Thomas J. Zuber, MD

In 1992, the American Medical Association (AMA) adopted modifications to the Current Procedural Terminology (CPT) codes used for reporting physician evaluation and management (E & M) services.^1-5 Code levels were determined in the majority of visits by the degree of history taking, the physical examination performed, and the complexity of the medical decision making required to provide patient care.^4,6-8 These 3 contributory factors in code determination became known as the key components. The Health Care Financing Administration (HCFA) provided guidelines for the required documentation to support the level of visit coded, and modified the guidelines further in 1995 and 1997.⁹ The 1997 guidelines were not well received by many medical groups,^3,9-12 and the adoption of those guidelines was postponed.⁹ The HCFA released a new proposed set of documentation guidelines in late 1998,^12,13 but at the time of this study those guidelines had not been field tested.

Increasing concern about inaccurate coding and fraudulent billing has led to audits of physician billing practices.^14,15 The Office of the Inspector General of the Department of Health and Human Resources has stated that approximately 40% of physician payments questioned at audit involved insufficient chart documentation.¹⁵ Huge dollar losses to the Medicare program have been extrapolated from small federally sponsored audits.¹⁵

Studies of billing and coding practices have demonstrated poor agreement between the code level and medical record documentation.^1,16,17 A recent study by Chao and colleagues¹ demonstrated only a 55% concordance rate between the billing codes assigned in the office and those assigned by direct observation.

Our study was undertaken to evaluate the level of agreement among physician and nonphysician auditors with office code selections by practicing family physicians. Chart evaluations were performed by physician reviewers with extensive practice experience (faculty), younger and less-experienced physicians (residents), and trained chart auditors (professional coders). We performed direct comparison of the 1995 and 1998 documentation guidelines.

Methods

Source of Data

Five private family physician offices, including 10 medical providers (9 physicians, 1 family nurse practitioner) volunteered to participate. No documentation training was given to the medical providers before initiating the study.

Consecutive billing sheets or computer sheets describing the level of established patient evaluation and management service billed between July 1998 and January 1999 were selected and placed on top of the corresponding chart. The medical providers were blinded to which charts would be audited.

Supporting documentation for this audit was limited to information recorded in the progress note on the date of service, unless the note directed the reviewer to another section of the chart. For example, medications recorded on a separate page were not included as documentation unless a specific reference was made to that page.

Documentation

The 1995 documentation guidelines were selected because of the poor physician acceptance of the 1997 guidelines. The checklist for the 1995 guidelines was created using a modification of the worksheet of the E/M Documentation Auditors’ Instructions.¹⁸ The checklist for the proposed 1998 guidelines was created from a document released by the AMA in November 1998.¹⁹

On the basis of the documentation, the auditor selected the level for each of the 3 key components for both sets of guidelines. Code selection was based on published coding guidelines from the 1999 CPT.²⁰ Because the proposed 1998 guidelines eliminated the straightforward level of medical decision making, only 3 levels of medical decision making were used to determine the 1998 code selections.

Chart Evaluators

Four faculty physicians, 6 resident physicians, and 6 professional coders volunteered to participate as chart auditors in our study. Auditors attended 2 training sessions on the application of the HCFA chart documentation guidelines conducted by T.Z. An additional pilot audit was performed by the reviewers at one site to gain experience with the work sheet. Auditors were permitted to use copies of the 1995 auditor’s instructions and the AMA summary of the 1998 guidelines to assist them during the audits. Once the work sheets were completed, they were not seen by the other auditors.

Results

Evaluation of the Medical Provider Code Selection

Audits were performed on 1069 charts. Agreement with the medical provider code selection was better for all 3 auditors (faculty, residents, and coders) with the 1998 guidelines. The coders differed significantly (P <.001) from the faculty and resident physician auditors in their greater agreement with the code selected by the medical provider. There was no statistical difference between the faculty and resident physician auditors (P=.604 for 1995; P=.799 for 1998) in their level of agreement with the medical provider’s code selection.

The [Figure] shows the numbers and percentage of each established patient code selection found by the auditors to have sufficient documentation to meet or exceed that level. In 69.5% of established patient visits reviewed, the code 99213 was billed. For this code, faculty noted that the documentation was sufficient for 92.7% (1995) and 91.0% (1998) of the visits, resident physicians for 81.3% (1995) and 83.6% (1998), and professional coders for 82.9% (1995) and 95.6% (1998).

All auditors agreed with the medical provider code selection for only 15.2% (1995) and 29.2% (1998) of the charts reviewed. Concordance among auditors for code determination was 30.4% (1995) and 42.5% (1998). When auditors disagreed with the code selected by the medical provider, the documentation supported a code higher than the one billed 4 times as often as insufficient documentation.

