John A. Batsis, MD

Public health concerns such as the aging population¹ and the increasing prevalence of obesity² are also important issues to hospitals. However, little attention has been given to the interface of obesity and the elderly, largely due to the dearth of studies that include elderly patients. An aging population leads to an increase in geriatric syndromes, such as osteoporosis³ and its most devastating complication, hip fracture.⁴ These frail, hip‐fracture patients pose management challenges to practicing geriatricians and hospitalists.^5,6 Furthermore, although fracture risk is inversely correlated to body mass index (BMI),⁷ this relationship has yet to be fully examined in the postoperative hip‐fracture population. In other surgical settings, there is disagreement as to whether underweight or obese patients are at higher risk of developing medical complications,^8‐11 but for orthopedic patients, data have been limited to elective orthopedic populations.^12‐14 We previously demonstrated that underweight hip‐fracture patients are at higher risk of postoperative cardiac complications at 1 year,¹⁵ consistent with studies of cardiac risk indices determining long‐term events. Because of different pathophysiologic mechanisms, the purpose of this study was to ascertain the influence of BMI on inpatient postoperative noncardiac medical complications and to assess predictors of such complications following urgent hip fracture repair.

Patients and Methods

All Olmsted County, Minnesota, residents undergoing urgent hip repair due to fracture were identified using the Rochester Epidemiology Project, a medical‐record linkage system funded by the Federal government since 1966 to support disease‐related epidemiology studies.¹⁶ All patient medical care is indexed, and both inpatient and outpatient visits are captured and available for review, allowing for complete case ascertainment. Medical care in Olmsted County is primarily provided by Mayo Clinic with its affiliated hospitals (St. Mary's and Rochester Methodist) and the Olmsted Medical Center, in addition to a few individual providers. Over 95% of all Olmsted County hip fracture surgeries are ultimately managed at St. Mary's Hospital.

Following approval by the Institutional Review Board we used this unique data resource to identify all residents with an International Classification of Diseases, 9th edition (ICD‐9) diagnosis code of 820 to 829 for hip fracture (n = 1310). Both sexes were included, and all patients included in the study provided research authorization for use of their medical records for research purposes.¹⁷ We excluded patients who were managed conservatively (n = 56), had a pathological fracture (n = 20), had multiple injuries (n = 19), were operated on >72 hours after fracture (n = 5), were aged <65 years (n = 2), or were admitted for reasons other than a fracture and experienced an in‐hospital fracture (n = 3). We subsequently excluded patients with missing information (n = 10). World Health Organization (WHO) criteria were used for classifying BMI: underweight (BMI < 18.5); normal (BMI = 18.5‐24.9); overweight (BMI = 25.0‐29.9); and obese (BMI 30.0).¹⁸

All data were abstracted using standardized collection forms by trained nurse abstractors blinded to the study hypothesis. Patients' admission height and weight were documented; if unavailable, the nearest data within 2 months prior to surgery were recorded. Patients' preadmission residence, functional status, baseline comorbidities, admission medications, discharge destination, as well as whether patients had an intensive care unit stay or any major surgeries in the past 90 days were abstracted. In addition, American Society of Anesthesia (ASA) class, type of anesthesia, and length of stay were also obtained. Inpatient complications that had been identified by the treating physicians and documented in the medical record or identified on imaging studies were assessed from the time of hip fracture repair to the time of discharge using standardized clinical criteria (Table 1). For criteria that were based on either objective findings or clinical documentation/suspicion, the patient was considered to meet the criteria of having a complication if they fulfilled either one. We did not include any cardiac outcomes, including congestive heart failure, angina, myocardial infarction, or arrhythmias that had been previously reported.¹⁵ Noncardiac complications were classified broadly: respiratory (respiratory failure, respiratory depression, or pulmonary hypoxemia); neurologic (any cerebral event including hemorrhagic or ischemic stroke, transient ischemic attack, or delirium); gastrointestinal (ileus or gastrointestinal bleeding); vascular (pulmonary embolus, or deep vein thrombosis); infectious (pneumonia, sepsis, urinary tract, wound, or cellulitis); renal/metabolic (acute renal failure, dehydration, or electrolyte abnormalities); or other (fractures or falls).

Continuous data are presented as means standard deviation and categorical data as counts and percentages. In testing for differences in patient demographics, past medical history, and baseline clinical data among BMI groups, Kruskal‐Wallis tests were performed for continuous variables and Fisher's Exact or Cochran‐Mantel‐Haenszel tests were used for discrete variables. Bonferroni adjustments were performed where appropriate. The primary outcome was the risk of any noncardiac medical complication during the postoperative hospitalization, based on patients with complications. Incidence rates were calculated for the overall group as well as for each BMI category. BMI was evaluated categorically according to the WHO criteria, as a continuous variable dichotomized as a BMI 18.5 kg/m² to 24.9 kg/m² (normal) vs. all others, and above/below 25.0 kg/m². The effect of BMI and other potential risk factors on the complication rate was evaluated using logistic regression. The effect of BMI category on the overall complication rate was adjusted for the a priori risk factors of age, sex, surgical year, and ASA class both univariately (Model 1) and multivariately (Model 2). In addition to these variables, we also evaluated other potential risk factors, including baseline demographic and baseline clinical variables that were significant (P < 0.05) univariately using a stepwise selection; first forcing in BMI as a categorical variable (Model 3), then repeating the stepwise selection process without forcing in BMI (Model 4). Using data from Lawrence et al.,¹⁹ we estimated that we would have 80% power to detect differences in rates of inpatient noncardiac complications equal to an odds ratio (OR) = 2.2 (normal vs. underweight), OR = 2.0 (normal vs. overweight), and OR = 2.4 (normal vs. obese). Finally, because of power considerations, as an exploratory analysis, we additionally identified predictors of inpatient complications within each BMI category using stepwise selection. All statistical tests were 2‐sided, and P values <0.05 were considered significant. All analyses were performed using SAS for UNIX (version 9.1.3; SAS Institute, Inc., Cary, NC).

Results

Between 1988 and 2002, 1195 urgent repairs for hip fracture met our inclusion/exclusion criteria. We subsequently excluded 15 repairs with missing BMI data, and, of the 7 patients with >1 repair, we included only their first fracture episode in our analysis. Two were subsequently excluded due to an administrative error. Ultimately, 1180 hip fracture repairs were included in the analysis cohort. There were 184 (15.6%) patients in the underweight group, 640 (54.2%) with normal BMI, 251 (21.3%) with a BMI 25.0 to 29.9 kg/m², and 105 (8.9%) with a BMI 30 kg/m². Baseline characteristics are otherwise shown in Table 2. Normal BMI patients were significantly older than the other groups, and underweight patients were less likely to be admitted from home. Past history of having a cardiovascular risk factor or a cardiovascular diagnosis appeared to increase with increasing BMI. Underweight patients were more likely to have chronic obstructive pulmonary disease (COPD) than patients with normal BMI (P = 0.03) or overweight patients (P = 0.009), but not more than obese patients (P = 0.21). There were no differences across BMI groups in ASA class, type of anesthesia, intensive care unit stay, or length of stay.

There were 77 (41.8%) postoperative inpatient noncardiac complications in the underweight group, 234 (36.6%) in the normal BMI group, 90 (35.9%) in the overweight group, and 42 (40.0%) in the obese group (P = 0.49). Figure 1 demonstrates the main subcategory complication rates by BMI group, and Table 3 outlines the univariate unadjusted complication rates. Other than gastrointestinal complications being more prevalent as BMI increases (P = 0.005), there were no significant differences in crude complication rates across BMI categories (all P > 0.05) for the other complication subcategories. A multiple comparisons analysis did not demonstrate any differences between normal and any of the other BMI categories for ileus. Normal BMI patients were more likely to be discharged to a nursing facility than overweight or obese patients (85.5% vs. 79.3%, P = 0.03; and 85.5% vs. 79.0%, P = 0.03, respectively). The proportion of in‐hospital deaths among underweight patients was significantly higher than in any of the other groups (9.3% vs. 4.4%; P = 0.01), but mean length of stay was not significantly different.

Figure 1

Significant univariate predictors of the composite outcome of any noncardiac complication included: age (OR, 1.04 95% confidence interval [CI>], 1.02‐1.06; P < 0.001), age 75 years (OR, 2.25; 95% CI, 1.52‐3.33; P < 0.001), age 85 years (OR, 1.49; 95% CI, 1.17‐1.89; P < 0.001), male sex (OR, 1.41; 95% CI, 1.05‐1.90; P = 0.02), admission from home (OR, 0.77; 95% CI, 0.61‐0.98; P = 0.03), a history of cerebrovascular disease (OR, 1.41; 95% CI, 1.08‐1.83; P = 0.01), myocardial infarction (OR, 1.41; 95% CI, 1.07‐1.86; P = 0.02), angina (OR, 1.32; 95% CI, 1.03‐1.69; P = 0.03), congestive heart failure (OR, 1.45; 95% CI, 1.11‐1.89; P = 0.006), dementia (OR, 1.39; 95% CI, 1.08‐1.78; P = 0.01), peripheral vascular disease (OR, 1.47; 95% CI, 1.06‐2.03; P = 0.02), COPD/asthma (OR, 1.56; 95% CI, 1.18‐2.08; P = 0.002), osteoarthritis (OR, 1.29; 95% CI, 1.01‐1.65; P = 0.04), code status as Do Not Resuscitate (OR, 0.74; 95% CI, 0.58‐0.94; P = 0.015), or ASA class III‐V (OR, 2.24; 95% CI, 1.53‐3.29; P < 0.001). Results were no different after using the Charlson comorbidity index in place of ASA class (data not shown). No significant differences in overall noncardiac complications were observed when examining BMI as a continuous variable, as a categorical variable, as 25 kg/m² vs. <25 kg/m², or as 18.5 kg/m² to 24.9 kg/m² vs. all others. Examining renal, respiratory, peripheral vascular, or neurologic complications univariately within these aforementioned strata also did not demonstrate any significant differences among BMI categories (data not shown).

