient to make this distinction.^20,36 Furthermore, the lack of macroscopic abnormalities does not rule out immune-mediated colitis.²⁰

When endoscopic biopsies are obtained, histologic findings for anti-CTLA-4 medications (eg, ipilimumab) usually follow 3 patterns: neutrophilic infiltrate (46%), lymphocytic infiltrate (15%), and mixed infiltrate (38%).⁴¹ Other findings include crypt abscesses and tissue destruction.²⁰ No biopsy-specific pattern has been described with anti-PD-1/PD-L1 medications, such as nivolumab or pembrolizumab.¹⁸ A normal colonic tissue does not exclude the presence of an irAE, as cases of isolated ileitis⁴⁸ or enteritis⁴⁹ without colitis can also occur.

 

CASE MANAGEMENT STRATEGY

The patient was started on intravenous (IV) methylprednisolone 2 mg/kg twice a day. After 48 hours, he still had more than 7 episodes of diarrhea per day, so he was treated with 1 dose of infliximab 5 mg/kg without stopping corticosteroids. Within 72 hours, the patient’s abdominal pain improved and his diarrhea stopped. He was discharged on an 8-week taper of prednisone starting at 1 mg/kg/day, pneumocystis pneumonia (PCP) prophylaxis was started, and ICI therapy was discontinued indefinitely.

MANAGEMENT OF COLITIS

Several principles should be considered in managing immune-mediated colitis: (1) management for adverse events of anti-CTLA-4 and anti-PD-1/PD-L1 should be the same; (2) though guidelines were made for patients with melanoma, they can be used to treat patients with other types of cancer; and (3) treatment should be started as early as possible, ideally within 5 days of symptom onset, as this hastens clinical improvement and decreases the incidence of complications.²⁰ Treatment is summarized in Table 3.

Management of grade 1 and 2 colitis is mainly supportive, consisting of fluid and electrolyte replacement, the American Dietetic Association colitis diet, and antimotility agents, such as loperamide, oral diphenoxylate hydrochloride, or atropine sulfate.^36,37 Persistent grade 2 symptoms (lasting >3 days), should prompt initiation of 0.5 to 1 mg/kg/day of oral prednisone or an equivalent.¹⁹ If symptoms do not improve with oral corticosteroids, patient hospitalization for IV corticosteroids should be considered.³⁷ Importantly, opioids and antidiarrheals may mask the pain and severity of symptoms and, therefore, should be used cautiously.¹⁹

Patients with grade 3 and 4 colitis (≥7 stools per day, severe abdominal pain, or complications) require the use of systemic corticosteroids at a dose of 1 to 2 mg/kg/day of prednisone or an equivalent.¹⁵ Patients who fail to respond to prednisone alone may benefit from the addition of oral budesonide at a dose of 9 to 12 mg/day.⁵⁰ In severe cases of colitis, hospitalization may be necessary for IV hydration, electrolyte replacement, and IV methylprednisolone at a starting dose of 2 mg/kg twice a day for 1 to 2 days before transitioning to oral corticosteroids.^12,15 Though improvement is usually noted within the first 2 weeks of treatment, prednisone should be slowly tapered over a period of 4 to 8 weeks to ensure complete healing and prevent relapse.^20,36 Patients who receive an equivalent dose of prednisone 20 mg daily during a period of 4 weeks or more should receive PCP prophylaxis.⁵¹ Some patients fail to respond to IV corticosteroids despite adequate dosing. Many of these patients have severe disease, possibly because of delayed recognition and initiation of treatment.¹⁹ As with IBD, the addition of infliximab to corticosteroids at 5 mg/kg as a single dose is usually successful for this population subset.^52-54 Although a response is seen within 1 to 3 days,⁴¹ some patients benefit from an additional dose of infliximab 2 weeks after the initial dose.¹⁹ If sepsis or perforation is suspected at any point, corticosteroids or infliximab should be avoided and antibiotics should be started immediately.^15,19 Patients with a medically unresponsive disease may require partial or complete colectomy.²⁰ The use of prophylactic budesonide to prevent diarrhea or colitis has not been proven effective and should not be used.⁵⁵ Despite complications, mortality from colitis has markedly decreased given the increased awareness of this adverse event, reduction in the time to recognition and treatment, and increased adherence to corticosteroids.¹²

Treating physicians may be delayed in starting appropriate therapy because patients are concerned that using corticosteroids will negatively impact immunotherapy efficacy. Current evidence shows that the use of temporary immunosuppression to treat irAEs does not affect overall survival, efficacy, or time to treatment failure of the ICI.^12,56 Restarting ICI therapy is a complex decision and should always be individualized. In grade 1 and 2 colitis, ICI therapy is typically restarted after symptoms have improved.⁵ In grade 3 and 4 colitis, ICI therapy is often permanently discontinued.²⁰

CONCLUSION

ICIs have not only increased our understanding of the biology of cancer, but they have also improved survival in advanced stages of malignancies like melanoma, NSCLC, and renal cell carcinoma. The expanding use of these medications increases the likelihood that healthcare providers will encounter patients experiencing their adverse events.

Immune-mediated GI adverse events include a wide range of symptoms, from mild diarrhea to severe colitis complicated by perforation and death. Diagnosis requires exclusion of an infectious process. Early recognition and treatment with corticosteroids or another immunosuppressant such as infliximab hastens recovery and decreases complications and mortality. Treatment should be started within 5 days of symptom onset. Corticosteroids should be slowly tapered for no less than 4 weeks to prevent relapse and PCP prophylaxis administered in appropriate patients. Restarting ICI therapy may be considered in cases of mild colitis, but in severe cases, ICI therapy is usually discontinued.

 

Disclosure

Julian Marin-Acevedo, Dana Harris, and M. Caroline Burton have no conflicts of interest or funding sources to declare.

Immune checkpoint inhibitors (ICIs), a form of immunotherapy, have changed the management of cancer since their introduction in 2011.¹ They were initially tested on melanoma.² Their use in the advanced stages of the disease demonstrated a 2-year survival of 18% compared with 5% by using other therapies.³ Similar results were observed in nonsmall cell lung carcinoma (NSCLC); the overall survival benefit was 3 months with the use of ICIs compared with traditional chemotherapy (42% and 24% at 1 year, respectively).⁴ Antitumor activity has also been seen in the treatment of other malignancies, including renal cell carcinoma,⁵ bladder carcinoma,^6,7 head and neck carcinoma,⁸ colorectal cancer,⁹ Hodgkin lymphoma,¹⁰ and, more recently, hepatocellular carcinoma.¹¹ The use of ICIs has also been linked to serious complications.¹² Although the skin, kidneys, lungs, and endocrine and nervous systems may be affected, complications of the gastrointestinal (GI) tract are frequent and can be life-threatening.^12-16 We performed a thorough review of the literature to familiarize hospitalists with the mechanism of action and uses of ICIs, the clinical presentation of their GI toxicity, and the current recommendations regarding diagnosis and treatment.

CASE PRESENTATION

A 66-year-old man was admitted to our institution with a 1-week history of severe, diffuse abdominal pain and profuse watery diarrhea. He reported having more than 8 watery bowel movements per day and denied fever, recent travel, ill contacts, or ingestion of undercooked food. He had a history of metastatic melanoma and was undergoing treatment with both nivolumab and ipilimumab; the drugs were started 6 weeks prior to presentation. Physical examination revealed a heart rate of 110 beats/minute while supine and 123 beats/minute while standing, blood pressure of 112/69 mm Hg while supine and 92/62 mm Hg while standing, and a temperature of 37.2°C. He was in mild distress and had dry oral mucosa. Abdominal examination revealed hyperactive bowel sounds and mild diffuse abdominal tenderness with no guarding or rebound. His extremities were cool, but peripheral pulses were present. Initial laboratory results included a hemoglobin level of 15.3 g/dL (range 12.0-16.0 mg/dL), white blood cell count 14.2 × 10⁹/L (range 4.5-11.0 × 10⁹/L), and platelet count 236 × 10⁹/L (range 150-400 × 10⁹/L); other test results included a sodium level of 130 mmol/L (range 135-145 mmol/L), potassium 2.3 mmol/L (range 3.5-5.5 mmol/L), serum creatinine 2.2 mg/dL (range 0.8-1.3 mg/dL), blood urea nitrogen 72 mg/dL (range 8-21 mg/dL), and serum venous lactate 5.9 mmol/L (range 0.9-1.7 mmol/L).

MECHANISM OF ACTION AND USES OF ICIS

T-cell lymphocytes play a pivotal role in acquired immunity, but their function requires an appropriate balance between stimulatory and inhibitory signals to prevent autoimmunity.¹⁷ Immune checkpoint molecules are used by the immune system to assist with this balance.¹⁸ Although several of these molecules exist, the cytotoxic T-lymphocyte antigen-4 (CTLA-4) and programmed cell death-1 (PD-1) are among the most widely studied.¹²

Activation or inhibition of T cells depends on the interaction of their receptors with ligands located on the surface of other cells. Both CTLA-4 and PD-1 are receptors located on the surface of T-cell lymphocytes that inhibit the function of T cells after binding with their ligands.^19-21 Cancer cells often use this mechanism to avoid immune recognition and promote their survival.^18,21,22 Importantly, ligands that bind CTLA-4 are expressed by numerous tissues throughout the body, contrary to ligands that bind to PD-1 (PD-L1 or PD-L2), which are more specific to tumor cells (Figure).^21-23 ICIs are monoclonal antibodies that block these pathways and increase T-cell activity.¹⁸

Ipilimumab is a monoclonal antibody directed against CTLA-4.²⁴ After demonstrating survival benefits in patients with unresectable and metastatic melanoma, ipilimumab was the first ICI approved for use by the US Food and Drug Administration (FDA).^1,3 Another monoclonal antibody directed against CTLA-4, tremelimumab, is not currently approved for use by the FDA.

Pembrolizumab and nivolumab are monoclonal antibodies against PD-1. The FDA approved them for the treatment of advanced melanoma in 2014¹⁷ and metastatic NSCLC in 2015.¹² Nivolumab was also approved for the treatment of renal cell carcinoma and for advanced-stage melanoma in combination with ipilimumab.^12,17 Atezolizumab, avelumab, and durvalumab are PD-L1 inhibitors. All 3 are approved by the FDA for treatment of advanced urothelial carcinoma.^25-27 Atezolizumab is also approved for the treatment of metastatic NSCLC,²⁸ and avelumab is approved for treatment of metastatic Merkel cell carcinoma.²⁹ Table 1 summarizes the medications, their target, and FDA-approved indications.^{1,12,17,26,27,29,30}

 

TOXIC PROFILE

Because of the sustained T-cell activation, ICIs have been associated with autoimmune-like toxicities known as immune-related adverse events (irAEs).^19,31 Because the PD-1/PD-L1 pathway is more tumor-specific than the CTLA-4 pathway,^21-23 there is a higher incidence of serious irAEs seen with ipilimumab, reported to be around 27%.^18,22 Furthermore, the risk of developing irAEs is dose-dependent and can increase up to 55% when anti-CTLA-4 are used with other ICIs such as nivolumab.^13,32-34

The skin and GI tract are the most commonly involved organs.^14-16 Skin is affected in 50% of patients receiving ipilimumab and 40% of patients on nivolumab or pembrolizumab, often in the form of a rash or pruritus.^12,35-37 The rash is often described as faintly erythematous, reticular, and maculopapular and typically affects the trunk and extremities.³⁸ Importantly, these events usually occur within the first 2 weeks of treatment, and fewer than 5% are severe.^12,36,39 A higher percentage of severe adverse events occurs in the GI tract, with a reported incidence of 12%.^3,14,36,39

CLINICAL PRESENTATION

Although any portion of the GI tract can be affected by ICIs, the lower GI tract is most commonly involved. Clinical signs include watery diarrhea, colitis, and enteritis.^15,19 Less commonly, the upper GI tract is involved, and clinical manifestations include aphthous ulcers, esophagitis, and gastritis.^40,41 GI symptoms usually begin 6 weeks after the initial dose of ICIs and typically follow cutaneous manifestations.^15,20,36,37 However, they can occur as late as 4 months after the last dose.¹⁹ Watery, nonbloody diarrhea is the most common presentation of GI involvement, occurring in 19% of patients receiving anti-PD-1/PD-L1 and 33% receiving anti-CTLA-4 medications.^19,37 When patients receive both therapies, the incidence rate increases to 44%.³² The clinical severity of diarrhea can be graded on a scale of 1 to 5 according to the Common Terminology Criteria for Adverse Events (CTCAE) (Table 2).⁴² Though most patients have mild disease (grade 1 or 2), close to 3% develop severe diarrhea (grade 3 or higher) with electrolyte disturbances and weight loss.^14,20

Colitis, defined by either the presence of symptoms or radiologic findings suggestive of inflammation, occurs less often than diarrhea alone, with a reported incidence of 2.3%.^37,43 This incidence increases to almost 12% when anti-CTLA-4 and anti-PD-1/PD-L1 are combined.³² Colitis symptoms include abdominal pain (20%), nausea and vomiting (15%), fever (12%), and, less often, bloody diarrhea or rectal bleeding.^19,20 Colitis severity is graded according to the CTCAE (Table 2).⁴² Most patients have mild colitis (grade 1 or 2).¹⁹ The risk for developing severe colitis (grade 3 or higher) is almost 10 times higher with the use of anti-CTLA-4 compared with anti-PD-1/PD-L1 agents.⁴³ Patients with severe disease are at risk of developing life-threatening complications, such as ileus, toxic megacolon, bowel ischemia, necrosis, or even perforation, which has been reported in up to 5% of patients with colitis because of ipilimumab.^13,17

CASE APPROACH STRATEGY

Based on the patient’s symptoms, physical findings, and temporal relationship to ICI therapy, he was believed to have immune-mediated colitis. Stool studies, including those looking for ova and parasites, Clostridium difficile polymerase chain reaction (PCR), and stool cultures were negative.