Evaluation of the 1995 and 1998 Documentation Guidelines

The visits billed as 99213 by the medical provider yielded the largest amount of data in comparing the documentation systems. Few differences were noted between the 1995 and 1998 documentation systems for the levels of history. The change in the counting of examination items in the 1998 guidelines produced lower levels of examinations. For all auditors, there was a 34-fold drop (68.6% vs 2%) in detailed examinations from the 1995 guidelines to the 1998 guidelines. Higher levels of medical decision making were achieved using the 1998 guidelines. Moderate complexity medical decision making was noted 6 times more often using the 1998 guidelines.

Discussion

Code Selection

The selection of office evaluation and management codes appears to be subjective, with significant degrees of difference noted between physicians and professional coders using the same guidelines. There was no statistical difference in code selection between resident physicians (with only 2.3 years of coding experience) and attending physicians (with 23.3 years of coding experience). The variance we noted between physicians and professional coders is a significant issue in the specialty of family practice, where the majority of patient care involves the reporting of E & M codes. Future systems used to evaluate chart documentation must demonstrate greater interobserver agreement.

Inadequate Chart Documentation

Auditors often observed that documentation of work performed during the visit appeared in the chart outside the progress note (such as on medication lists or on vital signs sheets), but this information was not referenced according to the audit rules. Physicians might significantly improve documentation by routinely incorporating some reference in the progress note to alternate information in the chart. The future use of electronic medical records may eliminate this problem by allowing chart reviewers to access more of the work actually being performed by medical providers. Alternatively, the HCFA could modify its rules to include this information automatically.

Underreporting of Work Performed

We noted that the documentation often supported a higher code than was originally billed. This tendency toward undercoding may reflect physician fear of audits or regulatory agencies. Physicians also may be uncomfortable or lacking in knowledge of cumbersome coding rules and documentation guidelines, and guessing or routine code assignments may play a role in code selection. The high degree of underreporting in our study makes it unlikely that intentional efforts exist to bill for work not performed.

This is one of the largest physician-directed documentation audits undertaken. It is possible that physician auditors view documentation differently from nonphysician auditors. Our finding that the professional coders agreed with the medical provider’s code selection more often than the faculty or resident physicians for both the 1995 and 1998 documentation guidelines was unexpected. The assumption that the coders would assign lower code levels was not confirmed by our study.

1995 and 1998 Chart Documentation Systems

The 1998 proposed HCFA documentation guidelines appear to provide better agreement with current family physician billing patterns than the 1995 guidelines. The levels of history achieved were very similar between the 1995 and 1998 guidelines. Higher levels of examination were much more difficult to achieve with the 1998 guidelines because of changes in the counting of examination items.

The determination of medical decision making changed in the 1998 guidelines to encompass just 3 types (low complexity, moderate complexity, and high complexity). In addition, the highest level achieved for any of the 3 components of medical decision making was enough to define the level, while 2 of 3 levels of the components were required with the 1995 guidelines to achieve a level of medical decision making. These changes made it easier to reach higher levels of medical decision making using the 1998 guidelines. The shift upward in levels of medical decision making using the 1998 guidelines was offset by the downward shift in the levels of examination.

Conclusions

Using the 1995 and 1998 guidelines, documentation supported the level of code selected in the majority of cases. When auditors disagreed with the code selected by the medical care provider, the documentation supported selection of a higher code than originally billed 4 times as often as it supported a lower code.

National coding systems used to determine levels of physician service should produce substantial interobserver agreement among reviewers and practicing physicians. If this system is used to determine physician fraud and assess penalties, consistency is tantamount. The 1995 and 1998 documentation guidelines fail to produce adequate concurrence among chart auditors for family practice patient office visits.

Acknowledgments

We would like to acknowledge those individuals who significantly assisted with the performance and publication of our study: Shirley Cresswell, Debbie Sanders, Darlene McComb, Sue Dalek, Larry Lalonde, Jan Goldberger, Jacqueline Schultz, Lynn Goldberger, Cyndi Coates, Robert Vitu, Sharon Sawyers, Peter Vasilenko, James Lafleur, John Clements, Kathi Kumar, Sue Davis, Jorge Plasencia, Michael Butman, Mitch Freeman, and John Cavendish.

In 1992, the American Medical Association (AMA) adopted modifications to the Current Procedural Terminology (CPT) codes used for reporting physician evaluation and management (E & M) services.^1-5 Code levels were determined in the majority of visits by the degree of history taking, the physical examination performed, and the complexity of the medical decision making required to provide patient care.^4,6-8 These 3 contributory factors in code determination became known as the key components. The Health Care Financing Administration (HCFA) provided guidelines for the required documentation to support the level of visit coded, and modified the guidelines further in 1995 and 1997.⁹ The 1997 guidelines were not well received by many medical groups,^3,9-12 and the adoption of those guidelines was postponed.⁹ The HCFA released a new proposed set of documentation guidelines in late 1998,^12,13 but at the time of this study those guidelines had not been field tested.