Multivariable analyses (Models 1‐4) are shown for any overall noncardiac inpatient medical complication in Table 4. BMI was not a significant predictor in any of our models, specifically in our main model that examined the effect of BMI adjusting for a priori variables (Model 2). However, older age, male sex, and ASA class were highly significant predictors of complications in all four models; however, surgical year was nonsignificant. Notably, after stepwise selection for other demographic and premorbid variables, a history of COPD or asthma was found to be an additional significant factor both in Model 3 (forcing BMI in the model) and Model 4 (without BMI in the model). Exploratory analysis of individual predictors of inpatient noncardiac complications within each BMI category demonstrated that, in underweight patients, admission use of ‐blockers was a significant predictor of having any medical complication (OR, 3.1; 95% CI, 1.1‐8.60; P = 0.03). In normal BMI patients, age 75 years (OR, 2.6; 95% CI, 1.4‐4.9; P = 0.003), ASA class III‐V (OR, 2.3; 95% CI, 1.3‐3.9; P = 0.003), and a history of cerebrovascular disease (OR, 1.5; 95%CI, 1.04‐2.1; P = 0.03) were predictors; and, in obese patients, only age (OR, 1.1; 95% CI, 1.00‐1.12; P = 0.05) was significant. There were no significant predictors of having a medical complication in the overweight group.

Discussion

Most research describing the association of BMI with postoperative outcomes has concentrated on cardiac surgery, general surgical procedures, and intensive care unit utilization.^8‐11,20 In the orthopedic literature, an elevated BMI has been associated with a higher number of short‐term complications, but this was limited to elective knee arthroplasty and spine surgery populations.^12,13,21 Conversely, no differences were observed in obese patients undergoing hip arthroplasties.^14,22 To the best of our knowledge, this study may be the first to examine the impact of BMI on inpatient hospital outcomes following urgent hip fracture repair. Our results suggest the risk of developing a noncardiac medical complication is the same regardless of BMI.

Our overall complication rate was higher (38%) than previous reports by others.^19,23‐26 Thus, Lawrence et al.,¹⁹ in their retrospective study of 20 facilities, demonstrated an overall complication rate of 17%, even though they also included postoperative cardiac complications. Although their study period overlapped our own (1982‐1993), they additionally included patients aged 60 to 65 years, a population known to have fewer comorbidities and fewer postoperative complications than the elderly hip‐fracture patients studied here. In addition, their population may have been healthier at baseline, in that a higher proportion lived at home (73%) and a lower percentage were ASA class III‐V (71%) than our cohort. These differences in baseline characteristics may explain the higher complication rates observed in our study.

Our findings did not suggest any relationship of BMI with noncardiac postoperative medical complications in any of the 4 methods we used to stratify BMI (continuous, categorical, normal vs. abnormal, and 25 kg/m²). Evidence is contradictory as to what the effect of BMI has on postoperative complications. An elevated BMI (30 kg/m²) has been shown to lead to increased sternal wound infection and saphenous vein harvest infection in a cardiac surgery population,²⁷ but other studies^10,28,29 have demonstrated the opposite effect. Among 6336 patients undergoing elective general surgery procedures, the incidence of complications were similar by body mass.³⁰ A matched study design that included urgent and emergent surgeries also did not find any appreciable increased perioperative risk in noncardiac surgery.²⁸ Whether this may be due to the elective nature of the surgeries in these studies, hence leading to selection bias, is unknown.

In geriatric patients, multiple baseline comorbid conditions often are reflected in a higher ASA class, which increases the risk of significant perioperative complications. Our multivariate modeling showed that a high ASA class strongly predicts morbidity and mortality following hip fracture repair, in line with other studies.^19,31,32 Although the Charlson comorbidity index could alternatively been used, we elected to adjust for ASA class as it is more commonly used and is simple to use. Interestingly, surgical year did not significantly predict any complication, which can suggest that practice changes play a minimal impact on patient outcomes. However, we caution that because the individual event rates, particularly vascular, were low, we were unable to fully determine whether changes in practice management, such as improved thromboprophylaxis, would impact event rates over time. Finally, other predictors such as older age³³ and a concomitant history of either COPD or asthma,³⁴ are well‐accepted predictors of inpatient complications. Our attempt to examine specific predictors of complications in each BMI category revealed differing results, making interpretations difficult. Because of power considerations, this was meant solely as an exploratory analysis, and larger cohorts are needed to further ascertain whether predictors are different in these groups. Such a study may in fact identify perioperative issues that allow practitioners caring for this population to modify these factors.

One of the major limitations in our study was our inability to adjust for individual complications using multivariable models, such as deep vein thrombosis or delirium, within each BMI stratum, because of statistical power issues. Such a study would require large numbers of individual complications or events to allow for appropriate adjustments. The authors acknowledge that such individual complication rates may vary dramatically. We were aware of this potential problem, and therefore a priori ascertained a composite outcome of any noncardiac medical complication. However, our results do provide preliminary information regarding the impact of BMI on noncardiac medical complications. Further studies would be needed, though, to fully determine the effect of BMI on the number of cases with each complication.

Obesity (or BMI) is a known cardiovascular risk factor, and our previous study's aim was to determine cardiovascular events in a comparable manner to the way risk indices, such as the Goldman, Lee, or the AHA preoperative algorithm function. The surgical literature often presents noncardiac complications separately, allowing us to directly compare our own data to other published studies. We used 2 separate approaches, focusing on the inpatient stay (ascertaining noncardiac complications) and 1‐year cardiac outcomes (cardiac complications), as these are mediated by different mechanisms and factors. Furthermore, the intent of both studies was to dispel any concerns that an elevated BMI would in fact lead to an increased number of complications. Whether cardiac complications, though, would impact noncardiac complications, or vice‐versa, is unknown, and would require further investigation.

Although we relied on well‐established definitions for body mass, they have often been challenged, as they may underestimate adiposity in the elderly population due to age‐related reductions in lean mass.^35,36 Studies have demonstrated a poor correlation between percent body fat and BMI in the >65 year age group,³⁷ which could impact our results and outcomes by misclassifying patients. Yet, as an anthropometric measurement, BMI is easily obtainable and its variables are routinely documented in patients' medical records, as compared to other anthropometric measurements. Other means of estimating adiposity, such as densitometry or computed tomography (CT) scanning, are impractical, expensive, and not used clinically but routinely in research settings. The lack of standardization in obtaining height and weight, despite nurse‐initiated protocols for bed calibration, may have introduced a degree of measurement bias. Furthermore, the extent of lean mass lost and volume status changes lead to further challenges of using BMI in hospital settings. Whether other anthropometric measurements, including hip circumference, waist circumference, or waist‐hip ratio, should be used in this group of patients requires further examination. However, despite its shortcomings in elderly patients, BMI is still deemed an appropriate surrogate for obesity.

Our main strength was the use of the Rochester Epidemiology Project medical record linkage system to ascertain all patient data. This focuses on patients from a single geographically‐defined community minimizing referral biases often observed in studies originating from a tertiary care referral center. Previous disease‐related epidemiology studies using the Olmsted County population have demonstrated excellent external validity to the U.S. white population.¹⁶ We relied on the medical documentation of the treating clinician for many diagnoses in our data abstraction. Although we attempted to use standardized definitions, clinicians may have inadvertently forgotten to document subjective signs or symptoms that would assist in the categorization of these complications. Hence, added inpatient complications may have been overlooked, suggesting that our results may slightly underestimate the true incidence in this population. Additionally, certain complications may overlap categories, such as pneumonia and infections. We agree with Lawrence et al.¹⁹ that long periods of time are necessary to accumulate data of this kind in an effort to describe complication rates epidemiologically.

Despite no difference in outcomes among BMI categories, our results have striking implications for the hospitalized patient. Thus, underweight elderly patients, often considered frail with minimal functional reserve, are at no higher risk for developing inpatient medical complications than patients with higher BMIs. This is contrary to our study focusing on cardiac complications, where underweight patients were at higher risk.¹⁵ Conversely, obese patients, who have been demonstrated to be at higher risk of medical complications (particularly pulmonary), had no greater risk than patients with normal BMI. To the practicing geriatrician and hospitalist, this information provides important prognostication regarding additional perioperative measures that need to be implemented in these different groups. Based on our results, BMI does not play a particular role in noncardiac medical complications, dispelling any myths of the added burden of excess weight on surgical outcomes in this population. From a hospital perspective, this may be important since additional testing or preventative management in these patients may lead to additional resource use. However, in‐hospital deaths were higher in underweight patients than in patients with a normal BMI. Although we were underpowered to detect any differences in mortality between groups and could therefore not adjust for additional variables, it is unknown whether cardiac or noncardiac complications may be a stronger predictor of death in the underweight patient population. Further studies would be needed to better ascertain this relationship.

Conclusions

In elderly patients undergoing urgent hip fracture repair, BMI does not appear to lead to an excess rate of inpatient noncardiac complications. Our results are the first to demonstrate that BMI has no impact on morbidity in this patient population. Further research on the influence of body composition on inpatient complications in this population is needed to accurately allow for appropriate perioperative prophylaxis. Whether BMI impacts specific complications or in‐patient mortality in this population still requires investigation.

Public health concerns such as the aging population¹ and the increasing prevalence of obesity² are also important issues to hospitals. However, little attention has been given to the interface of obesity and the elderly, largely due to the dearth of studies that include elderly patients. An aging population leads to an increase in geriatric syndromes, such as osteoporosis³ and its most devastating complication, hip fracture.⁴ These frail, hip‐fracture patients pose management challenges to practicing geriatricians and hospitalists.^5,6 Furthermore, although fracture risk is inversely correlated to body mass index (BMI),⁷ this relationship has yet to be fully examined in the postoperative hip‐fracture population. In other surgical settings, there is disagreement as to whether underweight or obese patients are at higher risk of developing medical complications,^8‐11 but for orthopedic patients, data have been limited to elective orthopedic populations.^12‐14 We previously demonstrated that underweight hip‐fracture patients are at higher risk of postoperative cardiac complications at 1 year,¹⁵ consistent with studies of cardiac risk indices determining long‐term events. Because of different pathophysiologic mechanisms, the purpose of this study was to ascertain the influence of BMI on inpatient postoperative noncardiac medical complications and to assess predictors of such complications following urgent hip fracture repair.