DIAGNOSIS

In a patient undergoing ICI treatment who has diarrhea, the initial assessment should exclude C. difficile and Salmonella by stool culture, PCR, or pathogenic antigens.¹⁹ Cytomegalovirus reactivation should also be considered. Immune-mediated colitis and infection can coexist; thus, a positive infectious etiology does not rule out the presence of immune colitis or vice versa.⁴⁴ Fecal calprotectin, a marker of neutrophil-associated inflammation, is nonspecific for ICI-induced colitis; however, it may help to distinguish inflammatory from noninflammatory diarrhea.^33,45

No clear guideline exists for the use of abdominal imaging. Some experts suggest using computed tomography in patients with severe, persistent, or progressive symptoms in order to exclude bowel obstruction, toxic megacolon, or perforation.^19,46

In patients with typical symptoms, and after infectious etiologies are ruled out, empiric use of corticosteroids can be initiated without an endoscopic evaluation, which is not necessary to establish a diagnosis and rarely changes management.^12,37,47 In patients with atypical presentations or for whom the diagnosis remains in question, endoscopic evaluation with biopsies may be required. Macroscopic findings may be similar to those seen with inflammatory bowel disease (IBD), including erythema, edema, ulceration, granularity, or loss of vascular pattern. Although immune-mediated colitis affects the descending colon more often than IBD, this feature and any macroscopic findings are insufficient to make this distinction.^20,36 Furthermore, the lack of macroscopic abnormalities does not rule out immune-mediated colitis.²⁰

When endoscopic biopsies are obtained, histologic findings for anti-CTLA-4 medications (eg, ipilimumab) usually follow 3 patterns: neutrophilic infiltrate (46%), lymphocytic infiltrate (15%), and mixed infiltrate (38%).⁴¹ Other findings include crypt abscesses and tissue destruction.²⁰ No biopsy-specific pattern has been described with anti-PD-1/PD-L1 medications, such as nivolumab or pembrolizumab.¹⁸ A normal colonic tissue does not exclude the presence of an irAE, as cases of isolated ileitis⁴⁸ or enteritis⁴⁹ without colitis can also occur.

 

CASE MANAGEMENT STRATEGY

The patient was started on intravenous (IV) methylprednisolone 2 mg/kg twice a day. After 48 hours, he still had more than 7 episodes of diarrhea per day, so he was treated with 1 dose of infliximab 5 mg/kg without stopping corticosteroids. Within 72 hours, the patient’s abdominal pain improved and his diarrhea stopped. He was discharged on an 8-week taper of prednisone starting at 1 mg/kg/day, pneumocystis pneumonia (PCP) prophylaxis was started, and ICI therapy was discontinued indefinitely.

MANAGEMENT OF COLITIS

Several principles should be considered in managing immune-mediated colitis: (1) management for adverse events of anti-CTLA-4 and anti-PD-1/PD-L1 should be the same; (2) though guidelines were made for patients with melanoma, they can be used to treat patients with other types of cancer; and (3) treatment should be started as early as possible, ideally within 5 days of symptom onset, as this hastens clinical improvement and decreases the incidence of complications.²⁰ Treatment is summarized in Table 3.

Management of grade 1 and 2 colitis is mainly supportive, consisting of fluid and electrolyte replacement, the American Dietetic Association colitis diet, and antimotility agents, such as loperamide, oral diphenoxylate hydrochloride, or atropine sulfate.^36,37 Persistent grade 2 symptoms (lasting >3 days), should prompt initiation of 0.5 to 1 mg/kg/day of oral prednisone or an equivalent.¹⁹ If symptoms do not improve with oral corticosteroids, patient hospitalization for IV corticosteroids should be considered.³⁷ Importantly, opioids and antidiarrheals may mask the pain and severity of symptoms and, therefore, should be used cautiously.¹⁹

Patients with grade 3 and 4 colitis (≥7 stools per day, severe abdominal pain, or complications) require the use of systemic corticosteroids at a dose of 1 to 2 mg/kg/day of prednisone or an equivalent.¹⁵ Patients who fail to respond to prednisone alone may benefit from the addition of oral budesonide at a dose of 9 to 12 mg/day.⁵⁰ In severe cases of colitis, hospitalization may be necessary for IV hydration, electrolyte replacement, and IV methylprednisolone at a starting dose of 2 mg/kg twice a day for 1 to 2 days before transitioning to oral corticosteroids.^12,15 Though improvement is usually noted within the first 2 weeks of treatment, prednisone should be slowly tapered over a period of 4 to 8 weeks to ensure complete healing and prevent relapse.^20,36 Patients who receive an equivalent dose of prednisone 20 mg daily during a period of 4 weeks or more should receive PCP prophylaxis.⁵¹ Some patients fail to respond to IV corticosteroids despite adequate dosing. Many of these patients have severe disease, possibly because of delayed recognition and initiation of treatment.¹⁹ As with IBD, the addition of infliximab to corticosteroids at 5 mg/kg as a single dose is usually successful for this population subset.^52-54 Although a response is seen within 1 to 3 days,⁴¹ some patients benefit from an additional dose of infliximab 2 weeks after the initial dose.¹⁹ If sepsis or perforation is suspected at any point, corticosteroids or infliximab should be avoided and antibiotics should be started immediately.^15,19 Patients with a medically unresponsive disease may require partial or complete colectomy.²⁰ The use of prophylactic budesonide to prevent diarrhea or colitis has not been proven effective and should not be used.⁵⁵ Despite complications, mortality from colitis has markedly decreased given the increased awareness of this adverse event, reduction in the time to recognition and treatment, and increased adherence to corticosteroids.¹²

Treating physicians may be delayed in starting appropriate therapy because patients are concerned that using corticosteroids will negatively impact immunotherapy efficacy. Current evidence shows that the use of temporary immunosuppression to treat irAEs does not affect overall survival, efficacy, or time to treatment failure of the ICI.^12,56 Restarting ICI therapy is a complex decision and should always be individualized. In grade 1 and 2 colitis, ICI therapy is typically restarted after symptoms have improved.⁵ In grade 3 and 4 colitis, ICI therapy is often permanently discontinued.²⁰

CONCLUSION

ICIs have not only increased our understanding of the biology of cancer, but they have also improved survival in advanced stages of malignancies like melanoma, NSCLC, and renal cell carcinoma. The expanding use of these medications increases the likelihood that healthcare providers will encounter patients experiencing their adverse events.

Immune-mediated GI adverse events include a wide range of symptoms, from mild diarrhea to severe colitis complicated by perforation and death. Diagnosis requires exclusion of an infectious process. Early recognition and treatment with corticosteroids or another immunosuppressant such as infliximab hastens recovery and decreases complications and mortality. Treatment should be started within 5 days of symptom onset. Corticosteroids should be slowly tapered for no less than 4 weeks to prevent relapse and PCP prophylaxis administered in appropriate patients. Restarting ICI therapy may be considered in cases of mild colitis, but in severe cases, ICI therapy is usually discontinued.

 

Disclosure

Julian Marin-Acevedo, Dana Harris, and M. Caroline Burton have no conflicts of interest or funding sources to declare.

Immune checkpoint inhibitors (ICIs), a form of immunotherapy, have changed the management of cancer since their introduction in 2011.¹ They were initially tested on melanoma.² Their use in the advanced stages of the disease demonstrated a 2-year survival of 18% compared with 5% by using other therapies.³ Similar results were observed in nonsmall cell lung carcinoma (NSCLC); the overall survival benefit was 3 months with the use of ICIs compared with traditional chemotherapy (42% and 24% at 1 year, respectively).⁴ Antitumor activity has also been seen in the treatment of other malignancies, including renal cell carcinoma,⁵ bladder carcinoma,^6,7 head and neck carcinoma,⁸ colorectal cancer,⁹ Hodgkin lymphoma,¹⁰ and, more recently, hepatocellular carcinoma.¹¹ The use of ICIs has also been linked to serious complications.¹² Although the skin, kidneys, lungs, and endocrine and nervous systems may be affected, complications of the gastrointestinal (GI) tract are frequent and can be life-threatening.^12-16 We performed a thorough review of the literature to familiarize hospitalists with the mechanism of action and uses of ICIs, the clinical presentation of their GI toxicity, and the current recommendations regarding diagnosis and treatment.

CASE PRESENTATION

A 66-year-old man was admitted to our institution with a 1-week history of severe, diffuse abdominal pain and profuse watery diarrhea. He reported having more than 8 watery bowel movements per day and denied fever, recent travel, ill contacts, or ingestion of undercooked food. He had a history of metastatic melanoma and was undergoing treatment with both nivolumab and ipilimumab; the drugs were started 6 weeks prior to presentation. Physical examination revealed a heart rate of 110 beats/minute while supine and 123 beats/minute while standing, blood pressure of 112/69 mm Hg while supine and 92/62 mm Hg while standing, and a temperature of 37.2°C. He was in mild distress and had dry oral mucosa. Abdominal examination revealed hyperactive bowel sounds and mild diffuse abdominal tenderness with no guarding or rebound. His extremities were cool, but peripheral pulses were present. Initial laboratory results included a hemoglobin level of 15.3 g/dL (range 12.0-16.0 mg/dL), white blood cell count 14.2 × 10⁹/L (range 4.5-11.0 × 10⁹/L), and platelet count 236 × 10⁹/L (range 150-400 × 10⁹/L); other test results included a sodium level of 130 mmol/L (range 135-145 mmol/L), potassium 2.3 mmol/L (range 3.5-5.5 mmol/L), serum creatinine 2.2 mg/dL (range 0.8-1.3 mg/dL), blood urea nitrogen 72 mg/dL (range 8-21 mg/dL), and serum venous lactate 5.9 mmol/L (range 0.9-1.7 mmol/L).

MECHANISM OF ACTION AND USES OF ICIS

T-cell lymphocytes play a pivotal role in acquired immunity, but their function requires an appropriate balance between stimulatory and inhibitory signals to prevent autoimmunity.¹⁷ Immune checkpoint molecules are used by the immune system to assist with this balance.¹⁸ Although several of these molecules exist, the cytotoxic T-lymphocyte antigen-4 (CTLA-4) and programmed cell death-1 (PD-1) are among the most widely studied.¹²

Activation or inhibition of T cells depends on the interaction of their receptors with ligands located on the surface of other cells. Both CTLA-4 and PD-1 are receptors located on the surface of T-cell lymphocytes that inhibit the function of T cells after binding with their ligands.^19-21 Cancer cells often use this mechanism to avoid immune recognition and promote their survival.^18,21,22 Importantly, ligands that bind CTLA-4 are expressed by numerous tissues throughout the body, contrary to ligands that bind to PD-1 (PD-L1 or PD-L2), which are more specific to tumor cells (Figure).^21-23 ICIs are monoclonal antibodies that block these pathways and increase T-cell activity.¹⁸

Ipilimumab is a monoclonal antibody directed against CTLA-4.²⁴ After demonstrating survival benefits in patients with unresectable and metastatic melanoma, ipilimumab was the first ICI approved for use by the US Food and Drug Administration (FDA).^1,3 Another monoclonal antibody directed against CTLA-4, tremelimumab, is not currently approved for use by the FDA.

Pembrolizumab and nivolumab are monoclonal antibodies against PD-1. The FDA approved them for the treatment of advanced melanoma in 2014¹⁷ and metastatic NSCLC in 2015.¹² Nivolumab was also approved for the treatment of renal cell carcinoma and for advanced-stage melanoma in combination with ipilimumab.^12,17 Atezolizumab, avelumab, and durvalumab are PD-L1 inhibitors. All 3 are approved by the FDA for treatment of advanced urothelial carcinoma.^25-27 Atezolizumab is also approved for the treatment of metastatic NSCLC,²⁸ and avelumab is approved for treatment of metastatic Merkel cell carcinoma.²⁹ Table 1 summarizes the medications, their target, and FDA-approved indications.^{1,12,17,26,27,29,30}

 

TOXIC PROFILE

Because of the sustained T-cell activation, ICIs have been associated with autoimmune-like toxicities known as immune-related adverse events (irAEs).^19,31 Because the PD-1/PD-L1 pathway is more tumor-specific than the CTLA-4 pathway,^21-23 there is a higher incidence of serious irAEs seen with ipilimumab, reported to be around 27%.^18,22 Furthermore, the risk of developing irAEs is dose-dependent and can increase up to 55% when anti-CTLA-4 are used with other ICIs such as nivolumab.^13,32-34

The skin and GI tract are the most commonly involved organs.^14-16 Skin is affected in 50% of patients receiving ipilimumab and 40% of patients on nivolumab or pembrolizumab, often in the form of a rash or pruritus.^12,35-37 The rash is often described as faintly erythematous, reticular, and maculopapular and typically affects the trunk and extremities.³⁸ Importantly, these events usually occur within the first 2 weeks of treatment, and fewer than 5% are severe.^12,36,39 A higher percentage of severe adverse events occurs in the GI tract, with a reported incidence of 12%.^3,14,36,39

CLINICAL PRESENTATION

Although any portion of the GI tract can be affected by ICIs, the lower GI tract is most commonly involved. Clinical signs include watery diarrhea, colitis, and enteritis.^15,19 Less commonly, the upper GI tract is involved, and clinical manifestations include aphthous ulcers, esophagitis, and gastritis.^40,41 GI symptoms usually begin 6 weeks after the initial dose of ICIs and typically follow cutaneous manifestations.^15,20,36,37 However, they can occur as late as 4 months after the last dose.¹⁹ Watery, nonbloody diarrhea is the most common presentation of GI involvement, occurring in 19% of patients receiving anti-PD-1/PD-L1 and 33% receiving anti-CTLA-4 medications.^19,37 When patients receive both therapies, the incidence rate increases to 44%.³² The clinical severity of diarrhea can be graded on a scale of 1 to 5 according to the Common Terminology Criteria for Adverse Events (CTCAE) (Table 2).⁴² Though most patients have mild disease (grade 1 or 2), close to 3% develop severe diarrhea (grade 3 or higher) with electrolyte disturbances and weight loss.^14,20

Colitis, defined by either the presence of symptoms or radiologic findings suggestive of inflammation, occurs less often than diarrhea alone, with a reported incidence of 2.3%.^37,43 This incidence increases to almost 12% when anti-CTLA-4 and anti-PD-1/PD-L1 are combined.³² Colitis symptoms include abdominal pain (20%), nausea and vomiting (15%), fever (12%), and, less often, bloody diarrhea or rectal bleeding.^19,20 Colitis severity is graded according to the CTCAE (Table 2).⁴² Most patients have mild colitis (grade 1 or 2).¹⁹ The risk for developing severe colitis (grade 3 or higher) is almost 10 times higher with the use of anti-CTLA-4 compared with anti-PD-1/PD-L1 agents.⁴³ Patients with severe disease are at risk of developing life-threatening complications, such as ileus, toxic megacolon, bowel ischemia, necrosis, or even perforation, which has been reported in up to 5% of patients with colitis because of ipilimumab.^13,17

CASE APPROACH STRATEGY

Based on the patient’s symptoms, physical findings, and temporal relationship to ICI therapy, he was believed to have immune-mediated colitis. Stool studies, including those looking for ova and parasites, Clostridium difficile polymerase chain reaction (PCR), and stool cultures were negative.