Patients and Methods

All Olmsted County, Minnesota, residents undergoing urgent hip repair due to fracture were identified using the Rochester Epidemiology Project, a medical‐record linkage system funded by the Federal government since 1966 to support disease‐related epidemiology studies.¹⁶ All patient medical care is indexed, and both inpatient and outpatient visits are captured and available for review, allowing for complete case ascertainment. Medical care in Olmsted County is primarily provided by Mayo Clinic with its affiliated hospitals (St. Mary's and Rochester Methodist) and the Olmsted Medical Center, in addition to a few individual providers. Over 95% of all Olmsted County hip fracture surgeries are ultimately managed at St. Mary's Hospital.

Following approval by the Institutional Review Board we used this unique data resource to identify all residents with an International Classification of Diseases, 9th edition (ICD‐9) diagnosis code of 820 to 829 for hip fracture (n = 1310). Both sexes were included, and all patients included in the study provided research authorization for use of their medical records for research purposes.¹⁷ We excluded patients who were managed conservatively (n = 56), had a pathological fracture (n = 20), had multiple injuries (n = 19), were operated on >72 hours after fracture (n = 5), were aged <65 years (n = 2), or were admitted for reasons other than a fracture and experienced an in‐hospital fracture (n = 3). We subsequently excluded patients with missing information (n = 10). World Health Organization (WHO) criteria were used for classifying BMI: underweight (BMI < 18.5); normal (BMI = 18.5‐24.9); overweight (BMI = 25.0‐29.9); and obese (BMI 30.0).¹⁸

All data were abstracted using standardized collection forms by trained nurse abstractors blinded to the study hypothesis. Patients' admission height and weight were documented; if unavailable, the nearest data within 2 months prior to surgery were recorded. Patients' preadmission residence, functional status, baseline comorbidities, admission medications, discharge destination, as well as whether patients had an intensive care unit stay or any major surgeries in the past 90 days were abstracted. In addition, American Society of Anesthesia (ASA) class, type of anesthesia, and length of stay were also obtained. Inpatient complications that had been identified by the treating physicians and documented in the medical record or identified on imaging studies were assessed from the time of hip fracture repair to the time of discharge using standardized clinical criteria (Table 1). For criteria that were based on either objective findings or clinical documentation/suspicion, the patient was considered to meet the criteria of having a complication if they fulfilled either one. We did not include any cardiac outcomes, including congestive heart failure, angina, myocardial infarction, or arrhythmias that had been previously reported.¹⁵ Noncardiac complications were classified broadly: respiratory (respiratory failure, respiratory depression, or pulmonary hypoxemia); neurologic (any cerebral event including hemorrhagic or ischemic stroke, transient ischemic attack, or delirium); gastrointestinal (ileus or gastrointestinal bleeding); vascular (pulmonary embolus, or deep vein thrombosis); infectious (pneumonia, sepsis, urinary tract, wound, or cellulitis); renal/metabolic (acute renal failure, dehydration, or electrolyte abnormalities); or other (fractures or falls).

Continuous data are presented as means standard deviation and categorical data as counts and percentages. In testing for differences in patient demographics, past medical history, and baseline clinical data among BMI groups, Kruskal‐Wallis tests were performed for continuous variables and Fisher's Exact or Cochran‐Mantel‐Haenszel tests were used for discrete variables. Bonferroni adjustments were performed where appropriate. The primary outcome was the risk of any noncardiac medical complication during the postoperative hospitalization, based on patients with complications. Incidence rates were calculated for the overall group as well as for each BMI category. BMI was evaluated categorically according to the WHO criteria, as a continuous variable dichotomized as a BMI 18.5 kg/m² to 24.9 kg/m² (normal) vs. all others, and above/below 25.0 kg/m². The effect of BMI and other potential risk factors on the complication rate was evaluated using logistic regression. The effect of BMI category on the overall complication rate was adjusted for the a priori risk factors of age, sex, surgical year, and ASA class both univariately (Model 1) and multivariately (Model 2). In addition to these variables, we also evaluated other potential risk factors, including baseline demographic and baseline clinical variables that were significant (P < 0.05) univariately using a stepwise selection; first forcing in BMI as a categorical variable (Model 3), then repeating the stepwise selection process without forcing in BMI (Model 4). Using data from Lawrence et al.,¹⁹ we estimated that we would have 80% power to detect differences in rates of inpatient noncardiac complications equal to an odds ratio (OR) = 2.2 (normal vs. underweight), OR = 2.0 (normal vs. overweight), and OR = 2.4 (normal vs. obese). Finally, because of power considerations, as an exploratory analysis, we additionally identified predictors of inpatient complications within each BMI category using stepwise selection. All statistical tests were 2‐sided, and P values <0.05 were considered significant. All analyses were performed using SAS for UNIX (version 9.1.3; SAS Institute, Inc., Cary, NC).

Results

Between 1988 and 2002, 1195 urgent repairs for hip fracture met our inclusion/exclusion criteria. We subsequently excluded 15 repairs with missing BMI data, and, of the 7 patients with >1 repair, we included only their first fracture episode in our analysis. Two were subsequently excluded due to an administrative error. Ultimately, 1180 hip fracture repairs were included in the analysis cohort. There were 184 (15.6%) patients in the underweight group, 640 (54.2%) with normal BMI, 251 (21.3%) with a BMI 25.0 to 29.9 kg/m², and 105 (8.9%) with a BMI 30 kg/m². Baseline characteristics are otherwise shown in Table 2. Normal BMI patients were significantly older than the other groups, and underweight patients were less likely to be admitted from home. Past history of having a cardiovascular risk factor or a cardiovascular diagnosis appeared to increase with increasing BMI. Underweight patients were more likely to have chronic obstructive pulmonary disease (COPD) than patients with normal BMI (P = 0.03) or overweight patients (P = 0.009), but not more than obese patients (P = 0.21). There were no differences across BMI groups in ASA class, type of anesthesia, intensive care unit stay, or length of stay.

There were 77 (41.8%) postoperative inpatient noncardiac complications in the underweight group, 234 (36.6%) in the normal BMI group, 90 (35.9%) in the overweight group, and 42 (40.0%) in the obese group (P = 0.49). Figure 1 demonstrates the main subcategory complication rates by BMI group, and Table 3 outlines the univariate unadjusted complication rates. Other than gastrointestinal complications being more prevalent as BMI increases (P = 0.005), there were no significant differences in crude complication rates across BMI categories (all P > 0.05) for the other complication subcategories. A multiple comparisons analysis did not demonstrate any differences between normal and any of the other BMI categories for ileus. Normal BMI patients were more likely to be discharged to a nursing facility than overweight or obese patients (85.5% vs. 79.3%, P = 0.03; and 85.5% vs. 79.0%, P = 0.03, respectively). The proportion of in‐hospital deaths among underweight patients was significantly higher than in any of the other groups (9.3% vs. 4.4%; P = 0.01), but mean length of stay was not significantly different.

Figure 1

Significant univariate predictors of the composite outcome of any noncardiac complication included: age (OR, 1.04 95% confidence interval [CI>], 1.02‐1.06; P < 0.001), age 75 years (OR, 2.25; 95% CI, 1.52‐3.33; P < 0.001), age 85 years (OR, 1.49; 95% CI, 1.17‐1.89; P < 0.001), male sex (OR, 1.41; 95% CI, 1.05‐1.90; P = 0.02), admission from home (OR, 0.77; 95% CI, 0.61‐0.98; P = 0.03), a history of cerebrovascular disease (OR, 1.41; 95% CI, 1.08‐1.83; P = 0.01), myocardial infarction (OR, 1.41; 95% CI, 1.07‐1.86; P = 0.02), angina (OR, 1.32; 95% CI, 1.03‐1.69; P = 0.03), congestive heart failure (OR, 1.45; 95% CI, 1.11‐1.89; P = 0.006), dementia (OR, 1.39; 95% CI, 1.08‐1.78; P = 0.01), peripheral vascular disease (OR, 1.47; 95% CI, 1.06‐2.03; P = 0.02), COPD/asthma (OR, 1.56; 95% CI, 1.18‐2.08; P = 0.002), osteoarthritis (OR, 1.29; 95% CI, 1.01‐1.65; P = 0.04), code status as Do Not Resuscitate (OR, 0.74; 95% CI, 0.58‐0.94; P = 0.015), or ASA class III‐V (OR, 2.24; 95% CI, 1.53‐3.29; P < 0.001). Results were no different after using the Charlson comorbidity index in place of ASA class (data not shown). No significant differences in overall noncardiac complications were observed when examining BMI as a continuous variable, as a categorical variable, as 25 kg/m² vs. <25 kg/m², or as 18.5 kg/m² to 24.9 kg/m² vs. all others. Examining renal, respiratory, peripheral vascular, or neurologic complications univariately within these aforementioned strata also did not demonstrate any significant differences among BMI categories (data not shown).

Multivariable analyses (Models 1‐4) are shown for any overall noncardiac inpatient medical complication in Table 4. BMI was not a significant predictor in any of our models, specifically in our main model that examined the effect of BMI adjusting for a priori variables (Model 2). However, older age, male sex, and ASA class were highly significant predictors of complications in all four models; however, surgical year was nonsignificant. Notably, after stepwise selection for other demographic and premorbid variables, a history of COPD or asthma was found to be an additional significant factor both in Model 3 (forcing BMI in the model) and Model 4 (without BMI in the model). Exploratory analysis of individual predictors of inpatient noncardiac complications within each BMI category demonstrated that, in underweight patients, admission use of ‐blockers was a significant predictor of having any medical complication (OR, 3.1; 95% CI, 1.1‐8.60; P = 0.03). In normal BMI patients, age 75 years (OR, 2.6; 95% CI, 1.4‐4.9; P = 0.003), ASA class III‐V (OR, 2.3; 95% CI, 1.3‐3.9; P = 0.003), and a history of cerebrovascular disease (OR, 1.5; 95%CI, 1.04‐2.1; P = 0.03) were predictors; and, in obese patients, only age (OR, 1.1; 95% CI, 1.00‐1.12; P = 0.05) was significant. There were no significant predictors of having a medical complication in the overweight group.