DIAGNOSIS

In a patient undergoing ICI treatment who has diarrhea, the initial assessment should exclude C. difficile and Salmonella by stool culture, PCR, or pathogenic antigens.¹⁹ Cytomegalovirus reactivation should also be considered. Immune-mediated colitis and infection can coexist; thus, a positive infectious etiology does not rule out the presence of immune colitis or vice versa.⁴⁴ Fecal calprotectin, a marker of neutrophil-associated inflammation, is nonspecific for ICI-induced colitis; however, it may help to distinguish inflammatory from noninflammatory diarrhea.^33,45

No clear guideline exists for the use of abdominal imaging. Some experts suggest using computed tomography in patients with severe, persistent, or progressive symptoms in order to exclude bowel obstruction, toxic megacolon, or perforation.^19,46

In patients with typical symptoms, and after infectious etiologies are ruled out, empiric use of corticosteroids can be initiated without an endoscopic evaluation, which is not necessary to establish a diagnosis and rarely changes management.^12,37,47 In patients with atypical presentations or for whom the diagnosis remains in question, endoscopic evaluation with biopsies may be required. Macroscopic findings may be similar to those seen with inflammatory bowel disease (IBD), including erythema, edema, ulceration, granularity, or loss of vascular pattern. Although immune-mediated colitis affects the descending colon more often than IBD, this feature and any macroscopic findings are insufficient to make this distinction.^20,36 Furthermore, the lack of macroscopic abnormalities does not rule out immune-mediated colitis.²⁰

When endoscopic biopsies are obtained, histologic findings for anti-CTLA-4 medications (eg, ipilimumab) usually follow 3 patterns: neutrophilic infiltrate (46%), lymphocytic infiltrate (15%), and mixed infiltrate (38%).⁴¹ Other findings include crypt abscesses and tissue destruction.²⁰ No biopsy-specific pattern has been described with anti-PD-1/PD-L1 medications, such as nivolumab or pembrolizumab.¹⁸ A normal colonic tissue does not exclude the presence of an irAE, as cases of isolated ileitis⁴⁸ or enteritis⁴⁹ without colitis can also occur.

 

CASE MANAGEMENT STRATEGY

The patient was started on intravenous (IV) methylprednisolone 2 mg/kg twice a day. After 48 hours, he still had more than 7 episodes of diarrhea per day, so he was treated with 1 dose of infliximab 5 mg/kg without stopping corticosteroids. Within 72 hours, the patient’s abdominal pain improved and his diarrhea stopped. He was discharged on an 8-week taper of prednisone starting at 1 mg/kg/day, pneumocystis pneumonia (PCP) prophylaxis was started, and ICI therapy was discontinued indefinitely.

MANAGEMENT OF COLITIS

Several principles should be considered in managing immune-mediated colitis: (1) management for adverse events of anti-CTLA-4 and anti-PD-1/PD-L1 should be the same; (2) though guidelines were made for patients with melanoma, they can be used to treat patients with other types of cancer; and (3) treatment should be started as early as possible, ideally within 5 days of symptom onset, as this hastens clinical improvement and decreases the incidence of complications.²⁰ Treatment is summarized in Table 3.

Management of grade 1 and 2 colitis is mainly supportive, consisting of fluid and electrolyte replacement, the American Dietetic Association colitis diet, and antimotility agents, such as loperamide, oral diphenoxylate hydrochloride, or atropine sulfate.^36,37 Persistent grade 2 symptoms (lasting >3 days), should prompt initiation of 0.5 to 1 mg/kg/day of oral prednisone or an equivalent.¹⁹ If symptoms do not improve with oral corticosteroids, patient hospitalization for IV corticosteroids should be considered.³⁷ Importantly, opioids and antidiarrheals may mask the pain and severity of symptoms and, therefore, should be used cautiously.¹⁹

Patients with grade 3 and 4 colitis (≥7 stools per day, severe abdominal pain, or complications) require the use of systemic corticosteroids at a dose of 1 to 2 mg/kg/day of prednisone or an equivalent.¹⁵ Patients who fail to respond to prednisone alone may benefit from the addition of oral budesonide at a dose of 9 to 12 mg/day.⁵⁰ In severe cases of colitis, hospitalization may be necessary for IV hydration, electrolyte replacement, and IV methylprednisolone at a starting dose of 2 mg/kg twice a day for 1 to 2 days before transitioning to oral corticosteroids.^12,15 Though improvement is usually noted within the first 2 weeks of treatment, prednisone should be slowly tapered over a period of 4 to 8 weeks to ensure complete healing and prevent relapse.^20,36 Patients who receive an equivalent dose of prednisone 20 mg daily during a period of 4 weeks or more should receive PCP prophylaxis.⁵¹ Some patients fail to respond to IV corticosteroids despite adequate dosing. Many of these patients have severe disease, possibly because of delayed recognition and initiation of treatment.¹⁹ As with IBD, the addition of infliximab to corticosteroids at 5 mg/kg as a single dose is usually successful for this population subset.^52-54 Although a response is seen within 1 to 3 days,⁴¹ some patients benefit from an additional dose of infliximab 2 weeks after the initial dose.¹⁹ If sepsis or perforation is suspected at any point, corticosteroids or infliximab should be avoided and antibiotics should be started immediately.^15,19 Patients with a medically unresponsive disease may require partial or complete colectomy.²⁰ The use of prophylactic budesonide to prevent diarrhea or colitis has not been proven effective and should not be used.⁵⁵ Despite complications, mortality from colitis has markedly decreased given the increased awareness of this adverse event, reduction in the time to recognition and treatment, and increased adherence to corticosteroids.¹²

Treating physicians may be delayed in starting appropriate therapy because patients are concerned that using corticosteroids will negatively impact immunotherapy efficacy. Current evidence shows that the use of temporary immunosuppression to treat irAEs does not affect overall survival, efficacy, or time to treatment failure of the ICI.^12,56 Restarting ICI therapy is a complex decision and should always be individualized. In grade 1 and 2 colitis, ICI therapy is typically restarted after symptoms have improved.⁵ In grade 3 and 4 colitis, ICI therapy is often permanently discontinued.²⁰

CONCLUSION

ICIs have not only increased our understanding of the biology of cancer, but they have also improved survival in advanced stages of malignancies like melanoma, NSCLC, and renal cell carcinoma. The expanding use of these medications increases the likelihood that healthcare providers will encounter patients experiencing their adverse events.

Immune-mediated GI adverse events include a wide range of symptoms, from mild diarrhea to severe colitis complicated by perforation and death. Diagnosis requires exclusion of an infectious process. Early recognition and treatment with corticosteroids or another immunosuppressant such as infliximab hastens recovery and decreases complications and mortality. Treatment should be started within 5 days of symptom onset. Corticosteroids should be slowly tapered for no less than 4 weeks to prevent relapse and PCP prophylaxis administered in appropriate patients. Restarting ICI therapy may be considered in cases of mild colitis, but in severe cases, ICI therapy is usually discontinued.

 

Disclosure

Julian Marin-Acevedo, Dana Harris, and M. Caroline Burton have no conflicts of interest or funding sources to declare.

The approach to clinical conundrums by an expert clinician is revealed through the presentation of an actual patient’s case in an approach typical of a morning report. Similarly to patient care, sequential pieces of information are provided to the clinician, who is unfamiliar with the case. The focus is on the thought processes of both the clinical team caring for the patient and the discussant. The bolded text represents the patient’s case. Each paragraph that follows represents the discussant’s thoughts.

A 63-year-old man at an inpatient rehabilitation center was transferred to an academic tertiary care center for evaluation of slurred speech and episodic confusion. He was accompanied by his wife, who provided the history. Three weeks earlier, the patient had fallen, sustaining a right femur fracture. He underwent surgery and was discharged to rehabilitation on postoperative day 3. During the second week of rehabilitation, he developed a cough and low-grade fevers, which prompted treatment with cefpodoxime for 5 days for presumed pneumonia. The day after completing antimicrobial therapy, he became confused and began to slur his words.

Confusion is a nonspecific symptom that typically has a diffuse or multifocal localization within the cerebral hemispheres and is unlikely to be caused by a single lesion. Slurred speech may accompany global metabolic dysfunction. However, slurred speech typically localizes to the brainstem, the cerebellum in the posterior fossa, the nuclei, or the course of cranial nerves VII, X, or XII, including where these nerves pass through the subarachnoid space.

It seems this patient’s new neurologic symptoms have some relationship to his fall. Long-bone fractures and altered mental status (AMS) lead to consideration of fat emboli, but this syndrome typically presents in the acute period after the fracture. The patient is at risk for a number of complications, related to recent surgery and hospitalization, that could affect the central nervous system (CNS), including systemic infection (possibly with associated meningeal involvement) and venous thromboembolism with concomitant stroke by paradoxical emboli. The episodic nature of the confusion leads to consideration of seizures from structural lesions in the brain. Finally, the circumstances of the fall itself should be explored to determine whether an underlying neurologic dysfunction led to imbalance and gait difficulty.

Over the next 3 days at the inpatient rehabilitation center, the patient’s slurred speech became unintelligible, and he experienced intermittent disorientation to person, place, and time. There was no concomitant fever, dizziness, headache, neck pain, weakness, dyspnea, diarrhea, dysuria, or change in hearing or vision.

Progressive dysarthria argues for an expanding lesion in the posterior fossa, worsening metabolic disturbance, or a problem affecting the cranial nerves (eg, Guillain-Barré syndrome) or neuromuscular junctions (eg, myasthenia gravis). Lack of headache makes a CNS localization less likely, though disorientation must localize to the brain itself. The transient nature of the AMS could signal an ictal phenomenon or a fluctuating toxic or metabolic condition, such as hyperammonemia, drug reaction, or healthcare–acquired delirium.

His past medical history included end-stage liver disease secondary to nonalcoholic steatohepatitis status post transjugular intrahepatic portosystemic shunt (TIPS) procedure three years prior, hepatic encephalopathy, diabetes mellitus type 2, hypertension, previous melanoma excision on his back, and recurrent Clostridium difficile colitis. Two years prior to admission he had been started on an indefinite course of metronidazole 500 mg twice daily without any recurrence. The patient’s other medications were aspirin, furosemide, insulin, lactulose, mirtazapine, pantoprazole, propranolol, spironolactone, and zinc. At the rehabilitation center, he was prescribed oral oxycodone 5 mg as needed every 4 hours for pain. He denied use of tobacco, alcohol, and recreational drugs. He previously worked as a funeral home director and embalmer.

Hyperammonemia and hepatic encephalopathy can present with a fluctuating mental state that often correlates to dietary protein intake or the frequency of bowel movements; the previous TIPS history places the patient at further risk. Use of oxycodone or another narcotic commonly leads to confusion, , especially in patients who are older, have preexisting cognitive decline, or have concomitant medical comorbidities. Mirtazapine and propranolol have been associated more rarely with encephalopathy, and therefore a careful history of adherence, drug interactions, and appropriate dosing should be obtained. Metronidazole is most often associated neurologically with a peripheral neuropathy; however, it is increasingly recognized that some patients can develop a CNS syndrome that features an AMS, which can be severe and accompanied by ataxia, dysarthria, and characteristic brain magnetic resonance imaging (MRI) findings, including hyperintensity surrounding the fourth ventricle on T₂-weighted images.

 

Embalming fluid has a high concentration of formaldehyde, and a recent epidemiologic study suggested a link between formaldehyde exposure and increased risk for amyotrophic lateral sclerosis (ALS). ALS uncommonly presents with isolated dysarthria, but its bulbar form can, usually over a much longer course than is demonstrated here. Finally, the patient’s history of melanoma places him at risk for stroke from hypercoagulability as well as potential brain metastases or carcinomatous meningitis.

Evaluation was initiated at the rehabilitation facility at the onset of the patient’s slurred speech and confusion. Physical examination were negative for focal neurologic deficits, asterixis, and jaundice. Ammonia level was 41 µmol/L (reference range, 11-35 µmol/L). Noncontrast computed tomography (CT) of the head showed no signs of acute infarct or hemorrhage. Symptoms were attributed to hepatic encephalopathy; lactulose was up-titrated to ensure 2 or 3 bowel movements per day, and rifaximin was started.

Hyperammonemia is a cause of non-inflammatory relapsing encephalopathy, but an elevated level is neither a sensitive nor specific indicator of hepatic encephalopathy. Levels of ammonia can fluctuate widely during the day based on the frequency of bowel movements as well as dietary protein intake. In addition, proper handling of samples with prompt delivery to the laboratory is essential to minimize errors.

The ammonia level of 41 µmol/L discovered here is only modestly elevated, but given the patient’s history of TIPS as well as the clinical picture, it is reasonable to aggressively treat hepatic encephalopathy with lactulose to reduce ammonia levels. If he does not improve, an MRI of the brain to exclude a structural lesion and spinal fluid examination looking for inflammatory or infectious conditions would be important next steps. Although CT excludes a large hemorrhage or mass, this screening examination does not visualize many of the findings of the metabolic etiology and the other etiologies under consideration here.

Despite 3 days of therapy for presumed hepatic encephalopathy, the patient’s slurred speech worsened, and he was transferred to an academic tertiary care center for further evaluation. On admission, his temperature was 36.9°C, heart rate was 80 beats per minute, blood pressure was 139/67 mm Hg, respiratory rate was 10 breaths per minute, and oxygen saturation was 99% on room air. He was alert, awake, and oriented to person, place, and time. He was not jaundiced. He exhibited a moderate dysarthria characterized by monotone speech, decreased volume, decreased breath support, and a hoarse vocal quality with intact language function. Motor control of the lips, tongue, and mandible were normal. Motor strength was 5/5 bilaterally in the upper and lower extremities with the exception of right hip flexion, which was 4/5. The patient exhibited mild bilateral dysmetria on finger-to-nose examination, consistent with appendicular ataxia of the upper extremities. Reflexes were depressed throughout, and there was no asterixis. He had 2+ pulses in all extremities and 1+ pitting edema of the right lower extremity to the mid leg. Pulmonary examination revealed inspiratory crackles at the left base. The rest of the examination findings were normal.

The patient’s altered mental state appears to have resolved, and the neurological examination is now mainly characterized by signs that point to the cerebellum. The description of monotone speech typically refers to loss of prosody, the variable stress or intonation of speech, which is characteristic of a cerebellar speech pattern. The hoarseness should be explored to determine if it is a feature of the patient’s speech or is a separate process. Hoarseness may involve the vocal cord and therefore, potentially, cranial nerve X or its nuclei in the brainstem. The appendicular ataxia of the limbs points definitively to the cerebellar hemispheres or their pathways through the brainstem.