Discussion

Most research describing the association of BMI with postoperative outcomes has concentrated on cardiac surgery, general surgical procedures, and intensive care unit utilization.^8‐11,20 In the orthopedic literature, an elevated BMI has been associated with a higher number of short‐term complications, but this was limited to elective knee arthroplasty and spine surgery populations.^12,13,21 Conversely, no differences were observed in obese patients undergoing hip arthroplasties.^14,22 To the best of our knowledge, this study may be the first to examine the impact of BMI on inpatient hospital outcomes following urgent hip fracture repair. Our results suggest the risk of developing a noncardiac medical complication is the same regardless of BMI.

Our overall complication rate was higher (38%) than previous reports by others.^19,23‐26 Thus, Lawrence et al.,¹⁹ in their retrospective study of 20 facilities, demonstrated an overall complication rate of 17%, even though they also included postoperative cardiac complications. Although their study period overlapped our own (1982‐1993), they additionally included patients aged 60 to 65 years, a population known to have fewer comorbidities and fewer postoperative complications than the elderly hip‐fracture patients studied here. In addition, their population may have been healthier at baseline, in that a higher proportion lived at home (73%) and a lower percentage were ASA class III‐V (71%) than our cohort. These differences in baseline characteristics may explain the higher complication rates observed in our study.

Our findings did not suggest any relationship of BMI with noncardiac postoperative medical complications in any of the 4 methods we used to stratify BMI (continuous, categorical, normal vs. abnormal, and 25 kg/m²). Evidence is contradictory as to what the effect of BMI has on postoperative complications. An elevated BMI (30 kg/m²) has been shown to lead to increased sternal wound infection and saphenous vein harvest infection in a cardiac surgery population,²⁷ but other studies^10,28,29 have demonstrated the opposite effect. Among 6336 patients undergoing elective general surgery procedures, the incidence of complications were similar by body mass.³⁰ A matched study design that included urgent and emergent surgeries also did not find any appreciable increased perioperative risk in noncardiac surgery.²⁸ Whether this may be due to the elective nature of the surgeries in these studies, hence leading to selection bias, is unknown.

In geriatric patients, multiple baseline comorbid conditions often are reflected in a higher ASA class, which increases the risk of significant perioperative complications. Our multivariate modeling showed that a high ASA class strongly predicts morbidity and mortality following hip fracture repair, in line with other studies.^19,31,32 Although the Charlson comorbidity index could alternatively been used, we elected to adjust for ASA class as it is more commonly used and is simple to use. Interestingly, surgical year did not significantly predict any complication, which can suggest that practice changes play a minimal impact on patient outcomes. However, we caution that because the individual event rates, particularly vascular, were low, we were unable to fully determine whether changes in practice management, such as improved thromboprophylaxis, would impact event rates over time. Finally, other predictors such as older age³³ and a concomitant history of either COPD or asthma,³⁴ are well‐accepted predictors of inpatient complications. Our attempt to examine specific predictors of complications in each BMI category revealed differing results, making interpretations difficult. Because of power considerations, this was meant solely as an exploratory analysis, and larger cohorts are needed to further ascertain whether predictors are different in these groups. Such a study may in fact identify perioperative issues that allow practitioners caring for this population to modify these factors.

One of the major limitations in our study was our inability to adjust for individual complications using multivariable models, such as deep vein thrombosis or delirium, within each BMI stratum, because of statistical power issues. Such a study would require large numbers of individual complications or events to allow for appropriate adjustments. The authors acknowledge that such individual complication rates may vary dramatically. We were aware of this potential problem, and therefore a priori ascertained a composite outcome of any noncardiac medical complication. However, our results do provide preliminary information regarding the impact of BMI on noncardiac medical complications. Further studies would be needed, though, to fully determine the effect of BMI on the number of cases with each complication.

Obesity (or BMI) is a known cardiovascular risk factor, and our previous study's aim was to determine cardiovascular events in a comparable manner to the way risk indices, such as the Goldman, Lee, or the AHA preoperative algorithm function. The surgical literature often presents noncardiac complications separately, allowing us to directly compare our own data to other published studies. We used 2 separate approaches, focusing on the inpatient stay (ascertaining noncardiac complications) and 1‐year cardiac outcomes (cardiac complications), as these are mediated by different mechanisms and factors. Furthermore, the intent of both studies was to dispel any concerns that an elevated BMI would in fact lead to an increased number of complications. Whether cardiac complications, though, would impact noncardiac complications, or vice‐versa, is unknown, and would require further investigation.

Although we relied on well‐established definitions for body mass, they have often been challenged, as they may underestimate adiposity in the elderly population due to age‐related reductions in lean mass.^35,36 Studies have demonstrated a poor correlation between percent body fat and BMI in the >65 year age group,³⁷ which could impact our results and outcomes by misclassifying patients. Yet, as an anthropometric measurement, BMI is easily obtainable and its variables are routinely documented in patients' medical records, as compared to other anthropometric measurements. Other means of estimating adiposity, such as densitometry or computed tomography (CT) scanning, are impractical, expensive, and not used clinically but routinely in research settings. The lack of standardization in obtaining height and weight, despite nurse‐initiated protocols for bed calibration, may have introduced a degree of measurement bias. Furthermore, the extent of lean mass lost and volume status changes lead to further challenges of using BMI in hospital settings. Whether other anthropometric measurements, including hip circumference, waist circumference, or waist‐hip ratio, should be used in this group of patients requires further examination. However, despite its shortcomings in elderly patients, BMI is still deemed an appropriate surrogate for obesity.

Our main strength was the use of the Rochester Epidemiology Project medical record linkage system to ascertain all patient data. This focuses on patients from a single geographically‐defined community minimizing referral biases often observed in studies originating from a tertiary care referral center. Previous disease‐related epidemiology studies using the Olmsted County population have demonstrated excellent external validity to the U.S. white population.¹⁶ We relied on the medical documentation of the treating clinician for many diagnoses in our data abstraction. Although we attempted to use standardized definitions, clinicians may have inadvertently forgotten to document subjective signs or symptoms that would assist in the categorization of these complications. Hence, added inpatient complications may have been overlooked, suggesting that our results may slightly underestimate the true incidence in this population. Additionally, certain complications may overlap categories, such as pneumonia and infections. We agree with Lawrence et al.¹⁹ that long periods of time are necessary to accumulate data of this kind in an effort to describe complication rates epidemiologically.

Despite no difference in outcomes among BMI categories, our results have striking implications for the hospitalized patient. Thus, underweight elderly patients, often considered frail with minimal functional reserve, are at no higher risk for developing inpatient medical complications than patients with higher BMIs. This is contrary to our study focusing on cardiac complications, where underweight patients were at higher risk.¹⁵ Conversely, obese patients, who have been demonstrated to be at higher risk of medical complications (particularly pulmonary), had no greater risk than patients with normal BMI. To the practicing geriatrician and hospitalist, this information provides important prognostication regarding additional perioperative measures that need to be implemented in these different groups. Based on our results, BMI does not play a particular role in noncardiac medical complications, dispelling any myths of the added burden of excess weight on surgical outcomes in this population. From a hospital perspective, this may be important since additional testing or preventative management in these patients may lead to additional resource use. However, in‐hospital deaths were higher in underweight patients than in patients with a normal BMI. Although we were underpowered to detect any differences in mortality between groups and could therefore not adjust for additional variables, it is unknown whether cardiac or noncardiac complications may be a stronger predictor of death in the underweight patient population. Further studies would be needed to better ascertain this relationship.

Conclusions

In elderly patients undergoing urgent hip fracture repair, BMI does not appear to lead to an excess rate of inpatient noncardiac complications. Our results are the first to demonstrate that BMI has no impact on morbidity in this patient population. Further research on the influence of body composition on inpatient complications in this population is needed to accurately allow for appropriate perioperative prophylaxis. Whether BMI impacts specific complications or in‐patient mortality in this population still requires investigation.

Public health concerns such as the aging population¹ and the increasing prevalence of obesity² are also important issues to hospitals. However, little attention has been given to the interface of obesity and the elderly, largely due to the dearth of studies that include elderly patients. An aging population leads to an increase in geriatric syndromes, such as osteoporosis³ and its most devastating complication, hip fracture.⁴ These frail, hip‐fracture patients pose management challenges to practicing geriatricians and hospitalists.^5,6 Furthermore, although fracture risk is inversely correlated to body mass index (BMI),⁷ this relationship has yet to be fully examined in the postoperative hip‐fracture population. In other surgical settings, there is disagreement as to whether underweight or obese patients are at higher risk of developing medical complications,^8‐11 but for orthopedic patients, data have been limited to elective orthopedic populations.^12‐14 We previously demonstrated that underweight hip‐fracture patients are at higher risk of postoperative cardiac complications at 1 year,¹⁵ consistent with studies of cardiac risk indices determining long‐term events. Because of different pathophysiologic mechanisms, the purpose of this study was to ascertain the influence of BMI on inpatient postoperative noncardiac medical complications and to assess predictors of such complications following urgent hip fracture repair.

Patients and Methods

All Olmsted County, Minnesota, residents undergoing urgent hip repair due to fracture were identified using the Rochester Epidemiology Project, a medical‐record linkage system funded by the Federal government since 1966 to support disease‐related epidemiology studies.¹⁶ All patient medical care is indexed, and both inpatient and outpatient visits are captured and available for review, allowing for complete case ascertainment. Medical care in Olmsted County is primarily provided by Mayo Clinic with its affiliated hospitals (St. Mary's and Rochester Methodist) and the Olmsted Medical Center, in addition to a few individual providers. Over 95% of all Olmsted County hip fracture surgeries are ultimately managed at St. Mary's Hospital.

Following approval by the Institutional Review Board we used this unique data resource to identify all residents with an International Classification of Diseases, 9th edition (ICD‐9) diagnosis code of 820 to 829 for hip fracture (n = 1310). Both sexes were included, and all patients included in the study provided research authorization for use of their medical records for research purposes.¹⁷ We excluded patients who were managed conservatively (n = 56), had a pathological fracture (n = 20), had multiple injuries (n = 19), were operated on >72 hours after fracture (n = 5), were aged <65 years (n = 2), or were admitted for reasons other than a fracture and experienced an in‐hospital fracture (n = 3). We subsequently excluded patients with missing information (n = 10). World Health Organization (WHO) criteria were used for classifying BMI: underweight (BMI < 18.5); normal (BMI = 18.5‐24.9); overweight (BMI = 25.0‐29.9); and obese (BMI 30.0).¹⁸

All data were abstracted using standardized collection forms by trained nurse abstractors blinded to the study hypothesis. Patients' admission height and weight were documented; if unavailable, the nearest data within 2 months prior to surgery were recorded. Patients' preadmission residence, functional status, baseline comorbidities, admission medications, discharge destination, as well as whether patients had an intensive care unit stay or any major surgeries in the past 90 days were abstracted. In addition, American Society of Anesthesia (ASA) class, type of anesthesia, and length of stay were also obtained. Inpatient complications that had been identified by the treating physicians and documented in the medical record or identified on imaging studies were assessed from the time of hip fracture repair to the time of discharge using standardized clinical criteria (Table 1). For criteria that were based on either objective findings or clinical documentation/suspicion, the patient was considered to meet the criteria of having a complication if they fulfilled either one. We did not include any cardiac outcomes, including congestive heart failure, angina, myocardial infarction, or arrhythmias that had been previously reported.¹⁵ Noncardiac complications were classified broadly: respiratory (respiratory failure, respiratory depression, or pulmonary hypoxemia); neurologic (any cerebral event including hemorrhagic or ischemic stroke, transient ischemic attack, or delirium); gastrointestinal (ileus or gastrointestinal bleeding); vascular (pulmonary embolus, or deep vein thrombosis); infectious (pneumonia, sepsis, urinary tract, wound, or cellulitis); renal/metabolic (acute renal failure, dehydration, or electrolyte abnormalities); or other (fractures or falls).

Continuous data are presented as means standard deviation and categorical data as counts and percentages. In testing for differences in patient demographics, past medical history, and baseline clinical data among BMI groups, Kruskal‐Wallis tests were performed for continuous variables and Fisher's Exact or Cochran‐Mantel‐Haenszel tests were used for discrete variables. Bonferroni adjustments were performed where appropriate. The primary outcome was the risk of any noncardiac medical complication during the postoperative hospitalization, based on patients with complications. Incidence rates were calculated for the overall group as well as for each BMI category. BMI was evaluated categorically according to the WHO criteria, as a continuous variable dichotomized as a BMI 18.5 kg/m² to 24.9 kg/m² (normal) vs. all others, and above/below 25.0 kg/m². The effect of BMI and other potential risk factors on the complication rate was evaluated using logistic regression. The effect of BMI category on the overall complication rate was adjusted for the a priori risk factors of age, sex, surgical year, and ASA class both univariately (Model 1) and multivariately (Model 2). In addition to these variables, we also evaluated other potential risk factors, including baseline demographic and baseline clinical variables that were significant (P < 0.05) univariately using a stepwise selection; first forcing in BMI as a categorical variable (Model 3), then repeating the stepwise selection process without forcing in BMI (Model 4). Using data from Lawrence et al.,¹⁹ we estimated that we would have 80% power to detect differences in rates of inpatient noncardiac complications equal to an odds ratio (OR) = 2.2 (normal vs. underweight), OR = 2.0 (normal vs. overweight), and OR = 2.4 (normal vs. obese). Finally, because of power considerations, as an exploratory analysis, we additionally identified predictors of inpatient complications within each BMI category using stepwise selection. All statistical tests were 2‐sided, and P values <0.05 were considered significant. All analyses were performed using SAS for UNIX (version 9.1.3; SAS Institute, Inc., Cary, NC).

Results

Between 1988 and 2002, 1195 urgent repairs for hip fracture met our inclusion/exclusion criteria. We subsequently excluded 15 repairs with missing BMI data, and, of the 7 patients with >1 repair, we included only their first fracture episode in our analysis. Two were subsequently excluded due to an administrative error. Ultimately, 1180 hip fracture repairs were included in the analysis cohort. There were 184 (15.6%) patients in the underweight group, 640 (54.2%) with normal BMI, 251 (21.3%) with a BMI 25.0 to 29.9 kg/m², and 105 (8.9%) with a BMI 30 kg/m². Baseline characteristics are otherwise shown in Table 2. Normal BMI patients were significantly older than the other groups, and underweight patients were less likely to be admitted from home. Past history of having a cardiovascular risk factor or a cardiovascular diagnosis appeared to increase with increasing BMI. Underweight patients were more likely to have chronic obstructive pulmonary disease (COPD) than patients with normal BMI (P = 0.03) or overweight patients (P = 0.009), but not more than obese patients (P = 0.21). There were no differences across BMI groups in ASA class, type of anesthesia, intensive care unit stay, or length of stay.

There were 77 (41.8%) postoperative inpatient noncardiac complications in the underweight group, 234 (36.6%) in the normal BMI group, 90 (35.9%) in the overweight group, and 42 (40.0%) in the obese group (P = 0.49). Figure 1 demonstrates the main subcategory complication rates by BMI group, and Table 3 outlines the univariate unadjusted complication rates. Other than gastrointestinal complications being more prevalent as BMI increases (P = 0.005), there were no significant differences in crude complication rates across BMI categories (all P > 0.05) for the other complication subcategories. A multiple comparisons analysis did not demonstrate any differences between normal and any of the other BMI categories for ileus. Normal BMI patients were more likely to be discharged to a nursing facility than overweight or obese patients (85.5% vs. 79.3%, P = 0.03; and 85.5% vs. 79.0%, P = 0.03, respectively). The proportion of in‐hospital deaths among underweight patients was significantly higher than in any of the other groups (9.3% vs. 4.4%; P = 0.01), but mean length of stay was not significantly different.

Figure 1

Significant univariate predictors of the composite outcome of any noncardiac complication included: age (OR, 1.04 95% confidence interval [CI>], 1.02‐1.06; P < 0.001), age 75 years (OR, 2.25; 95% CI, 1.52‐3.33; P < 0.001), age 85 years (OR, 1.49; 95% CI, 1.17‐1.89; P < 0.001), male sex (OR, 1.41; 95% CI, 1.05‐1.90; P = 0.02), admission from home (OR, 0.77; 95% CI, 0.61‐0.98; P = 0.03), a history of cerebrovascular disease (OR, 1.41; 95% CI, 1.08‐1.83; P = 0.01), myocardial infarction (OR, 1.41; 95% CI, 1.07‐1.86; P = 0.02), angina (OR, 1.32; 95% CI, 1.03‐1.69; P = 0.03), congestive heart failure (OR, 1.45; 95% CI, 1.11‐1.89; P = 0.006), dementia (OR, 1.39; 95% CI, 1.08‐1.78; P = 0.01), peripheral vascular disease (OR, 1.47; 95% CI, 1.06‐2.03; P = 0.02), COPD/asthma (OR, 1.56; 95% CI, 1.18‐2.08; P = 0.002), osteoarthritis (OR, 1.29; 95% CI, 1.01‐1.65; P = 0.04), code status as Do Not Resuscitate (OR, 0.74; 95% CI, 0.58‐0.94; P = 0.015), or ASA class III‐V (OR, 2.24; 95% CI, 1.53‐3.29; P < 0.001). Results were no different after using the Charlson comorbidity index in place of ASA class (data not shown). No significant differences in overall noncardiac complications were observed when examining BMI as a continuous variable, as a categorical variable, as 25 kg/m² vs. <25 kg/m², or as 18.5 kg/m² to 24.9 kg/m² vs. all others. Examining renal, respiratory, peripheral vascular, or neurologic complications univariately within these aforementioned strata also did not demonstrate any significant differences among BMI categories (data not shown).

Multivariable analyses (Models 1‐4) are shown for any overall noncardiac inpatient medical complication in Table 4. BMI was not a significant predictor in any of our models, specifically in our main model that examined the effect of BMI adjusting for a priori variables (Model 2). However, older age, male sex, and ASA class were highly significant predictors of complications in all four models; however, surgical year was nonsignificant. Notably, after stepwise selection for other demographic and premorbid variables, a history of COPD or asthma was found to be an additional significant factor both in Model 3 (forcing BMI in the model) and Model 4 (without BMI in the model). Exploratory analysis of individual predictors of inpatient noncardiac complications within each BMI category demonstrated that, in underweight patients, admission use of ‐blockers was a significant predictor of having any medical complication (OR, 3.1; 95% CI, 1.1‐8.60; P = 0.03). In normal BMI patients, age 75 years (OR, 2.6; 95% CI, 1.4‐4.9; P = 0.003), ASA class III‐V (OR, 2.3; 95% CI, 1.3‐3.9; P = 0.003), and a history of cerebrovascular disease (OR, 1.5; 95%CI, 1.04‐2.1; P = 0.03) were predictors; and, in obese patients, only age (OR, 1.1; 95% CI, 1.00‐1.12; P = 0.05) was significant. There were no significant predictors of having a medical complication in the overweight group.

Discussion

Most research describing the association of BMI with postoperative outcomes has concentrated on cardiac surgery, general surgical procedures, and intensive care unit utilization.^8‐11,20 In the orthopedic literature, an elevated BMI has been associated with a higher number of short‐term complications, but this was limited to elective knee arthroplasty and spine surgery populations.^12,13,21 Conversely, no differences were observed in obese patients undergoing hip arthroplasties.^14,22 To the best of our knowledge, this study may be the first to examine the impact of BMI on inpatient hospital outcomes following urgent hip fracture repair. Our results suggest the risk of developing a noncardiac medical complication is the same regardless of BMI.