Unilateral lower extremity edema, especially in the context of a recent fracture, raises the possibility of deep vein thrombosis. If this patient has a right-to-left intracardiac or intrapulmonary shunt, embolization could lead to an ischemic stroke of the brainstem or cerebellum, potentially causing dysarthria.

Laboratory evaluation revealed hemoglobin level of 10.9 g/dL, white blood cell count of 5.3 × 10 ⁹ /L, platelet count of 169 × 10 ⁹ /L, glucose level of 177 mg/dL, corrected calcium level of 9.0 mg/dL, sodium level of 135 mmol/L, bicarbonate level of 30 mmol/L, creatinine level of 0.9 mg/dL, total bilirubin level of 1.3 mg/dL, direct bilirubin level of 0.4 mg/dL, alkaline phosphatase level of 503 U/L, alanine aminotransferase level of 12 U/L, aspartate aminotransferase level of 33 U/L, ammonia level of 49 µmol/L (range, 0-30 µ mol/L), international normalized ratio of 1.2, and troponin level of <0.01 ng/mL. Electrocardiogram showed normal sinus rhythm.

 

Some patients with bacterial meningitis do not have a leukocytosis, but patients with meningitis caused by seeding from a systemic infection nearly always do. In this patient’s case, lack of a leukocytosis makes bacterial meningitis very unlikely. The elevated alkaline phosphatase level is expected, as this level peaks about 3 weeks after a long-bone fracture and returns to normal over a few months.

Non-contrast CT scan of the head performed on admission demonstrated no large vessel cortical-based infarct, intracranial hemorrhage, hydrocephalus, mass effect, midline shift, or extra-axial fluid. There was mild cortical atrophy as well as very mild periventricular white matter hypodensity.

The atrophy and mild white-matter hypodensities seen on repeat noncontrast CT are nonspecific for any particular entity in this patient’s age group. MRI is more effective in evaluating toxic encephalopathies, including metronidazole toxicity or Wernicke encephalopathy, and in characterizing small infarcts or inflammatory conditions of the brainstem and cerebellum, which are poorly evaluated by CT due to the bone surrounded space of the posterior fossa. An urgent lumbar puncture is not necessary due to the slow pace of illness, lack of fever, nuchal rigidity, or serum elevated white blood cell count. Rather, performing MRI should be prioritized. If MRI is nondiagnostic, then spinal fluid should be evaluated for evidence of an infectious, autoimmune, paraneoplastic, or neoplastic process.

MRI was subsequently performed. It showed symmetric abnormal T₂ hyperintensities involving dentate nuclei (Figure 1), left inferior olivary nuclei (Figure 2), restiform bodies, pontine tegmentum, superior cerebellar peduncles, oculomotor nuclei, and subthalamic nuclei. The most prominent hyperintensity was in the dentate nuclei.

The clinical and radiographic features confirm a diagnosis of metronidazole-associated CNS neurotoxicity. The reason for the predilection for edema in these specific areas of the brainstem and midline cerebellum is unclear but likely is related to selective neuronal vulnerability in these structures. The treatment is to stop metronidazole. In addition, the fluctuating mental status should be evaluated with electroencephalogram to ensure concomitant seizures are not occurring.

These MRI findings were consistent with metronidazole toxicity. Metronidazole was discontinued, and 2 days later the patient’s speech improved. Two weeks after medication discontinuation, his speech was normal. There were no more episodes of confusion.

DISCUSSION

Metronidazole was originally developed in France during the 1950s as an anti-parasitic medication to treat trichomonas infections. In 1962, its antibacterial properties were discovered after a patient with bacterial gingivitis improved while taking metronidazole for treatment of Trichomonas vaginalis.¹ Since that time metronidazole has become a first-line treatment for anaerobic bacteria and is now recommended by the Infectious Diseases Society of America² and the American College of Gastroenterology³ as a first-line therapy for mild and moderate C difficile infections.

Common side effects of metronidazole are nausea, vomiting, decreased appetite, diarrhea, headaches, peripheral neuropathy, and metallic taste; less common is CNS toxicity. Although the incidence of CNS toxicity is unknown, a systematic review of the literature found 64 cases reported between 1965 and 2011.⁴ CNS toxicity most often occurs between the fifth and sixth decades of life, and about two thirds of the people affected are men.⁴ CNS adverse effects characteristically fall into 4 categories: cerebellar dysfunction (eg, ataxia, dysarthria, dysmetria, nystagmus; 75%), AMS (33%), seizures (13%), and a combination of the first 3 categories.⁴

The exact mechanism of metronidazole CNS toxicity is unknown, but vasogenic or cytotoxic edema may be involved.^5,6 Other potential etiologies are neural protein inhibition, reversible mitochondrial dysfunction, and modifications of the inhibitory neurotransmitter gamma-aminobutyric acid receptor in the cerebellum.^7,8 There is no known genetic predisposition. Although the risk for CNS toxicity traditionally is thought to correlate with therapy duration and cumulative dose,^7,9 in 2011 a systemic review found no significant correlation.⁴ In fact, 26% of patients with CNS toxicity were treated with metronidazole for less than 1 week at time of diagnosis.⁴

Brain CT is typically normal. On brain MRI, lesions most commonly appear as bilateral symmetric T₂ hyperintensities, most often in the cerebellar dentate nuclei (85%) and less often in the midbrain (55%), the splenium of the corpus callosum (50%), the pons (35%), and the medulla (30%).^4,10 Radiographic changes have been noted as early as 3 days after symptom onset. Based on damage severity and area affected (white or gray matter), vasogenic edema and cytotoxic edema may in combination be contributing to MRI abnormalities.^6,10 Hyperintensities of the bilateral dentate nuclei can help in distinguishing metronidazole-induced encephalopathy from other potential disease processes, such as Wernicke encephalopathy.¹⁰

The prognosis for patients with metronidazole-induced neurotoxicity is favorable if metronidazole is discontinued. Approximately two-thirds of patients will have complete resolution of symptoms, which is more commonly observed when patients present with seizures or altered mental status. Approximately one-third will show partial improvement, particularly if the symptoms are due to cerebellar dysfunction. It is rare to experience permanent damage or death.⁴ Neurologic recovery usually begins within a week after medication discontinuation but may take months for complete recovery to occur.^6,8,9,11 Follow-up imaging typically shows reversal of the original lesions, but this does not always correlate with symptom improvement.^4,10

Despite its frequent use and long history, metronidazole can have potentially severe toxicity. When patients who are taking this medication present with new signs and symptoms of CNS dysfunction, hospitalists should include metronidazole CNS toxicity in the differential diagnosis and, if they suspect toxicity, have a brain MRI performed. Hospitalists often prescribe metronidazole because of the increasing number of patients being discharged from acute-care hospitals with a diagnosis of C difficile colitis.¹² Brain MRI remains the imaging modality of choice for diagnosis. Discontinuation of metronidazole is usually salutary in reversing symptoms. Being keenly aware of this toxicity will help clinicians avoid being rendered speechless by a patient rendered speechless.

 

TEACHING POINTS

• CNS toxicity is a rare but potentially devastating side effect of metronidazole exposure.

• Metronidazole CNS adverse effects characteristically fall under 4 categories:

○ Cerebellar dysfunction, such as ataxia, dysarthria, dysmetria, or nystagmus (75%).

○ AMS (33%).

○ Seizures (13%).

○ A combination of the first 3 categories.

• Typically lesions indicating metronidazole toxicity on brain MRI are bilateral symmetric hyperintensities on T2-weighted imaging in the cerebellar dentate nuclei, corpus callosum, midbrain, pons, or medulla.
• Treatment of CNS toxicity is metronidazole discontinuation, which results in a high rate of symptom resolution.

Disclosure

Nothing to report.

 

The approach to clinical conundrums by an expert clinician is revealed through the presentation of an actual patient’s case in an approach typical of a morning report. Similarly to patient care, sequential pieces of information are provided to the clinician, who is unfamiliar with the case. The focus is on the thought processes of both the clinical team caring for the patient and the discussant. The bolded text represents the patient’s case. Each paragraph that follows represents the discussant’s thoughts.

A 63-year-old man at an inpatient rehabilitation center was transferred to an academic tertiary care center for evaluation of slurred speech and episodic confusion. He was accompanied by his wife, who provided the history. Three weeks earlier, the patient had fallen, sustaining a right femur fracture. He underwent surgery and was discharged to rehabilitation on postoperative day 3. During the second week of rehabilitation, he developed a cough and low-grade fevers, which prompted treatment with cefpodoxime for 5 days for presumed pneumonia. The day after completing antimicrobial therapy, he became confused and began to slur his words.

Confusion is a nonspecific symptom that typically has a diffuse or multifocal localization within the cerebral hemispheres and is unlikely to be caused by a single lesion. Slurred speech may accompany global metabolic dysfunction. However, slurred speech typically localizes to the brainstem, the cerebellum in the posterior fossa, the nuclei, or the course of cranial nerves VII, X, or XII, including where these nerves pass through the subarachnoid space.

It seems this patient’s new neurologic symptoms have some relationship to his fall. Long-bone fractures and altered mental status (AMS) lead to consideration of fat emboli, but this syndrome typically presents in the acute period after the fracture. The patient is at risk for a number of complications, related to recent surgery and hospitalization, that could affect the central nervous system (CNS), including systemic infection (possibly with associated meningeal involvement) and venous thromboembolism with concomitant stroke by paradoxical emboli. The episodic nature of the confusion leads to consideration of seizures from structural lesions in the brain. Finally, the circumstances of the fall itself should be explored to determine whether an underlying neurologic dysfunction led to imbalance and gait difficulty.

Over the next 3 days at the inpatient rehabilitation center, the patient’s slurred speech became unintelligible, and he experienced intermittent disorientation to person, place, and time. There was no concomitant fever, dizziness, headache, neck pain, weakness, dyspnea, diarrhea, dysuria, or change in hearing or vision.

Progressive dysarthria argues for an expanding lesion in the posterior fossa, worsening metabolic disturbance, or a problem affecting the cranial nerves (eg, Guillain-Barré syndrome) or neuromuscular junctions (eg, myasthenia gravis). Lack of headache makes a CNS localization less likely, though disorientation must localize to the brain itself. The transient nature of the AMS could signal an ictal phenomenon or a fluctuating toxic or metabolic condition, such as hyperammonemia, drug reaction, or healthcare–acquired delirium.

His past medical history included end-stage liver disease secondary to nonalcoholic steatohepatitis status post transjugular intrahepatic portosystemic shunt (TIPS) procedure three years prior, hepatic encephalopathy, diabetes mellitus type 2, hypertension, previous melanoma excision on his back, and recurrent Clostridium difficile colitis. Two years prior to admission he had been started on an indefinite course of metronidazole 500 mg twice daily without any recurrence. The patient’s other medications were aspirin, furosemide, insulin, lactulose, mirtazapine, pantoprazole, propranolol, spironolactone, and zinc. At the rehabilitation center, he was prescribed oral oxycodone 5 mg as needed every 4 hours for pain. He denied use of tobacco, alcohol, and recreational drugs. He previously worked as a funeral home director and embalmer.

Hyperammonemia and hepatic encephalopathy can present with a fluctuating mental state that often correlates to dietary protein intake or the frequency of bowel movements; the previous TIPS history places the patient at further risk. Use of oxycodone or another narcotic commonly leads to confusion, , especially in patients who are older, have preexisting cognitive decline, or have concomitant medical comorbidities. Mirtazapine and propranolol have been associated more rarely with encephalopathy, and therefore a careful history of adherence, drug interactions, and appropriate dosing should be obtained. Metronidazole is most often associated neurologically with a peripheral neuropathy; however, it is increasingly recognized that some patients can develop a CNS syndrome that features an AMS, which can be severe and accompanied by ataxia, dysarthria, and characteristic brain magnetic resonance imaging (MRI) findings, including hyperintensity surrounding the fourth ventricle on T₂-weighted images.

 

Embalming fluid has a high concentration of formaldehyde, and a recent epidemiologic study suggested a link between formaldehyde exposure and increased risk for amyotrophic lateral sclerosis (ALS). ALS uncommonly presents with isolated dysarthria, but its bulbar form can, usually over a much longer course than is demonstrated here. Finally, the patient’s history of melanoma places him at risk for stroke from hypercoagulability as well as potential brain metastases or carcinomatous meningitis.

Evaluation was initiated at the rehabilitation facility at the onset of the patient’s slurred speech and confusion. Physical examination were negative for focal neurologic deficits, asterixis, and jaundice. Ammonia level was 41 µmol/L (reference range, 11-35 µmol/L). Noncontrast computed tomography (CT) of the head showed no signs of acute infarct or hemorrhage. Symptoms were attributed to hepatic encephalopathy; lactulose was up-titrated to ensure 2 or 3 bowel movements per day, and rifaximin was started.

Hyperammonemia is a cause of non-inflammatory relapsing encephalopathy, but an elevated level is neither a sensitive nor specific indicator of hepatic encephalopathy. Levels of ammonia can fluctuate widely during the day based on the frequency of bowel movements as well as dietary protein intake. In addition, proper handling of samples with prompt delivery to the laboratory is essential to minimize errors.

The ammonia level of 41 µmol/L discovered here is only modestly elevated, but given the patient’s history of TIPS as well as the clinical picture, it is reasonable to aggressively treat hepatic encephalopathy with lactulose to reduce ammonia levels. If he does not improve, an MRI of the brain to exclude a structural lesion and spinal fluid examination looking for inflammatory or infectious conditions would be important next steps. Although CT excludes a large hemorrhage or mass, this screening examination does not visualize many of the findings of the metabolic etiology and the other etiologies under consideration here.

Despite 3 days of therapy for presumed hepatic encephalopathy, the patient’s slurred speech worsened, and he was transferred to an academic tertiary care center for further evaluation. On admission, his temperature was 36.9°C, heart rate was 80 beats per minute, blood pressure was 139/67 mm Hg, respiratory rate was 10 breaths per minute, and oxygen saturation was 99% on room air. He was alert, awake, and oriented to person, place, and time. He was not jaundiced. He exhibited a moderate dysarthria characterized by monotone speech, decreased volume, decreased breath support, and a hoarse vocal quality with intact language function. Motor control of the lips, tongue, and mandible were normal. Motor strength was 5/5 bilaterally in the upper and lower extremities with the exception of right hip flexion, which was 4/5. The patient exhibited mild bilateral dysmetria on finger-to-nose examination, consistent with appendicular ataxia of the upper extremities. Reflexes were depressed throughout, and there was no asterixis. He had 2+ pulses in all extremities and 1+ pitting edema of the right lower extremity to the mid leg. Pulmonary examination revealed inspiratory crackles at the left base. The rest of the examination findings were normal.