Our overall complication rate was higher (38%) than previous reports by others.^19,23‐26 Thus, Lawrence et al.,¹⁹ in their retrospective study of 20 facilities, demonstrated an overall complication rate of 17%, even though they also included postoperative cardiac complications. Although their study period overlapped our own (1982‐1993), they additionally included patients aged 60 to 65 years, a population known to have fewer comorbidities and fewer postoperative complications than the elderly hip‐fracture patients studied here. In addition, their population may have been healthier at baseline, in that a higher proportion lived at home (73%) and a lower percentage were ASA class III‐V (71%) than our cohort. These differences in baseline characteristics may explain the higher complication rates observed in our study.

Our findings did not suggest any relationship of BMI with noncardiac postoperative medical complications in any of the 4 methods we used to stratify BMI (continuous, categorical, normal vs. abnormal, and 25 kg/m²). Evidence is contradictory as to what the effect of BMI has on postoperative complications. An elevated BMI (30 kg/m²) has been shown to lead to increased sternal wound infection and saphenous vein harvest infection in a cardiac surgery population,²⁷ but other studies^10,28,29 have demonstrated the opposite effect. Among 6336 patients undergoing elective general surgery procedures, the incidence of complications were similar by body mass.³⁰ A matched study design that included urgent and emergent surgeries also did not find any appreciable increased perioperative risk in noncardiac surgery.²⁸ Whether this may be due to the elective nature of the surgeries in these studies, hence leading to selection bias, is unknown.

In geriatric patients, multiple baseline comorbid conditions often are reflected in a higher ASA class, which increases the risk of significant perioperative complications. Our multivariate modeling showed that a high ASA class strongly predicts morbidity and mortality following hip fracture repair, in line with other studies.^19,31,32 Although the Charlson comorbidity index could alternatively been used, we elected to adjust for ASA class as it is more commonly used and is simple to use. Interestingly, surgical year did not significantly predict any complication, which can suggest that practice changes play a minimal impact on patient outcomes. However, we caution that because the individual event rates, particularly vascular, were low, we were unable to fully determine whether changes in practice management, such as improved thromboprophylaxis, would impact event rates over time. Finally, other predictors such as older age³³ and a concomitant history of either COPD or asthma,³⁴ are well‐accepted predictors of inpatient complications. Our attempt to examine specific predictors of complications in each BMI category revealed differing results, making interpretations difficult. Because of power considerations, this was meant solely as an exploratory analysis, and larger cohorts are needed to further ascertain whether predictors are different in these groups. Such a study may in fact identify perioperative issues that allow practitioners caring for this population to modify these factors.

One of the major limitations in our study was our inability to adjust for individual complications using multivariable models, such as deep vein thrombosis or delirium, within each BMI stratum, because of statistical power issues. Such a study would require large numbers of individual complications or events to allow for appropriate adjustments. The authors acknowledge that such individual complication rates may vary dramatically. We were aware of this potential problem, and therefore a priori ascertained a composite outcome of any noncardiac medical complication. However, our results do provide preliminary information regarding the impact of BMI on noncardiac medical complications. Further studies would be needed, though, to fully determine the effect of BMI on the number of cases with each complication.

Obesity (or BMI) is a known cardiovascular risk factor, and our previous study's aim was to determine cardiovascular events in a comparable manner to the way risk indices, such as the Goldman, Lee, or the AHA preoperative algorithm function. The surgical literature often presents noncardiac complications separately, allowing us to directly compare our own data to other published studies. We used 2 separate approaches, focusing on the inpatient stay (ascertaining noncardiac complications) and 1‐year cardiac outcomes (cardiac complications), as these are mediated by different mechanisms and factors. Furthermore, the intent of both studies was to dispel any concerns that an elevated BMI would in fact lead to an increased number of complications. Whether cardiac complications, though, would impact noncardiac complications, or vice‐versa, is unknown, and would require further investigation.

Although we relied on well‐established definitions for body mass, they have often been challenged, as they may underestimate adiposity in the elderly population due to age‐related reductions in lean mass.^35,36 Studies have demonstrated a poor correlation between percent body fat and BMI in the >65 year age group,³⁷ which could impact our results and outcomes by misclassifying patients. Yet, as an anthropometric measurement, BMI is easily obtainable and its variables are routinely documented in patients' medical records, as compared to other anthropometric measurements. Other means of estimating adiposity, such as densitometry or computed tomography (CT) scanning, are impractical, expensive, and not used clinically but routinely in research settings. The lack of standardization in obtaining height and weight, despite nurse‐initiated protocols for bed calibration, may have introduced a degree of measurement bias. Furthermore, the extent of lean mass lost and volume status changes lead to further challenges of using BMI in hospital settings. Whether other anthropometric measurements, including hip circumference, waist circumference, or waist‐hip ratio, should be used in this group of patients requires further examination. However, despite its shortcomings in elderly patients, BMI is still deemed an appropriate surrogate for obesity.

Our main strength was the use of the Rochester Epidemiology Project medical record linkage system to ascertain all patient data. This focuses on patients from a single geographically‐defined community minimizing referral biases often observed in studies originating from a tertiary care referral center. Previous disease‐related epidemiology studies using the Olmsted County population have demonstrated excellent external validity to the U.S. white population.¹⁶ We relied on the medical documentation of the treating clinician for many diagnoses in our data abstraction. Although we attempted to use standardized definitions, clinicians may have inadvertently forgotten to document subjective signs or symptoms that would assist in the categorization of these complications. Hence, added inpatient complications may have been overlooked, suggesting that our results may slightly underestimate the true incidence in this population. Additionally, certain complications may overlap categories, such as pneumonia and infections. We agree with Lawrence et al.¹⁹ that long periods of time are necessary to accumulate data of this kind in an effort to describe complication rates epidemiologically.

Despite no difference in outcomes among BMI categories, our results have striking implications for the hospitalized patient. Thus, underweight elderly patients, often considered frail with minimal functional reserve, are at no higher risk for developing inpatient medical complications than patients with higher BMIs. This is contrary to our study focusing on cardiac complications, where underweight patients were at higher risk.¹⁵ Conversely, obese patients, who have been demonstrated to be at higher risk of medical complications (particularly pulmonary), had no greater risk than patients with normal BMI. To the practicing geriatrician and hospitalist, this information provides important prognostication regarding additional perioperative measures that need to be implemented in these different groups. Based on our results, BMI does not play a particular role in noncardiac medical complications, dispelling any myths of the added burden of excess weight on surgical outcomes in this population. From a hospital perspective, this may be important since additional testing or preventative management in these patients may lead to additional resource use. However, in‐hospital deaths were higher in underweight patients than in patients with a normal BMI. Although we were underpowered to detect any differences in mortality between groups and could therefore not adjust for additional variables, it is unknown whether cardiac or noncardiac complications may be a stronger predictor of death in the underweight patient population. Further studies would be needed to better ascertain this relationship.

Conclusions

In elderly patients undergoing urgent hip fracture repair, BMI does not appear to lead to an excess rate of inpatient noncardiac complications. Our results are the first to demonstrate that BMI has no impact on morbidity in this patient population. Further research on the influence of body composition on inpatient complications in this population is needed to accurately allow for appropriate perioperative prophylaxis. Whether BMI impacts specific complications or in‐patient mortality in this population still requires investigation.

Hospital practices are increasingly responsible for ensuring enhanced patient safety, satisfaction, and cost containment. Recently developed models of care have achieved the necessary efficiency to attain these measures, not only in the use of hospitalists managing general medical1, 2 and postoperative orthopedic patients,3, 4 but also in the use of midlevel providers in busy primary care settings.5 In addition, stroke units6 and geriatric evaluation and management units7, 8 worldwide have demonstrated reduced disability and improved survival and importantly have been proven to provide cost‐effective care. Specialized orthopedic surgery (SOS) units may be a means to reproduce the results observed in these other models.

The economic potential of SOS units will become more significant with changing demographics. The percentage of patients greater than 65 years old will increase, from 12.3% in 2002 to 20% by 2030, with a parallel increase in the prevalence of osteoarthritis (OA).9 The World Health Organization has declared 2000‐2010 the Bone and Joint Decade,10, 11 reflecting that OA affects some 43 million people, with more than 60 million projected to be affected by 2020.12, 13 The National Center for Health Statistics reported that more than 280,000 total knee arthroplasties (TKAs) are performed annually in the United States, which marks an increase in frequency in the last decade that is likely to continue.1419

Approximately 75% of all TKAs are reimbursed under Medicare,17 whereas elective TKA continues to be one of the most common surgeries in the Medicare‐age patient population,20 foreshadowing the prominent cost burden of osteoarthritis in the aging population. The concomitant decreasing reimbursement for arthroplasty in general supports an examination of what constitutes efficient, high‐quality, and cost‐effective care21 for TKA. At our institution, patients undergoing TKA are preferentially triaged to an SOS nursing unit for postoperative care. As hospital bed capacity continues to decline, patients may be triaged to open beds at locations that may not be the optimal choice for nursing care. The primary purpose of this study was to determine the impact of SOS units versus nonorthopedic nursing (NON) units on resource utilization for and outcomes of patients undergoing elective knee arthroplasty. We hypothesized that length of stay would be shorter and cost of inpatient care would be lower for patients cared for on SOS units.

MATERIALS AND METHODS

RESULTS

Baseline patient characteristics are represented in Table 2. Five thousand and eighty‐two patients were admitted to an SOS unit, and 452 patients were admitted to a NON unit. The annual number of patients undergoing TKA increased during our study period, as did the number of patients cared for on NON units. There were no differences between groups in the number of local county patients or in the number of patients primarily referred by other providers for elective arthroplasty. Mean length of stay was 4.9 days in both groups. After adjusting for the specified covariates, including age, sex, year of surgery, Charlson comorbidities, ASA class, and type of anesthesia, LOS was 0.234 days shorter in the SOS group (95% confidence interval [CI]: 0.08, 0.39; P = .002). Overall and hospital costs were significantly lower in the SOS group, as outlined with the other costs in Table 3. Room‐and‐board costs were 5.3% lower for SOS patients than for patients on NON units, representing a per‐patient difference of $244 $87 (95% CI: $72, $415; P = .005).