The patient’s altered mental state appears to have resolved, and the neurological examination is now mainly characterized by signs that point to the cerebellum. The description of monotone speech typically refers to loss of prosody, the variable stress or intonation of speech, which is characteristic of a cerebellar speech pattern. The hoarseness should be explored to determine if it is a feature of the patient’s speech or is a separate process. Hoarseness may involve the vocal cord and therefore, potentially, cranial nerve X or its nuclei in the brainstem. The appendicular ataxia of the limbs points definitively to the cerebellar hemispheres or their pathways through the brainstem.

Unilateral lower extremity edema, especially in the context of a recent fracture, raises the possibility of deep vein thrombosis. If this patient has a right-to-left intracardiac or intrapulmonary shunt, embolization could lead to an ischemic stroke of the brainstem or cerebellum, potentially causing dysarthria.

Laboratory evaluation revealed hemoglobin level of 10.9 g/dL, white blood cell count of 5.3 × 10 ⁹ /L, platelet count of 169 × 10 ⁹ /L, glucose level of 177 mg/dL, corrected calcium level of 9.0 mg/dL, sodium level of 135 mmol/L, bicarbonate level of 30 mmol/L, creatinine level of 0.9 mg/dL, total bilirubin level of 1.3 mg/dL, direct bilirubin level of 0.4 mg/dL, alkaline phosphatase level of 503 U/L, alanine aminotransferase level of 12 U/L, aspartate aminotransferase level of 33 U/L, ammonia level of 49 µmol/L (range, 0-30 µ mol/L), international normalized ratio of 1.2, and troponin level of <0.01 ng/mL. Electrocardiogram showed normal sinus rhythm.

 

Some patients with bacterial meningitis do not have a leukocytosis, but patients with meningitis caused by seeding from a systemic infection nearly always do. In this patient’s case, lack of a leukocytosis makes bacterial meningitis very unlikely. The elevated alkaline phosphatase level is expected, as this level peaks about 3 weeks after a long-bone fracture and returns to normal over a few months.

Non-contrast CT scan of the head performed on admission demonstrated no large vessel cortical-based infarct, intracranial hemorrhage, hydrocephalus, mass effect, midline shift, or extra-axial fluid. There was mild cortical atrophy as well as very mild periventricular white matter hypodensity.

The atrophy and mild white-matter hypodensities seen on repeat noncontrast CT are nonspecific for any particular entity in this patient’s age group. MRI is more effective in evaluating toxic encephalopathies, including metronidazole toxicity or Wernicke encephalopathy, and in characterizing small infarcts or inflammatory conditions of the brainstem and cerebellum, which are poorly evaluated by CT due to the bone surrounded space of the posterior fossa. An urgent lumbar puncture is not necessary due to the slow pace of illness, lack of fever, nuchal rigidity, or serum elevated white blood cell count. Rather, performing MRI should be prioritized. If MRI is nondiagnostic, then spinal fluid should be evaluated for evidence of an infectious, autoimmune, paraneoplastic, or neoplastic process.

MRI was subsequently performed. It showed symmetric abnormal T₂ hyperintensities involving dentate nuclei (Figure 1), left inferior olivary nuclei (Figure 2), restiform bodies, pontine tegmentum, superior cerebellar peduncles, oculomotor nuclei, and subthalamic nuclei. The most prominent hyperintensity was in the dentate nuclei.

The clinical and radiographic features confirm a diagnosis of metronidazole-associated CNS neurotoxicity. The reason for the predilection for edema in these specific areas of the brainstem and midline cerebellum is unclear but likely is related to selective neuronal vulnerability in these structures. The treatment is to stop metronidazole. In addition, the fluctuating mental status should be evaluated with electroencephalogram to ensure concomitant seizures are not occurring.

These MRI findings were consistent with metronidazole toxicity. Metronidazole was discontinued, and 2 days later the patient’s speech improved. Two weeks after medication discontinuation, his speech was normal. There were no more episodes of confusion.

DISCUSSION

Metronidazole was originally developed in France during the 1950s as an anti-parasitic medication to treat trichomonas infections. In 1962, its antibacterial properties were discovered after a patient with bacterial gingivitis improved while taking metronidazole for treatment of Trichomonas vaginalis.¹ Since that time metronidazole has become a first-line treatment for anaerobic bacteria and is now recommended by the Infectious Diseases Society of America² and the American College of Gastroenterology³ as a first-line therapy for mild and moderate C difficile infections.

Common side effects of metronidazole are nausea, vomiting, decreased appetite, diarrhea, headaches, peripheral neuropathy, and metallic taste; less common is CNS toxicity. Although the incidence of CNS toxicity is unknown, a systematic review of the literature found 64 cases reported between 1965 and 2011.⁴ CNS toxicity most often occurs between the fifth and sixth decades of life, and about two thirds of the people affected are men.⁴ CNS adverse effects characteristically fall into 4 categories: cerebellar dysfunction (eg, ataxia, dysarthria, dysmetria, nystagmus; 75%), AMS (33%), seizures (13%), and a combination of the first 3 categories.⁴

The exact mechanism of metronidazole CNS toxicity is unknown, but vasogenic or cytotoxic edema may be involved.^5,6 Other potential etiologies are neural protein inhibition, reversible mitochondrial dysfunction, and modifications of the inhibitory neurotransmitter gamma-aminobutyric acid receptor in the cerebellum.^7,8 There is no known genetic predisposition. Although the risk for CNS toxicity traditionally is thought to correlate with therapy duration and cumulative dose,^7,9 in 2011 a systemic review found no significant correlation.⁴ In fact, 26% of patients with CNS toxicity were treated with metronidazole for less than 1 week at time of diagnosis.⁴

Brain CT is typically normal. On brain MRI, lesions most commonly appear as bilateral symmetric T₂ hyperintensities, most often in the cerebellar dentate nuclei (85%) and less often in the midbrain (55%), the splenium of the corpus callosum (50%), the pons (35%), and the medulla (30%).^4,10 Radiographic changes have been noted as early as 3 days after symptom onset. Based on damage severity and area affected (white or gray matter), vasogenic edema and cytotoxic edema may in combination be contributing to MRI abnormalities.^6,10 Hyperintensities of the bilateral dentate nuclei can help in distinguishing metronidazole-induced encephalopathy from other potential disease processes, such as Wernicke encephalopathy.¹⁰

The prognosis for patients with metronidazole-induced neurotoxicity is favorable if metronidazole is discontinued. Approximately two-thirds of patients will have complete resolution of symptoms, which is more commonly observed when patients present with seizures or altered mental status. Approximately one-third will show partial improvement, particularly if the symptoms are due to cerebellar dysfunction. It is rare to experience permanent damage or death.⁴ Neurologic recovery usually begins within a week after medication discontinuation but may take months for complete recovery to occur.^6,8,9,11 Follow-up imaging typically shows reversal of the original lesions, but this does not always correlate with symptom improvement.^4,10

Despite its frequent use and long history, metronidazole can have potentially severe toxicity. When patients who are taking this medication present with new signs and symptoms of CNS dysfunction, hospitalists should include metronidazole CNS toxicity in the differential diagnosis and, if they suspect toxicity, have a brain MRI performed. Hospitalists often prescribe metronidazole because of the increasing number of patients being discharged from acute-care hospitals with a diagnosis of C difficile colitis.¹² Brain MRI remains the imaging modality of choice for diagnosis. Discontinuation of metronidazole is usually salutary in reversing symptoms. Being keenly aware of this toxicity will help clinicians avoid being rendered speechless by a patient rendered speechless.

 

TEACHING POINTS

• CNS toxicity is a rare but potentially devastating side effect of metronidazole exposure.

• Metronidazole CNS adverse effects characteristically fall under 4 categories:

○ Cerebellar dysfunction, such as ataxia, dysarthria, dysmetria, or nystagmus (75%).

○ AMS (33%).

○ Seizures (13%).

○ A combination of the first 3 categories.

• Typically lesions indicating metronidazole toxicity on brain MRI are bilateral symmetric hyperintensities on T2-weighted imaging in the cerebellar dentate nuclei, corpus callosum, midbrain, pons, or medulla.
• Treatment of CNS toxicity is metronidazole discontinuation, which results in a high rate of symptom resolution.

Disclosure

Nothing to report.

 

The approach to clinical conundrums by an expert clinician is revealed through the presentation of an actual patient’s case in an approach typical of a morning report. Similarly to patient care, sequential pieces of information are provided to the clinician, who is unfamiliar with the case. The focus is on the thought processes of both the clinical team caring for the patient and the discussant. The bolded text represents the patient’s case. Each paragraph that follows represents the discussant’s thoughts.

A 63-year-old man at an inpatient rehabilitation center was transferred to an academic tertiary care center for evaluation of slurred speech and episodic confusion. He was accompanied by his wife, who provided the history. Three weeks earlier, the patient had fallen, sustaining a right femur fracture. He underwent surgery and was discharged to rehabilitation on postoperative day 3. During the second week of rehabilitation, he developed a cough and low-grade fevers, which prompted treatment with cefpodoxime for 5 days for presumed pneumonia. The day after completing antimicrobial therapy, he became confused and began to slur his words.

Confusion is a nonspecific symptom that typically has a diffuse or multifocal localization within the cerebral hemispheres and is unlikely to be caused by a single lesion. Slurred speech may accompany global metabolic dysfunction. However, slurred speech typically localizes to the brainstem, the cerebellum in the posterior fossa, the nuclei, or the course of cranial nerves VII, X, or XII, including where these nerves pass through the subarachnoid space.

It seems this patient’s new neurologic symptoms have some relationship to his fall. Long-bone fractures and altered mental status (AMS) lead to consideration of fat emboli, but this syndrome typically presents in the acute period after the fracture. The patient is at risk for a number of complications, related to recent surgery and hospitalization, that could affect the central nervous system (CNS), including systemic infection (possibly with associated meningeal involvement) and venous thromboembolism with concomitant stroke by paradoxical emboli. The episodic nature of the confusion leads to consideration of seizures from structural lesions in the brain. Finally, the circumstances of the fall itself should be explored to determine whether an underlying neurologic dysfunction led to imbalance and gait difficulty.

Over the next 3 days at the inpatient rehabilitation center, the patient’s slurred speech became unintelligible, and he experienced intermittent disorientation to person, place, and time. There was no concomitant fever, dizziness, headache, neck pain, weakness, dyspnea, diarrhea, dysuria, or change in hearing or vision.

Progressive dysarthria argues for an expanding lesion in the posterior fossa, worsening metabolic disturbance, or a problem affecting the cranial nerves (eg, Guillain-Barré syndrome) or neuromuscular junctions (eg, myasthenia gravis). Lack of headache makes a CNS localization less likely, though disorientation must localize to the brain itself. The transient nature of the AMS could signal an ictal phenomenon or a fluctuating toxic or metabolic condition, such as hyperammonemia, drug reaction, or healthcare–acquired delirium.

His past medical history included end-stage liver disease secondary to nonalcoholic steatohepatitis status post transjugular intrahepatic portosystemic shunt (TIPS) procedure three years prior, hepatic encephalopathy, diabetes mellitus type 2, hypertension, previous melanoma excision on his back, and recurrent Clostridium difficile colitis. Two years prior to admission he had been started on an indefinite course of metronidazole 500 mg twice daily without any recurrence. The patient’s other medications were aspirin, furosemide, insulin, lactulose, mirtazapine, pantoprazole, propranolol, spironolactone, and zinc. At the rehabilitation center, he was prescribed oral oxycodone 5 mg as needed every 4 hours for pain. He denied use of tobacco, alcohol, and recreational drugs. He previously worked as a funeral home director and embalmer.

Hyperammonemia and hepatic encephalopathy can present with a fluctuating mental state that often correlates to dietary protein intake or the frequency of bowel movements; the previous TIPS history places the patient at further risk. Use of oxycodone or another narcotic commonly leads to confusion, , especially in patients who are older, have preexisting cognitive decline, or have concomitant medical comorbidities. Mirtazapine and propranolol have been associated more rarely with encephalopathy, and therefore a careful history of adherence, drug interactions, and appropriate dosing should be obtained. Metronidazole is most often associated neurologically with a peripheral neuropathy; however, it is increasingly recognized that some patients can develop a CNS syndrome that features an AMS, which can be severe and accompanied by ataxia, dysarthria, and characteristic brain magnetic resonance imaging (MRI) findings, including hyperintensity surrounding the fourth ventricle on T₂-weighted images.

 

Embalming fluid has a high concentration of formaldehyde, and a recent epidemiologic study suggested a link between formaldehyde exposure and increased risk for amyotrophic lateral sclerosis (ALS). ALS uncommonly presents with isolated dysarthria, but its bulbar form can, usually over a much longer course than is demonstrated here. Finally, the patient’s history of melanoma places him at risk for stroke from hypercoagulability as well as potential brain metastases or carcinomatous meningitis.

Evaluation was initiated at the rehabilitation facility at the onset of the patient’s slurred speech and confusion. Physical examination were negative for focal neurologic deficits, asterixis, and jaundice. Ammonia level was 41 µmol/L (reference range, 11-35 µmol/L). Noncontrast computed tomography (CT) of the head showed no signs of acute infarct or hemorrhage. Symptoms were attributed to hepatic encephalopathy; lactulose was up-titrated to ensure 2 or 3 bowel movements per day, and rifaximin was started.

Hyperammonemia is a cause of non-inflammatory relapsing encephalopathy, but an elevated level is neither a sensitive nor specific indicator of hepatic encephalopathy. Levels of ammonia can fluctuate widely during the day based on the frequency of bowel movements as well as dietary protein intake. In addition, proper handling of samples with prompt delivery to the laboratory is essential to minimize errors.

The ammonia level of 41 µmol/L discovered here is only modestly elevated, but given the patient’s history of TIPS as well as the clinical picture, it is reasonable to aggressively treat hepatic encephalopathy with lactulose to reduce ammonia levels. If he does not improve, an MRI of the brain to exclude a structural lesion and spinal fluid examination looking for inflammatory or infectious conditions would be important next steps. Although CT excludes a large hemorrhage or mass, this screening examination does not visualize many of the findings of the metabolic etiology and the other etiologies under consideration here.

Despite 3 days of therapy for presumed hepatic encephalopathy, the patient’s slurred speech worsened, and he was transferred to an academic tertiary care center for further evaluation. On admission, his temperature was 36.9°C, heart rate was 80 beats per minute, blood pressure was 139/67 mm Hg, respiratory rate was 10 breaths per minute, and oxygen saturation was 99% on room air. He was alert, awake, and oriented to person, place, and time. He was not jaundiced. He exhibited a moderate dysarthria characterized by monotone speech, decreased volume, decreased breath support, and a hoarse vocal quality with intact language function. Motor control of the lips, tongue, and mandible were normal. Motor strength was 5/5 bilaterally in the upper and lower extremities with the exception of right hip flexion, which was 4/5. The patient exhibited mild bilateral dysmetria on finger-to-nose examination, consistent with appendicular ataxia of the upper extremities. Reflexes were depressed throughout, and there was no asterixis. He had 2+ pulses in all extremities and 1+ pitting edema of the right lower extremity to the mid leg. Pulmonary examination revealed inspiratory crackles at the left base. The rest of the examination findings were normal.

The patient’s altered mental state appears to have resolved, and the neurological examination is now mainly characterized by signs that point to the cerebellum. The description of monotone speech typically refers to loss of prosody, the variable stress or intonation of speech, which is characteristic of a cerebellar speech pattern. The hoarseness should be explored to determine if it is a feature of the patient’s speech or is a separate process. Hoarseness may involve the vocal cord and therefore, potentially, cranial nerve X or its nuclei in the brainstem. The appendicular ataxia of the limbs points definitively to the cerebellar hemispheres or their pathways through the brainstem.

Unilateral lower extremity edema, especially in the context of a recent fracture, raises the possibility of deep vein thrombosis. If this patient has a right-to-left intracardiac or intrapulmonary shunt, embolization could lead to an ischemic stroke of the brainstem or cerebellum, potentially causing dysarthria.

Laboratory evaluation revealed hemoglobin level of 10.9 g/dL, white blood cell count of 5.3 × 10 ⁹ /L, platelet count of 169 × 10 ⁹ /L, glucose level of 177 mg/dL, corrected calcium level of 9.0 mg/dL, sodium level of 135 mmol/L, bicarbonate level of 30 mmol/L, creatinine level of 0.9 mg/dL, total bilirubin level of 1.3 mg/dL, direct bilirubin level of 0.4 mg/dL, alkaline phosphatase level of 503 U/L, alanine aminotransferase level of 12 U/L, aspartate aminotransferase level of 33 U/L, ammonia level of 49 µmol/L (range, 0-30 µ mol/L), international normalized ratio of 1.2, and troponin level of <0.01 ng/mL. Electrocardiogram showed normal sinus rhythm.

 

Some patients with bacterial meningitis do not have a leukocytosis, but patients with meningitis caused by seeding from a systemic infection nearly always do. In this patient’s case, lack of a leukocytosis makes bacterial meningitis very unlikely. The elevated alkaline phosphatase level is expected, as this level peaks about 3 weeks after a long-bone fracture and returns to normal over a few months.

Non-contrast CT scan of the head performed on admission demonstrated no large vessel cortical-based infarct, intracranial hemorrhage, hydrocephalus, mass effect, midline shift, or extra-axial fluid. There was mild cortical atrophy as well as very mild periventricular white matter hypodensity.

The atrophy and mild white-matter hypodensities seen on repeat noncontrast CT are nonspecific for any particular entity in this patient’s age group. MRI is more effective in evaluating toxic encephalopathies, including metronidazole toxicity or Wernicke encephalopathy, and in characterizing small infarcts or inflammatory conditions of the brainstem and cerebellum, which are poorly evaluated by CT due to the bone surrounded space of the posterior fossa. An urgent lumbar puncture is not necessary due to the slow pace of illness, lack of fever, nuchal rigidity, or serum elevated white blood cell count. Rather, performing MRI should be prioritized. If MRI is nondiagnostic, then spinal fluid should be evaluated for evidence of an infectious, autoimmune, paraneoplastic, or neoplastic process.

MRI was subsequently performed. It showed symmetric abnormal T₂ hyperintensities involving dentate nuclei (Figure 1), left inferior olivary nuclei (Figure 2), restiform bodies, pontine tegmentum, superior cerebellar peduncles, oculomotor nuclei, and subthalamic nuclei. The most prominent hyperintensity was in the dentate nuclei.

The clinical and radiographic features confirm a diagnosis of metronidazole-associated CNS neurotoxicity. The reason for the predilection for edema in these specific areas of the brainstem and midline cerebellum is unclear but likely is related to selective neuronal vulnerability in these structures. The treatment is to stop metronidazole. In addition, the fluctuating mental status should be evaluated with electroencephalogram to ensure concomitant seizures are not occurring.

These MRI findings were consistent with metronidazole toxicity. Metronidazole was discontinued, and 2 days later the patient’s speech improved. Two weeks after medication discontinuation, his speech was normal. There were no more episodes of confusion.

DISCUSSION

Metronidazole was originally developed in France during the 1950s as an anti-parasitic medication to treat trichomonas infections. In 1962, its antibacterial properties were discovered after a patient with bacterial gingivitis improved while taking metronidazole for treatment of Trichomonas vaginalis.¹ Since that time metronidazole has become a first-line treatment for anaerobic bacteria and is now recommended by the Infectious Diseases Society of America² and the American College of Gastroenterology³ as a first-line therapy for mild and moderate C difficile infections.

Common side effects of metronidazole are nausea, vomiting, decreased appetite, diarrhea, headaches, peripheral neuropathy, and metallic taste; less common is CNS toxicity. Although the incidence of CNS toxicity is unknown, a systematic review of the literature found 64 cases reported between 1965 and 2011.⁴ CNS toxicity most often occurs between the fifth and sixth decades of life, and about two thirds of the people affected are men.⁴ CNS adverse effects characteristically fall into 4 categories: cerebellar dysfunction (eg, ataxia, dysarthria, dysmetria, nystagmus; 75%), AMS (33%), seizures (13%), and a combination of the first 3 categories.⁴

The exact mechanism of metronidazole CNS toxicity is unknown, but vasogenic or cytotoxic edema may be involved.^5,6 Other potential etiologies are neural protein inhibition, reversible mitochondrial dysfunction, and modifications of the inhibitory neurotransmitter gamma-aminobutyric acid receptor in the cerebellum.^7,8 There is no known genetic predisposition. Although the risk for CNS toxicity traditionally is thought to correlate with therapy duration and cumulative dose,^7,9 in 2011 a systemic review found no significant correlation.⁴ In fact, 26% of patients with CNS toxicity were treated with metronidazole for less than 1 week at time of diagnosis.⁴

Brain CT is typically normal. On brain MRI, lesions most commonly appear as bilateral symmetric T₂ hyperintensities, most often in the cerebellar dentate nuclei (85%) and less often in the midbrain (55%), the splenium of the corpus callosum (50%), the pons (35%), and the medulla (30%).^4,10 Radiographic changes have been noted as early as 3 days after symptom onset. Based on damage severity and area affected (white or gray matter), vasogenic edema and cytotoxic edema may in combination be contributing to MRI abnormalities.^6,10 Hyperintensities of the bilateral dentate nuclei can help in distinguishing metronidazole-induced encephalopathy from other potential disease processes, such as Wernicke encephalopathy.¹⁰

The prognosis for patients with metronidazole-induced neurotoxicity is favorable if metronidazole is discontinued. Approximately two-thirds of patients will have complete resolution of symptoms, which is more commonly observed when patients present with seizures or altered mental status. Approximately one-third will show partial improvement, particularly if the symptoms are due to cerebellar dysfunction. It is rare to experience permanent damage or death.⁴ Neurologic recovery usually begins within a week after medication discontinuation but may take months for complete recovery to occur.^6,8,9,11 Follow-up imaging typically shows reversal of the original lesions, but this does not always correlate with symptom improvement.^4,10

Despite its frequent use and long history, metronidazole can have potentially severe toxicity. When patients who are taking this medication present with new signs and symptoms of CNS dysfunction, hospitalists should include metronidazole CNS toxicity in the differential diagnosis and, if they suspect toxicity, have a brain MRI performed. Hospitalists often prescribe metronidazole because of the increasing number of patients being discharged from acute-care hospitals with a diagnosis of C difficile colitis.¹² Brain MRI remains the imaging modality of choice for diagnosis. Discontinuation of metronidazole is usually salutary in reversing symptoms. Being keenly aware of this toxicity will help clinicians avoid being rendered speechless by a patient rendered speechless.

 

TEACHING POINTS

• CNS toxicity is a rare but potentially devastating side effect of metronidazole exposure.

• Metronidazole CNS adverse effects characteristically fall under 4 categories:

○ Cerebellar dysfunction, such as ataxia, dysarthria, dysmetria, or nystagmus (75%).

○ AMS (33%).

○ Seizures (13%).

○ A combination of the first 3 categories.

• Typically lesions indicating metronidazole toxicity on brain MRI are bilateral symmetric hyperintensities on T2-weighted imaging in the cerebellar dentate nuclei, corpus callosum, midbrain, pons, or medulla.
• Treatment of CNS toxicity is metronidazole discontinuation, which results in a high rate of symptom resolution.

Disclosure

Nothing to report.

 

From the Divisions of Hospital Medicine (Drs. Dawson, Chirila, Bhide, and Burton) and Biomedical Statistics and Informatics (Ms. Thomas), Mayo Clinic, Jacksonville, FL, and the Division of Hospital Medicine, Mayo Clinic, Phoenix, AZ (Dr. Cannon).

 

Abstract

• Objective: To validate an electronic tool created to identify inpatients who are at risk of readmission within 30 days and quantify the predictive performance of the readmission risk score (RRS).
• Methods: Retrospective cohort study including inpa-tients who were discharged between 1 Nov 2012 and 31 Dec 2012. The ability of the RRS to discriminate between those who did and did not have a 30-day urgent readmission was quantified by the c statistic. Calibration was assessed by plotting the observed and predicted probability of 30-day urgent readmission. Predicted probabilities were obtained from generalized estimating equations, clustering on patient.
• Results: Of 1689 hospital inpatient discharges (1515 patients), 159 (9.4%) had a 30-day urgent readmission. The RRS had some discriminative ability (c statistic: 0.612; 95% confidence interval: 0.570–0.655) and good calibration.
• Conclusions: Our study shows that the RRS has some discriminative ability. The automated tool can be used to estimate the probability of a 30-day urgent readmission.

 

Hospital readmissions are increasingly scrutinized by the Center for Medicare and Medicaid Services and other payers due to their frequency and high cost. It is estimated that up to 25% of all patients discharged from acute care hospitals are readmitted within 30 days [1]. To address this problem, the Center for Medicare and Medicaid Services is using these rates as one of the benchmarks for quality for hospitals and health care organizations and has begun to assess penalties to those institutions with the highest rates. This scrutiny and the desire for better patient care transitions has resulted in most hospitals implementing various initiatives to reduce potentially avoidable readmissions.

Multiple interventions have been shown to reduce readmissions [2,3]. These interventions have varying effectiveness and are often labor intensive and thus costly to the institutions implementing them. In fact, no one intervention has been shown to be effective alone [4], and it may take several concurrent interventions targeting the highest risk patients to improve transitions of care at discharge that result in reduced readmissions. Many experts do recommend risk stratifying patients in order to target interventions to the highest risk patients for effective use of resources [5,6]. Several risk factor assessments have been proposed with varying success [7–13]. Multiple factors can limit the effectiveness of these risk stratification profiles. They may have low sensitivity and specificity, be based solely on retrospective data, be limited to certain populations, or be created from administrative data only without taking psychosocial factors into consideration [14].

An effective risk assessment ideally would encompass multiple known risk factors including certain comorbidities such as malignancy and heart failure, psychosocial factors such as health literacy and social support, and administrative data including payment source and demographics. All of these have been shown in prior studies to contribute to readmissions [7–13]. In addition, availability of the assessment early in the hospitalization would allow for interventions throughout the hospital stay to mitigate the effect of these factors where possible. To address these needs, our institution formed a readmission task force in January 2010 to review published literature on hospital 30-day readmissions and create a readmission risk score (RRS). The aim of this study was to quantify the predictive performance of the RRS after it was first implemented into the electronic medical record (EMR) in November 2012.

Methods

Study Design and Cohort

The Mayo Clinic institutional review board approved this study. The study was conducted at Mayo Clinic in Jacksonville, Florida, a tertiary care hospital in a community setting. The RRS (Table 1) was created as a modification of 2 published studies [9,15]. Using AMALGA software (Microsoft, Redmond, WA), an automated electronic tool was developed to collect the necessary clinical, psychosocial, and financial information on hospital admission to calculate the RRS. This initial risk score was run retrospectively on a data set of approximately 2000 patients who had been readmitted 2 years prior to the study to determine the cut-off values for high, medium, and low risk prior to the implementation of the electronic tool.

All consecutive adult inpatients who were discharged between 1 November 2012 and 31 December 2012 were included in this retrospective cohort study. This narrow time frame corresponded to the period from RRS tool implementation to the start of readmission interventions. We excluded hospitalizations if the patient died in the hospital.

Outcome Measures

The primary outcome was a 30-day urgent readmission, which included readmissions categorized as either emergency, urgent, or semi-urgent. Secondary outcomes included any 30-day readmission and 30-day death. Only readmissions to Mayo Clinic were examined.

Predictors

In collaboration with the information technology department, an algorithm was written to extract data from the EMR for each patient within 24 hours of admission to the hospital. This data was retrieved from existing repositories of patient information, such as demographic information, payer source, medication list, problem list, and past medical history. In addition, each patient was interviewed by a nurse at the time of admission, and the nurse completed an “admission profile” in the EMR that confirmed or entered past medical history, medications, social support at home, depression symptoms, and learning styles, among other information (Table 1). The algorithm was able to extract data from this evaluation also, so that each element of the risk score was correlated to at least one data source in the EMR. The algorithm then assigned the correct value to each element, and the total score was electronically calculated and placed in a discrete cell in each patient’s record. The algorithm was automatically run again 48 hours after the initial scoring in order to assure completeness of the information. If the patient had a length of stay greater than 5 days, an additional score was generated to include the length of stay component.