There were 83 patients (1.63%) transferred from SOS units to the ICU, compared with 14 patients (3.1%) transferred from NON units (P = .02), but no differences in the mean number of ICU days or associated costs between groups. A priori, the authors were aware of the small number of postoperative medical events in this population. In examining the combined endpoint of reoperations, readmissions, and mortality, there were no differences observed in our regression analysis between SOS patients and NON unit patients (0.03 events, standard error: 0.1859; odds ratio: 0.976). Table 4 demonstrates a higher percentage of patients discharged with home health on the NON units than on the SOS units (8.41% vs. 4.62%; P < .001).

DISCUSSION

To the best of our knowledge, this is the first study to examine the impact of specialized orthopedic surgery units on resource utilization in elective knee arthroplasty patients. Our findings demonstrate that patients admitted following elective TKA to SOS units will have a reduced length of stay, lower overall and hospital costs, and fewer unexpected transfers to higher levels of care (ICUs). We believe that these findings are a result in part of the specialized expertise allied health care providers develop by taking care of and focusing on a large volume of patients over time with the same group and type of surgeons. This multidisciplinary setting in which care providers are familiar not only with each other but with this specific population of patients creates the environment necessary for adherence to specialized clinical pathways.27

Patient LOS is an important determinant of resource utilization. In a study by Husted et al., the mean length of stay in Danish hospitals following TKA was 8.6 days in 2003.28 An epidemiological study using the Nationwide Inpatient Sample database of patients in the United States showed that from 1998 to 2000, the mean LOS was 4.3 days.18 In our study, the mean LOS was slightly higher (4.9 days), potentially reflecting referral bias. Achieving additional savings and improved outcomes by further reducing LOS in an environment in which care pathways are already in place is often difficult; hence, alternative approaches and strategies are often necessary.29, 30 Our results suggest that in TKA patients, after adjusting for other factors, there is a decrease in the length of stay of 0.234 days among those cared for on SOS units. However, we cannot state that the existence of the clinical pathway alone is responsible for our data differences because certain components of the care pathway for elective TKA patients are used throughout the hospital regardless of type of postoperative nursing unit. We believe that the interdisciplinary specialty care provided to orthopedic patients on SOS units is a critical component of a successfully implemented care pathway and not just a convenience or practice preference. The same surgeons admitting patients to the same nursing unit, with the same nurses, physical therapists and pharmacists providing care to the same type of patient population over time, leverages the collective experience of all care providers. This integrated, multidisciplinary teamwork may optimize timeliness, achieve incremental cost savings, and improve safety (including a decreased number of unanticipated transfers to an ICU setting).

Clinical pathways are known to reduce overall costs, normally by reducing LOS,29, 3133 and our results suggest approximately an incremental 6% cost reduction with the use of improving patient logistics by using SOS units. An economic evaluation study by Healy et al. suggests that focusing on nursing units may be a means of reducing total costs.29 Our cost savings were slightly lower than the reported savings by other practice assessments; however, we excluded operative and anesthesia costs, both significant contributors to overall and hospital costs. By eliminating these variables, our costs were specifically limited to the postoperative course, which is highly dependent on specialized interdisciplinary care.29

Providing specialized care has a significant impact on society. Although there is a per‐patient savings of only $600 when elective TKA patients are cared for on SOS units, this could be the difference between a positive and negative margin in the setting of fixed reimbursement. With a current average of 90 patients annually triaged postoperatively to NON units, there is a potential loss of $54,000 annually at our institution in just this single patient population with the current mechanisms of perioperative hospital flow. Multiply this potential savings to a national level, and the total is significant. With an aging population, the number of arthroplasties and concomitantly the number of hospitalizations in general are likely to increase, suggesting that changes in hospital flow are required to ensure optimal, cost‐effective care in the best setting available for patients. Such care is often related to surgical volume, and our institution observes such volume. Our results indicate that SOS units are one possible means of achieving this objective of fiscal sustainability, but further studies are needed to determine the indirect and hidden costs of sustaining such units in order to observe the actual cost savings.34 It could be argued that for elective TKA patients to have the most optimal outcomes and most efficient care, the surgical procedure should be performed only if beds are available on the nursing units whose staff has the most specific training.

CONCLUSIONS

In conclusion, postoperative patients after elective knee arthroplasty cared for on specialized orthopedic surgery units have shorter length of stays and cost hospitals less than patients admitted to nonspecialized orthopedic nursing units. In an era in which quality indicators and external reviews are forcing practitioners and health care organizations to become increasingly responsible for their own practices, more research is required to better address specific questions pertaining to different processes of care. Our study is meant to increase the attention paid to patient flow and postoperative logistics in the elective TKA population. SOS units, as a unique model of care, may become an additional step toward ensuring quality care and improved resource utilization.

Hospital practices are increasingly responsible for ensuring enhanced patient safety, satisfaction, and cost containment. Recently developed models of care have achieved the necessary efficiency to attain these measures, not only in the use of hospitalists managing general medical1, 2 and postoperative orthopedic patients,3, 4 but also in the use of midlevel providers in busy primary care settings.5 In addition, stroke units6 and geriatric evaluation and management units7, 8 worldwide have demonstrated reduced disability and improved survival and importantly have been proven to provide cost‐effective care. Specialized orthopedic surgery (SOS) units may be a means to reproduce the results observed in these other models.

The economic potential of SOS units will become more significant with changing demographics. The percentage of patients greater than 65 years old will increase, from 12.3% in 2002 to 20% by 2030, with a parallel increase in the prevalence of osteoarthritis (OA).9 The World Health Organization has declared 2000‐2010 the Bone and Joint Decade,10, 11 reflecting that OA affects some 43 million people, with more than 60 million projected to be affected by 2020.12, 13 The National Center for Health Statistics reported that more than 280,000 total knee arthroplasties (TKAs) are performed annually in the United States, which marks an increase in frequency in the last decade that is likely to continue.1419

Approximately 75% of all TKAs are reimbursed under Medicare,17 whereas elective TKA continues to be one of the most common surgeries in the Medicare‐age patient population,20 foreshadowing the prominent cost burden of osteoarthritis in the aging population. The concomitant decreasing reimbursement for arthroplasty in general supports an examination of what constitutes efficient, high‐quality, and cost‐effective care21 for TKA. At our institution, patients undergoing TKA are preferentially triaged to an SOS nursing unit for postoperative care. As hospital bed capacity continues to decline, patients may be triaged to open beds at locations that may not be the optimal choice for nursing care. The primary purpose of this study was to determine the impact of SOS units versus nonorthopedic nursing (NON) units on resource utilization for and outcomes of patients undergoing elective knee arthroplasty. We hypothesized that length of stay would be shorter and cost of inpatient care would be lower for patients cared for on SOS units.

MATERIALS AND METHODS

RESULTS

Baseline patient characteristics are represented in Table 2. Five thousand and eighty‐two patients were admitted to an SOS unit, and 452 patients were admitted to a NON unit. The annual number of patients undergoing TKA increased during our study period, as did the number of patients cared for on NON units. There were no differences between groups in the number of local county patients or in the number of patients primarily referred by other providers for elective arthroplasty. Mean length of stay was 4.9 days in both groups. After adjusting for the specified covariates, including age, sex, year of surgery, Charlson comorbidities, ASA class, and type of anesthesia, LOS was 0.234 days shorter in the SOS group (95% confidence interval [CI]: 0.08, 0.39; P = .002). Overall and hospital costs were significantly lower in the SOS group, as outlined with the other costs in Table 3. Room‐and‐board costs were 5.3% lower for SOS patients than for patients on NON units, representing a per‐patient difference of $244 $87 (95% CI: $72, $415; P = .005).

There were 83 patients (1.63%) transferred from SOS units to the ICU, compared with 14 patients (3.1%) transferred from NON units (P = .02), but no differences in the mean number of ICU days or associated costs between groups. A priori, the authors were aware of the small number of postoperative medical events in this population. In examining the combined endpoint of reoperations, readmissions, and mortality, there were no differences observed in our regression analysis between SOS patients and NON unit patients (0.03 events, standard error: 0.1859; odds ratio: 0.976). Table 4 demonstrates a higher percentage of patients discharged with home health on the NON units than on the SOS units (8.41% vs. 4.62%; P < .001).

DISCUSSION

To the best of our knowledge, this is the first study to examine the impact of specialized orthopedic surgery units on resource utilization in elective knee arthroplasty patients. Our findings demonstrate that patients admitted following elective TKA to SOS units will have a reduced length of stay, lower overall and hospital costs, and fewer unexpected transfers to higher levels of care (ICUs). We believe that these findings are a result in part of the specialized expertise allied health care providers develop by taking care of and focusing on a large volume of patients over time with the same group and type of surgeons. This multidisciplinary setting in which care providers are familiar not only with each other but with this specific population of patients creates the environment necessary for adherence to specialized clinical pathways.27

Patient LOS is an important determinant of resource utilization. In a study by Husted et al., the mean length of stay in Danish hospitals following TKA was 8.6 days in 2003.28 An epidemiological study using the Nationwide Inpatient Sample database of patients in the United States showed that from 1998 to 2000, the mean LOS was 4.3 days.18 In our study, the mean LOS was slightly higher (4.9 days), potentially reflecting referral bias. Achieving additional savings and improved outcomes by further reducing LOS in an environment in which care pathways are already in place is often difficult; hence, alternative approaches and strategies are often necessary.29, 30 Our results suggest that in TKA patients, after adjusting for other factors, there is a decrease in the length of stay of 0.234 days among those cared for on SOS units. However, we cannot state that the existence of the clinical pathway alone is responsible for our data differences because certain components of the care pathway for elective TKA patients are used throughout the hospital regardless of type of postoperative nursing unit. We believe that the interdisciplinary specialty care provided to orthopedic patients on SOS units is a critical component of a successfully implemented care pathway and not just a convenience or practice preference. The same surgeons admitting patients to the same nursing unit, with the same nurses, physical therapists and pharmacists providing care to the same type of patient population over time, leverages the collective experience of all care providers. This integrated, multidisciplinary teamwork may optimize timeliness, achieve incremental cost savings, and improve safety (including a decreased number of unanticipated transfers to an ICU setting).

Clinical pathways are known to reduce overall costs, normally by reducing LOS,29, 3133 and our results suggest approximately an incremental 6% cost reduction with the use of improving patient logistics by using SOS units. An economic evaluation study by Healy et al. suggests that focusing on nursing units may be a means of reducing total costs.29 Our cost savings were slightly lower than the reported savings by other practice assessments; however, we excluded operative and anesthesia costs, both significant contributors to overall and hospital costs. By eliminating these variables, our costs were specifically limited to the postoperative course, which is highly dependent on specialized interdisciplinary care.29

Providing specialized care has a significant impact on society. Although there is a per‐patient savings of only $600 when elective TKA patients are cared for on SOS units, this could be the difference between a positive and negative margin in the setting of fixed reimbursement. With a current average of 90 patients annually triaged postoperatively to NON units, there is a potential loss of $54,000 annually at our institution in just this single patient population with the current mechanisms of perioperative hospital flow. Multiply this potential savings to a national level, and the total is significant. With an aging population, the number of arthroplasties and concomitantly the number of hospitalizations in general are likely to increase, suggesting that changes in hospital flow are required to ensure optimal, cost‐effective care in the best setting available for patients. Such care is often related to surgical volume, and our institution observes such volume. Our results indicate that SOS units are one possible means of achieving this objective of fiscal sustainability, but further studies are needed to determine the indirect and hidden costs of sustaining such units in order to observe the actual cost savings.34 It could be argued that for elective TKA patients to have the most optimal outcomes and most efficient care, the surgical procedure should be performed only if beds are available on the nursing units whose staff has the most specific training.

CONCLUSIONS

In conclusion, postoperative patients after elective knee arthroplasty cared for on specialized orthopedic surgery units have shorter length of stays and cost hospitals less than patients admitted to nonspecialized orthopedic nursing units. In an era in which quality indicators and external reviews are forcing practitioners and health care organizations to become increasingly responsible for their own practices, more research is required to better address specific questions pertaining to different processes of care. Our study is meant to increase the attention paid to patient flow and postoperative logistics in the elective TKA population. SOS units, as a unique model of care, may become an additional step toward ensuring quality care and improved resource utilization.

Hospital practices are increasingly responsible for ensuring enhanced patient safety, satisfaction, and cost containment. Recently developed models of care have achieved the necessary efficiency to attain these measures, not only in the use of hospitalists managing general medical1, 2 and postoperative orthopedic patients,3, 4 but also in the use of midlevel providers in busy primary care settings.5 In addition, stroke units6 and geriatric evaluation and management units7, 8 worldwide have demonstrated reduced disability and improved survival and importantly have been proven to provide cost‐effective care. Specialized orthopedic surgery (SOS) units may be a means to reproduce the results observed in these other models.

The economic potential of SOS units will become more significant with changing demographics. The percentage of patients greater than 65 years old will increase, from 12.3% in 2002 to 20% by 2030, with a parallel increase in the prevalence of osteoarthritis (OA).9 The World Health Organization has declared 2000‐2010 the Bone and Joint Decade,10, 11 reflecting that OA affects some 43 million people, with more than 60 million projected to be affected by 2020.12, 13 The National Center for Health Statistics reported that more than 280,000 total knee arthroplasties (TKAs) are performed annually in the United States, which marks an increase in frequency in the last decade that is likely to continue.1419

Approximately 75% of all TKAs are reimbursed under Medicare,17 whereas elective TKA continues to be one of the most common surgeries in the Medicare‐age patient population,20 foreshadowing the prominent cost burden of osteoarthritis in the aging population. The concomitant decreasing reimbursement for arthroplasty in general supports an examination of what constitutes efficient, high‐quality, and cost‐effective care21 for TKA. At our institution, patients undergoing TKA are preferentially triaged to an SOS nursing unit for postoperative care. As hospital bed capacity continues to decline, patients may be triaged to open beds at locations that may not be the optimal choice for nursing care. The primary purpose of this study was to determine the impact of SOS units versus nonorthopedic nursing (NON) units on resource utilization for and outcomes of patients undergoing elective knee arthroplasty. We hypothesized that length of stay would be shorter and cost of inpatient care would be lower for patients cared for on SOS units.

MATERIALS AND METHODS

RESULTS

Baseline patient characteristics are represented in Table 2. Five thousand and eighty‐two patients were admitted to an SOS unit, and 452 patients were admitted to a NON unit. The annual number of patients undergoing TKA increased during our study period, as did the number of patients cared for on NON units. There were no differences between groups in the number of local county patients or in the number of patients primarily referred by other providers for elective arthroplasty. Mean length of stay was 4.9 days in both groups. After adjusting for the specified covariates, including age, sex, year of surgery, Charlson comorbidities, ASA class, and type of anesthesia, LOS was 0.234 days shorter in the SOS group (95% confidence interval [CI]: 0.08, 0.39; P = .002). Overall and hospital costs were significantly lower in the SOS group, as outlined with the other costs in Table 3. Room‐and‐board costs were 5.3% lower for SOS patients than for patients on NON units, representing a per‐patient difference of $244 $87 (95% CI: $72, $415; P = .005).

There were 83 patients (1.63%) transferred from SOS units to the ICU, compared with 14 patients (3.1%) transferred from NON units (P = .02), but no differences in the mean number of ICU days or associated costs between groups. A priori, the authors were aware of the small number of postoperative medical events in this population. In examining the combined endpoint of reoperations, readmissions, and mortality, there were no differences observed in our regression analysis between SOS patients and NON unit patients (0.03 events, standard error: 0.1859; odds ratio: 0.976). Table 4 demonstrates a higher percentage of patients discharged with home health on the NON units than on the SOS units (8.41% vs. 4.62%; P < .001).

DISCUSSION

To the best of our knowledge, this is the first study to examine the impact of specialized orthopedic surgery units on resource utilization in elective knee arthroplasty patients. Our findings demonstrate that patients admitted following elective TKA to SOS units will have a reduced length of stay, lower overall and hospital costs, and fewer unexpected transfers to higher levels of care (ICUs). We believe that these findings are a result in part of the specialized expertise allied health care providers develop by taking care of and focusing on a large volume of patients over time with the same group and type of surgeons. This multidisciplinary setting in which care providers are familiar not only with each other but with this specific population of patients creates the environment necessary for adherence to specialized clinical pathways.27

Patient LOS is an important determinant of resource utilization. In a study by Husted et al., the mean length of stay in Danish hospitals following TKA was 8.6 days in 2003.28 An epidemiological study using the Nationwide Inpatient Sample database of patients in the United States showed that from 1998 to 2000, the mean LOS was 4.3 days.18 In our study, the mean LOS was slightly higher (4.9 days), potentially reflecting referral bias. Achieving additional savings and improved outcomes by further reducing LOS in an environment in which care pathways are already in place is often difficult; hence, alternative approaches and strategies are often necessary.29, 30 Our results suggest that in TKA patients, after adjusting for other factors, there is a decrease in the length of stay of 0.234 days among those cared for on SOS units. However, we cannot state that the existence of the clinical pathway alone is responsible for our data differences because certain components of the care pathway for elective TKA patients are used throughout the hospital regardless of type of postoperative nursing unit. We believe that the interdisciplinary specialty care provided to orthopedic patients on SOS units is a critical component of a successfully implemented care pathway and not just a convenience or practice preference. The same surgeons admitting patients to the same nursing unit, with the same nurses, physical therapists and pharmacists providing care to the same type of patient population over time, leverages the collective experience of all care providers. This integrated, multidisciplinary teamwork may optimize timeliness, achieve incremental cost savings, and improve safety (including a decreased number of unanticipated transfers to an ICU setting).

Clinical pathways are known to reduce overall costs, normally by reducing LOS,29, 3133 and our results suggest approximately an incremental 6% cost reduction with the use of improving patient logistics by using SOS units. An economic evaluation study by Healy et al. suggests that focusing on nursing units may be a means of reducing total costs.29 Our cost savings were slightly lower than the reported savings by other practice assessments; however, we excluded operative and anesthesia costs, both significant contributors to overall and hospital costs. By eliminating these variables, our costs were specifically limited to the postoperative course, which is highly dependent on specialized interdisciplinary care.29

Providing specialized care has a significant impact on society. Although there is a per‐patient savings of only $600 when elective TKA patients are cared for on SOS units, this could be the difference between a positive and negative margin in the setting of fixed reimbursement. With a current average of 90 patients annually triaged postoperatively to NON units, there is a potential loss of $54,000 annually at our institution in just this single patient population with the current mechanisms of perioperative hospital flow. Multiply this potential savings to a national level, and the total is significant. With an aging population, the number of arthroplasties and concomitantly the number of hospitalizations in general are likely to increase, suggesting that changes in hospital flow are required to ensure optimal, cost‐effective care in the best setting available for patients. Such care is often related to surgical volume, and our institution observes such volume. Our results indicate that SOS units are one possible means of achieving this objective of fiscal sustainability, but further studies are needed to determine the indirect and hidden costs of sustaining such units in order to observe the actual cost savings.34 It could be argued that for elective TKA patients to have the most optimal outcomes and most efficient care, the surgical procedure should be performed only if beds are available on the nursing units whose staff has the most specific training.

CONCLUSIONS

In conclusion, postoperative patients after elective knee arthroplasty cared for on specialized orthopedic surgery units have shorter length of stays and cost hospitals less than patients admitted to nonspecialized orthopedic nursing units. In an era in which quality indicators and external reviews are forcing practitioners and health care organizations to become increasingly responsible for their own practices, more research is required to better address specific questions pertaining to different processes of care. Our study is meant to increase the attention paid to patient flow and postoperative logistics in the elective TKA population. SOS units, as a unique model of care, may become an additional step toward ensuring quality care and improved resource utilization.