Statistical Analysis

The predictive performance of the RRS was assessed by evaluating the discrimination and calibration. Discrimination is the ability of the RRS to separate those who had a 30-day urgent readmission and those who did not. Discrimination was quantified by the c statistic, which is equivalent to the area under the receiver operating characteristic curve in this study owing to the use of binary endpoints. A c statistic of 1.0 would indicate that the RRS perfectly predicts 30-day urgent readmission while a c statistic of 0.5 would indicate the RRS has no apparent accuracy in predicting 30-day urgent readmission. Calibration assesses how closely predicted outcomes agree with observed outcomes. The predicted probability of 30-day urgent readmission was estimated utilizing a generalized estimating equation model, clustering on patient, with RRS as the only predictor variable. Inpatient discharges were divided into deciles of the predicted probabilities for 30-day urgent readmission. Agreement of the predicted and observed outcomes was displayed graphically according to decile of the predicted outcomes. All analyses were performed using SAS (version 9.3, SAS Institute, Cary, NC) and R statistical software (version 3.1.1, R Foundation for Statistical Computing, Vienna, Austria).

 

Results

There were 1689 hospital inpatient discharges for 1515 patients during the study period. In this population, the mean age was 64 ± 17 years, 50% were female, and 87% were Caucasian. Additional characteristics are reported in Table 2. Of the 1689 hospital inpatient discharges, 159 (9.4%) resulted in a 30-day urgent readmission and 190 (11.2%) resulted in any 30-day readmission. Among the 1515 patients, 57 (3.8%) died within 30 days of discharge.

The RRS was significantly associated with 30-day urgent readmission (odds ratio [OR] for 1-point increase in the RRS, 1.07 [95% confidence interval {CI} 1.05–1.10]; P < 0.001). A c statistic of 0.612 (95% CI 0.570–0.655) indicates that the RRS has some ability to discriminate between those with and without a 30-day urgent readmission (Figure, Table 3). The expected and observed probabilities of 30-day urgent readmission were similar in each decile of the RRS. The calibration (Table 4) shows that although there is some deviation between the observed and expected probabilities, 

the calibration is fairly good, particularly at the higher risk levels, making the tool more valuable for the high-risk patients.

The RRS was also significantly associated with each of the secondary outcome measures. The odds ratios for a 1-point increase in the RRS for any 30-day readmission was 1.06 (95% CI 1.03–1.09, P < 0.001) and the c statistic was 0.591 (95% CI 0.551–0.631, Table 2). The odds ratios for a 1-point increase in the 

RRS for 30-day death was 1.13 (95% CI 1.08–1.18, P < 0.001) and the c statistic was 0.696 (95% CI 0.633–0.758, Table 2). The odds ratios for a 1-point increase in the RRS for 30-day death or 30-day urgent readmission was 1.09 (95% CI 1.07–1.12, P < 0.001) and the c statistic was 0.639 (95% CI 0.601–0.676, Table 2).

 

Discussion

Our study provides evidence that the RRS has some ability to discriminate between patients who did and did not have a 30-day urgent readmission (c statistic 0.612 [95% CI 0.570–0.655]). More importantly the calibration appears to be good particularly in the higher risk patients, which are the most crucial to identify in order to target interventions.

In addition to predicting the risk of readmission, our method of risk evaluation has several other advantages. First, the risk score is assigned to each patient within 24 to 48 hours of admission by using elements available at the time of, or soon after, admission. This early evaluation during the hospitalization identifies patients who could benefit from interventions throughout the stay that could help mitigate the risks and allow for a safer transition. Other studies have used elements available only at discharge, such as lab values and length of stay [7,11]. Donze et al used 7 elements in a validated scoring system, but several of the elements were discharge values and the risk assessment system had a fair discriminatory value with a c statistic of 0.71, similar to our results. The advantage to having the score available at admission is that several of the factors used to compose the RRS could be addressed during the hospitalization, including increased education for those with greater than 7 medications, intensive care management intervention for those with a lack of social support, and increased or modified education for those with low health literacy.

Second, the score is derived entirely from elements available in the EMR, thus the score is calculated automatically within 24 hours of admission and displayed in the chart for all providers to access. This eliminates any need for individual chart review or patient evaluation outside the normal admission process, making this system extremely efficient. Van Walraven et al [9] devised a scoring system using length of stay, acuity of admission, comorbidities and emergency department use (LACE index), with a validation c statistic of 0.684, which again is similar to our results. However, the LACE index uses the Charlson comorbidity index as a measure of patient comorbidity and this can be cumbersome to calculate in clinical practice. Having the score automatically available to all providers caring for the patient increases their awareness of the patient’s level of risk. Allaudeen and colleagues showed that providers are unable to intuitively predict those patients who are at high-risk for readmission [15]; therefore, an objective, readily available risk stratification is necessary to inform the providers.

 

Third, the risk scoring system uses elements from varied sources to include social, medical, and individual factors, all of which have been shown to increase risk of 30-day readmissions [9,15]. An accurate risk scoring system, ideally, should include elements from multiple sources, and use of the EMR allows for this varied compilation. The risk evaluation is done on every patient, regardless of admitting diagnosis, and in spite of this heterogeneous population, it was still found to be significantly accurate. Prior studies have looked at individual populations [7,10,12,13,16]; however, this can miss many patient populations that are also high-risk. Tailoring individual risk algorithms by diagnosis can also be labor intensive.

Our study has limitations. It is a retrospective study and included a relatively short study period of 2 months. This period was chosen because it represented the time from when the RRS was first implemented to when interventions to reduce readmission according to the RRS began, however, it still encompassed a significant number of discharges. We were only able to evaluate readmissions to our own facility; therefore, patients readmitted to other facilities were not included. Although readmission to any facility is undesirable, having a risk scoring system that can reliably predict readmission to the index admission hospital is still helpful. In addition, we only validated the risk score on patients in our own facility. A larger population from multiple facilities would be helpful for further validation. In spite of this limitation we would expect that most of our readmissions return to our own facility given our community setting. In fact, based on Medicare data for readmissions to all facilities, the difference in readmission rate between our facility and all facilities differs by less than 4%.

In summary, we developed a comprehensive risk scoring system that proved to be moderately predictive of readmission that encompasses multiple factors, is available to all providers early in a hospitalization, and is completely automated via the EMR. Further studies are ongoing to refine this score and improve the predictive performance.

 

Corresponding author: Nancy L. Dawson, MD, Division of Hospital Medicine, Mayo Clinic, 4500 San Pablo Road, Jacksonville, FL 32224, dawson.nancy11@mayo.edu.

Financial disclosures: None.

From the Divisions of Hospital Medicine (Drs. Dawson, Chirila, Bhide, and Burton) and Biomedical Statistics and Informatics (Ms. Thomas), Mayo Clinic, Jacksonville, FL, and the Division of Hospital Medicine, Mayo Clinic, Phoenix, AZ (Dr. Cannon).

 

Abstract

• Objective: To validate an electronic tool created to identify inpatients who are at risk of readmission within 30 days and quantify the predictive performance of the readmission risk score (RRS).
• Methods: Retrospective cohort study including inpa-tients who were discharged between 1 Nov 2012 and 31 Dec 2012. The ability of the RRS to discriminate between those who did and did not have a 30-day urgent readmission was quantified by the c statistic. Calibration was assessed by plotting the observed and predicted probability of 30-day urgent readmission. Predicted probabilities were obtained from generalized estimating equations, clustering on patient.
• Results: Of 1689 hospital inpatient discharges (1515 patients), 159 (9.4%) had a 30-day urgent readmission. The RRS had some discriminative ability (c statistic: 0.612; 95% confidence interval: 0.570–0.655) and good calibration.
• Conclusions: Our study shows that the RRS has some discriminative ability. The automated tool can be used to estimate the probability of a 30-day urgent readmission.

 

Hospital readmissions are increasingly scrutinized by the Center for Medicare and Medicaid Services and other payers due to their frequency and high cost. It is estimated that up to 25% of all patients discharged from acute care hospitals are readmitted within 30 days [1]. To address this problem, the Center for Medicare and Medicaid Services is using these rates as one of the benchmarks for quality for hospitals and health care organizations and has begun to assess penalties to those institutions with the highest rates. This scrutiny and the desire for better patient care transitions has resulted in most hospitals implementing various initiatives to reduce potentially avoidable readmissions.

Multiple interventions have been shown to reduce readmissions [2,3]. These interventions have varying effectiveness and are often labor intensive and thus costly to the institutions implementing them. In fact, no one intervention has been shown to be effective alone [4], and it may take several concurrent interventions targeting the highest risk patients to improve transitions of care at discharge that result in reduced readmissions. Many experts do recommend risk stratifying patients in order to target interventions to the highest risk patients for effective use of resources [5,6]. Several risk factor assessments have been proposed with varying success [7–13]. Multiple factors can limit the effectiveness of these risk stratification profiles. They may have low sensitivity and specificity, be based solely on retrospective data, be limited to certain populations, or be created from administrative data only without taking psychosocial factors into consideration [14].

An effective risk assessment ideally would encompass multiple known risk factors including certain comorbidities such as malignancy and heart failure, psychosocial factors such as health literacy and social support, and administrative data including payment source and demographics. All of these have been shown in prior studies to contribute to readmissions [7–13]. In addition, availability of the assessment early in the hospitalization would allow for interventions throughout the hospital stay to mitigate the effect of these factors where possible. To address these needs, our institution formed a readmission task force in January 2010 to review published literature on hospital 30-day readmissions and create a readmission risk score (RRS). The aim of this study was to quantify the predictive performance of the RRS after it was first implemented into the electronic medical record (EMR) in November 2012.

Methods

Study Design and Cohort

The Mayo Clinic institutional review board approved this study. The study was conducted at Mayo Clinic in Jacksonville, Florida, a tertiary care hospital in a community setting. The RRS (Table 1) was created as a modification of 2 published studies [9,15]. Using AMALGA software (Microsoft, Redmond, WA), an automated electronic tool was developed to collect the necessary clinical, psychosocial, and financial information on hospital admission to calculate the RRS. This initial risk score was run retrospectively on a data set of approximately 2000 patients who had been readmitted 2 years prior to the study to determine the cut-off values for high, medium, and low risk prior to the implementation of the electronic tool.

All consecutive adult inpatients who were discharged between 1 November 2012 and 31 December 2012 were included in this retrospective cohort study. This narrow time frame corresponded to the period from RRS tool implementation to the start of readmission interventions. We excluded hospitalizations if the patient died in the hospital.

Outcome Measures

The primary outcome was a 30-day urgent readmission, which included readmissions categorized as either emergency, urgent, or semi-urgent. Secondary outcomes included any 30-day readmission and 30-day death. Only readmissions to Mayo Clinic were examined.

Predictors

In collaboration with the information technology department, an algorithm was written to extract data from the EMR for each patient within 24 hours of admission to the hospital. This data was retrieved from existing repositories of patient information, such as demographic information, payer source, medication list, problem list, and past medical history. In addition, each patient was interviewed by a nurse at the time of admission, and the nurse completed an “admission profile” in the EMR that confirmed or entered past medical history, medications, social support at home, depression symptoms, and learning styles, among other information (Table 1). The algorithm was able to extract data from this evaluation also, so that each element of the risk score was correlated to at least one data source in the EMR. The algorithm then assigned the correct value to each element, and the total score was electronically calculated and placed in a discrete cell in each patient’s record. The algorithm was automatically run again 48 hours after the initial scoring in order to assure completeness of the information. If the patient had a length of stay greater than 5 days, an additional score was generated to include the length of stay component.

Statistical Analysis

The predictive performance of the RRS was assessed by evaluating the discrimination and calibration. Discrimination is the ability of the RRS to separate those who had a 30-day urgent readmission and those who did not. Discrimination was quantified by the c statistic, which is equivalent to the area under the receiver operating characteristic curve in this study owing to the use of binary endpoints. A c statistic of 1.0 would indicate that the RRS perfectly predicts 30-day urgent readmission while a c statistic of 0.5 would indicate the RRS has no apparent accuracy in predicting 30-day urgent readmission. Calibration assesses how closely predicted outcomes agree with observed outcomes. The predicted probability of 30-day urgent readmission was estimated utilizing a generalized estimating equation model, clustering on patient, with RRS as the only predictor variable. Inpatient discharges were divided into deciles of the predicted probabilities for 30-day urgent readmission. Agreement of the predicted and observed outcomes was displayed graphically according to decile of the predicted outcomes. All analyses were performed using SAS (version 9.3, SAS Institute, Cary, NC) and R statistical software (version 3.1.1, R Foundation for Statistical Computing, Vienna, Austria).

 

Results

There were 1689 hospital inpatient discharges for 1515 patients during the study period. In this population, the mean age was 64 ± 17 years, 50% were female, and 87% were Caucasian. Additional characteristics are reported in Table 2. Of the 1689 hospital inpatient discharges, 159 (9.4%) resulted in a 30-day urgent readmission and 190 (11.2%) resulted in any 30-day readmission. Among the 1515 patients, 57 (3.8%) died within 30 days of discharge.

The RRS was significantly associated with 30-day urgent readmission (odds ratio [OR] for 1-point increase in the RRS, 1.07 [95% confidence interval {CI} 1.05–1.10]; P < 0.001). A c statistic of 0.612 (95% CI 0.570–0.655) indicates that the RRS has some ability to discriminate between those with and without a 30-day urgent readmission (Figure, Table 3). The expected and observed probabilities of 30-day urgent readmission were similar in each decile of the RRS. The calibration (Table 4) shows that although there is some deviation between the observed and expected probabilities, 

the calibration is fairly good, particularly at the higher risk levels, making the tool more valuable for the high-risk patients.

The RRS was also significantly associated with each of the secondary outcome measures. The odds ratios for a 1-point increase in the RRS for any 30-day readmission was 1.06 (95% CI 1.03–1.09, P < 0.001) and the c statistic was 0.591 (95% CI 0.551–0.631, Table 2). The odds ratios for a 1-point increase in the 

RRS for 30-day death was 1.13 (95% CI 1.08–1.18, P < 0.001) and the c statistic was 0.696 (95% CI 0.633–0.758, Table 2). The odds ratios for a 1-point increase in the RRS for 30-day death or 30-day urgent readmission was 1.09 (95% CI 1.07–1.12, P < 0.001) and the c statistic was 0.639 (95% CI 0.601–0.676, Table 2).

 

Discussion

Our study provides evidence that the RRS has some ability to discriminate between patients who did and did not have a 30-day urgent readmission (c statistic 0.612 [95% CI 0.570–0.655]). More importantly the calibration appears to be good particularly in the higher risk patients, which are the most crucial to identify in order to target interventions.

In addition to predicting the risk of readmission, our method of risk evaluation has several other advantages. First, the risk score is assigned to each patient within 24 to 48 hours of admission by using elements available at the time of, or soon after, admission. This early evaluation during the hospitalization identifies patients who could benefit from interventions throughout the stay that could help mitigate the risks and allow for a safer transition. Other studies have used elements available only at discharge, such as lab values and length of stay [7,11]. Donze et al used 7 elements in a validated scoring system, but several of the elements were discharge values and the risk assessment system had a fair discriminatory value with a c statistic of 0.71, similar to our results. The advantage to having the score available at admission is that several of the factors used to compose the RRS could be addressed during the hospitalization, including increased education for those with greater than 7 medications, intensive care management intervention for those with a lack of social support, and increased or modified education for those with low health literacy.

Second, the score is derived entirely from elements available in the EMR, thus the score is calculated automatically within 24 hours of admission and displayed in the chart for all providers to access. This eliminates any need for individual chart review or patient evaluation outside the normal admission process, making this system extremely efficient. Van Walraven et al [9] devised a scoring system using length of stay, acuity of admission, comorbidities and emergency department use (LACE index), with a validation c statistic of 0.684, which again is similar to our results. However, the LACE index uses the Charlson comorbidity index as a measure of patient comorbidity and this can be cumbersome to calculate in clinical practice. Having the score automatically available to all providers caring for the patient increases their awareness of the patient’s level of risk. Allaudeen and colleagues showed that providers are unable to intuitively predict those patients who are at high-risk for readmission [15]; therefore, an objective, readily available risk stratification is necessary to inform the providers.

 

Third, the risk scoring system uses elements from varied sources to include social, medical, and individual factors, all of which have been shown to increase risk of 30-day readmissions [9,15]. An accurate risk scoring system, ideally, should include elements from multiple sources, and use of the EMR allows for this varied compilation. The risk evaluation is done on every patient, regardless of admitting diagnosis, and in spite of this heterogeneous population, it was still found to be significantly accurate. Prior studies have looked at individual populations [7,10,12,13,16]; however, this can miss many patient populations that are also high-risk. Tailoring individual risk algorithms by diagnosis can also be labor intensive.

Our study has limitations. It is a retrospective study and included a relatively short study period of 2 months. This period was chosen because it represented the time from when the RRS was first implemented to when interventions to reduce readmission according to the RRS began, however, it still encompassed a significant number of discharges. We were only able to evaluate readmissions to our own facility; therefore, patients readmitted to other facilities were not included. Although readmission to any facility is undesirable, having a risk scoring system that can reliably predict readmission to the index admission hospital is still helpful. In addition, we only validated the risk score on patients in our own facility. A larger population from multiple facilities would be helpful for further validation. In spite of this limitation we would expect that most of our readmissions return to our own facility given our community setting. In fact, based on Medicare data for readmissions to all facilities, the difference in readmission rate between our facility and all facilities differs by less than 4%.

In summary, we developed a comprehensive risk scoring system that proved to be moderately predictive of readmission that encompasses multiple factors, is available to all providers early in a hospitalization, and is completely automated via the EMR. Further studies are ongoing to refine this score and improve the predictive performance.

 

Corresponding author: Nancy L. Dawson, MD, Division of Hospital Medicine, Mayo Clinic, 4500 San Pablo Road, Jacksonville, FL 32224, dawson.nancy11@mayo.edu.

Financial disclosures: None.

From the Divisions of Hospital Medicine (Drs. Dawson, Chirila, Bhide, and Burton) and Biomedical Statistics and Informatics (Ms. Thomas), Mayo Clinic, Jacksonville, FL, and the Division of Hospital Medicine, Mayo Clinic, Phoenix, AZ (Dr. Cannon).

 

Abstract

• Objective: To validate an electronic tool created to identify inpatients who are at risk of readmission within 30 days and quantify the predictive performance of the readmission risk score (RRS).
• Methods: Retrospective cohort study including inpa-tients who were discharged between 1 Nov 2012 and 31 Dec 2012. The ability of the RRS to discriminate between those who did and did not have a 30-day urgent readmission was quantified by the c statistic. Calibration was assessed by plotting the observed and predicted probability of 30-day urgent readmission. Predicted probabilities were obtained from generalized estimating equations, clustering on patient.
• Results: Of 1689 hospital inpatient discharges (1515 patients), 159 (9.4%) had a 30-day urgent readmission. The RRS had some discriminative ability (c statistic: 0.612; 95% confidence interval: 0.570–0.655) and good calibration.
• Conclusions: Our study shows that the RRS has some discriminative ability. The automated tool can be used to estimate the probability of a 30-day urgent readmission.

 

Hospital readmissions are increasingly scrutinized by the Center for Medicare and Medicaid Services and other payers due to their frequency and high cost. It is estimated that up to 25% of all patients discharged from acute care hospitals are readmitted within 30 days [1]. To address this problem, the Center for Medicare and Medicaid Services is using these rates as one of the benchmarks for quality for hospitals and health care organizations and has begun to assess penalties to those institutions with the highest rates. This scrutiny and the desire for better patient care transitions has resulted in most hospitals implementing various initiatives to reduce potentially avoidable readmissions.

Multiple interventions have been shown to reduce readmissions [2,3]. These interventions have varying effectiveness and are often labor intensive and thus costly to the institutions implementing them. In fact, no one intervention has been shown to be effective alone [4], and it may take several concurrent interventions targeting the highest risk patients to improve transitions of care at discharge that result in reduced readmissions. Many experts do recommend risk stratifying patients in order to target interventions to the highest risk patients for effective use of resources [5,6]. Several risk factor assessments have been proposed with varying success [7–13]. Multiple factors can limit the effectiveness of these risk stratification profiles. They may have low sensitivity and specificity, be based solely on retrospective data, be limited to certain populations, or be created from administrative data only without taking psychosocial factors into consideration [14].

An effective risk assessment ideally would encompass multiple known risk factors including certain comorbidities such as malignancy and heart failure, psychosocial factors such as health literacy and social support, and administrative data including payment source and demographics. All of these have been shown in prior studies to contribute to readmissions [7–13]. In addition, availability of the assessment early in the hospitalization would allow for interventions throughout the hospital stay to mitigate the effect of these factors where possible. To address these needs, our institution formed a readmission task force in January 2010 to review published literature on hospital 30-day readmissions and create a readmission risk score (RRS). The aim of this study was to quantify the predictive performance of the RRS after it was first implemented into the electronic medical record (EMR) in November 2012.

Methods

Study Design and Cohort

The Mayo Clinic institutional review board approved this study. The study was conducted at Mayo Clinic in Jacksonville, Florida, a tertiary care hospital in a community setting. The RRS (Table 1) was created as a modification of 2 published studies [9,15]. Using AMALGA software (Microsoft, Redmond, WA), an automated electronic tool was developed to collect the necessary clinical, psychosocial, and financial information on hospital admission to calculate the RRS. This initial risk score was run retrospectively on a data set of approximately 2000 patients who had been readmitted 2 years prior to the study to determine the cut-off values for high, medium, and low risk prior to the implementation of the electronic tool.

All consecutive adult inpatients who were discharged between 1 November 2012 and 31 December 2012 were included in this retrospective cohort study. This narrow time frame corresponded to the period from RRS tool implementation to the start of readmission interventions. We excluded hospitalizations if the patient died in the hospital.

Outcome Measures

The primary outcome was a 30-day urgent readmission, which included readmissions categorized as either emergency, urgent, or semi-urgent. Secondary outcomes included any 30-day readmission and 30-day death. Only readmissions to Mayo Clinic were examined.

Predictors

In collaboration with the information technology department, an algorithm was written to extract data from the EMR for each patient within 24 hours of admission to the hospital. This data was retrieved from existing repositories of patient information, such as demographic information, payer source, medication list, problem list, and past medical history. In addition, each patient was interviewed by a nurse at the time of admission, and the nurse completed an “admission profile” in the EMR that confirmed or entered past medical history, medications, social support at home, depression symptoms, and learning styles, among other information (Table 1). The algorithm was able to extract data from this evaluation also, so that each element of the risk score was correlated to at least one data source in the EMR. The algorithm then assigned the correct value to each element, and the total score was electronically calculated and placed in a discrete cell in each patient’s record. The algorithm was automatically run again 48 hours after the initial scoring in order to assure completeness of the information. If the patient had a length of stay greater than 5 days, an additional score was generated to include the length of stay component.

Statistical Analysis

The predictive performance of the RRS was assessed by evaluating the discrimination and calibration. Discrimination is the ability of the RRS to separate those who had a 30-day urgent readmission and those who did not. Discrimination was quantified by the c statistic, which is equivalent to the area under the receiver operating characteristic curve in this study owing to the use of binary endpoints. A c statistic of 1.0 would indicate that the RRS perfectly predicts 30-day urgent readmission while a c statistic of 0.5 would indicate the RRS has no apparent accuracy in predicting 30-day urgent readmission. Calibration assesses how closely predicted outcomes agree with observed outcomes. The predicted probability of 30-day urgent readmission was estimated utilizing a generalized estimating equation model, clustering on patient, with RRS as the only predictor variable. Inpatient discharges were divided into deciles of the predicted probabilities for 30-day urgent readmission. Agreement of the predicted and observed outcomes was displayed graphically according to decile of the predicted outcomes. All analyses were performed using SAS (version 9.3, SAS Institute, Cary, NC) and R statistical software (version 3.1.1, R Foundation for Statistical Computing, Vienna, Austria).

 

Results

There were 1689 hospital inpatient discharges for 1515 patients during the study period. In this population, the mean age was 64 ± 17 years, 50% were female, and 87% were Caucasian. Additional characteristics are reported in Table 2. Of the 1689 hospital inpatient discharges, 159 (9.4%) resulted in a 30-day urgent readmission and 190 (11.2%) resulted in any 30-day readmission. Among the 1515 patients, 57 (3.8%) died within 30 days of discharge.

The RRS was significantly associated with 30-day urgent readmission (odds ratio [OR] for 1-point increase in the RRS, 1.07 [95% confidence interval {CI} 1.05–1.10]; P < 0.001). A c statistic of 0.612 (95% CI 0.570–0.655) indicates that the RRS has some ability to discriminate between those with and without a 30-day urgent readmission (Figure, Table 3). The expected and observed probabilities of 30-day urgent readmission were similar in each decile of the RRS. The calibration (Table 4) shows that although there is some deviation between the observed and expected probabilities, 

the calibration is fairly good, particularly at the higher risk levels, making the tool more valuable for the high-risk patients.

The RRS was also significantly associated with each of the secondary outcome measures. The odds ratios for a 1-point increase in the RRS for any 30-day readmission was 1.06 (95% CI 1.03–1.09, P < 0.001) and the c statistic was 0.591 (95% CI 0.551–0.631, Table 2). The odds ratios for a 1-point increase in the 

RRS for 30-day death was 1.13 (95% CI 1.08–1.18, P < 0.001) and the c statistic was 0.696 (95% CI 0.633–0.758, Table 2). The odds ratios for a 1-point increase in the RRS for 30-day death or 30-day urgent readmission was 1.09 (95% CI 1.07–1.12, P < 0.001) and the c statistic was 0.639 (95% CI 0.601–0.676, Table 2).

 

Discussion

Our study provides evidence that the RRS has some ability to discriminate between patients who did and did not have a 30-day urgent readmission (c statistic 0.612 [95% CI 0.570–0.655]). More importantly the calibration appears to be good particularly in the higher risk patients, which are the most crucial to identify in order to target interventions.

In addition to predicting the risk of readmission, our method of risk evaluation has several other advantages. First, the risk score is assigned to each patient within 24 to 48 hours of admission by using elements available at the time of, or soon after, admission. This early evaluation during the hospitalization identifies patients who could benefit from interventions throughout the stay that could help mitigate the risks and allow for a safer transition. Other studies have used elements available only at discharge, such as lab values and length of stay [7,11]. Donze et al used 7 elements in a validated scoring system, but several of the elements were discharge values and the risk assessment system had a fair discriminatory value with a c statistic of 0.71, similar to our results. The advantage to having the score available at admission is that several of the factors used to compose the RRS could be addressed during the hospitalization, including increased education for those with greater than 7 medications, intensive care management intervention for those with a lack of social support, and increased or modified education for those with low health literacy.

Second, the score is derived entirely from elements available in the EMR, thus the score is calculated automatically within 24 hours of admission and displayed in the chart for all providers to access. This eliminates any need for individual chart review or patient evaluation outside the normal admission process, making this system extremely efficient. Van Walraven et al [9] devised a scoring system using length of stay, acuity of admission, comorbidities and emergency department use (LACE index), with a validation c statistic of 0.684, which again is similar to our results. However, the LACE index uses the Charlson comorbidity index as a measure of patient comorbidity and this can be cumbersome to calculate in clinical practice. Having the score automatically available to all providers caring for the patient increases their awareness of the patient’s level of risk. Allaudeen and colleagues showed that providers are unable to intuitively predict those patients who are at high-risk for readmission [15]; therefore, an objective, readily available risk stratification is necessary to inform the providers.

 

Third, the risk scoring system uses elements from varied sources to include social, medical, and individual factors, all of which have been shown to increase risk of 30-day readmissions [9,15]. An accurate risk scoring system, ideally, should include elements from multiple sources, and use of the EMR allows for this varied compilation. The risk evaluation is done on every patient, regardless of admitting diagnosis, and in spite of this heterogeneous population, it was still found to be significantly accurate. Prior studies have looked at individual populations [7,10,12,13,16]; however, this can miss many patient populations that are also high-risk. Tailoring individual risk algorithms by diagnosis can also be labor intensive.

Our study has limitations. It is a retrospective study and included a relatively short study period of 2 months. This period was chosen because it represented the time from when the RRS was first implemented to when interventions to reduce readmission according to the RRS began, however, it still encompassed a significant number of discharges. We were only able to evaluate readmissions to our own facility; therefore, patients readmitted to other facilities were not included. Although readmission to any facility is undesirable, having a risk scoring system that can reliably predict readmission to the index admission hospital is still helpful. In addition, we only validated the risk score on patients in our own facility. A larger population from multiple facilities would be helpful for further validation. In spite of this limitation we would expect that most of our readmissions return to our own facility given our community setting. In fact, based on Medicare data for readmissions to all facilities, the difference in readmission rate between our facility and all facilities differs by less than 4%.

In summary, we developed a comprehensive risk scoring system that proved to be moderately predictive of readmission that encompasses multiple factors, is available to all providers early in a hospitalization, and is completely automated via the EMR. Further studies are ongoing to refine this score and improve the predictive performance.

 

Corresponding author: Nancy L. Dawson, MD, Division of Hospital Medicine, Mayo Clinic, 4500 San Pablo Road, Jacksonville, FL 32224, dawson.nancy11@mayo.edu.

Financial disclosures: None.