Increasing concern about inaccurate coding and fraudulent billing has led to audits of physician billing practices.^14,15 The Office of the Inspector General of the Department of Health and Human Resources has stated that approximately 40% of physician payments questioned at audit involved insufficient chart documentation.¹⁵ Huge dollar losses to the Medicare program have been extrapolated from small federally sponsored audits.¹⁵

Studies of billing and coding practices have demonstrated poor agreement between the code level and medical record documentation.^1,16,17 A recent study by Chao and colleagues¹ demonstrated only a 55% concordance rate between the billing codes assigned in the office and those assigned by direct observation.

Our study was undertaken to evaluate the level of agreement among physician and nonphysician auditors with office code selections by practicing family physicians. Chart evaluations were performed by physician reviewers with extensive practice experience (faculty), younger and less-experienced physicians (residents), and trained chart auditors (professional coders). We performed direct comparison of the 1995 and 1998 documentation guidelines.

Methods

Source of Data

Five private family physician offices, including 10 medical providers (9 physicians, 1 family nurse practitioner) volunteered to participate. No documentation training was given to the medical providers before initiating the study.

Consecutive billing sheets or computer sheets describing the level of established patient evaluation and management service billed between July 1998 and January 1999 were selected and placed on top of the corresponding chart. The medical providers were blinded to which charts would be audited.

Supporting documentation for this audit was limited to information recorded in the progress note on the date of service, unless the note directed the reviewer to another section of the chart. For example, medications recorded on a separate page were not included as documentation unless a specific reference was made to that page.

Documentation

The 1995 documentation guidelines were selected because of the poor physician acceptance of the 1997 guidelines. The checklist for the 1995 guidelines was created using a modification of the worksheet of the E/M Documentation Auditors’ Instructions.¹⁸ The checklist for the proposed 1998 guidelines was created from a document released by the AMA in November 1998.¹⁹

On the basis of the documentation, the auditor selected the level for each of the 3 key components for both sets of guidelines. Code selection was based on published coding guidelines from the 1999 CPT.²⁰ Because the proposed 1998 guidelines eliminated the straightforward level of medical decision making, only 3 levels of medical decision making were used to determine the 1998 code selections.

Chart Evaluators

Four faculty physicians, 6 resident physicians, and 6 professional coders volunteered to participate as chart auditors in our study. Auditors attended 2 training sessions on the application of the HCFA chart documentation guidelines conducted by T.Z. An additional pilot audit was performed by the reviewers at one site to gain experience with the work sheet. Auditors were permitted to use copies of the 1995 auditor’s instructions and the AMA summary of the 1998 guidelines to assist them during the audits. Once the work sheets were completed, they were not seen by the other auditors.

Results

Evaluation of the Medical Provider Code Selection

Audits were performed on 1069 charts. Agreement with the medical provider code selection was better for all 3 auditors (faculty, residents, and coders) with the 1998 guidelines. The coders differed significantly (P <.001) from the faculty and resident physician auditors in their greater agreement with the code selected by the medical provider. There was no statistical difference between the faculty and resident physician auditors (P=.604 for 1995; P=.799 for 1998) in their level of agreement with the medical provider’s code selection.

The [Figure] shows the numbers and percentage of each established patient code selection found by the auditors to have sufficient documentation to meet or exceed that level. In 69.5% of established patient visits reviewed, the code 99213 was billed. For this code, faculty noted that the documentation was sufficient for 92.7% (1995) and 91.0% (1998) of the visits, resident physicians for 81.3% (1995) and 83.6% (1998), and professional coders for 82.9% (1995) and 95.6% (1998).

All auditors agreed with the medical provider code selection for only 15.2% (1995) and 29.2% (1998) of the charts reviewed. Concordance among auditors for code determination was 30.4% (1995) and 42.5% (1998). When auditors disagreed with the code selected by the medical provider, the documentation supported a code higher than the one billed 4 times as often as insufficient documentation.

Evaluation of the 1995 and 1998 Documentation Guidelines

The visits billed as 99213 by the medical provider yielded the largest amount of data in comparing the documentation systems. Few differences were noted between the 1995 and 1998 documentation systems for the levels of history. The change in the counting of examination items in the 1998 guidelines produced lower levels of examinations. For all auditors, there was a 34-fold drop (68.6% vs 2%) in detailed examinations from the 1995 guidelines to the 1998 guidelines. Higher levels of medical decision making were achieved using the 1998 guidelines. Moderate complexity medical decision making was noted 6 times more often using the 1998 guidelines.

Discussion

Code Selection

The selection of office evaluation and management codes appears to be subjective, with significant degrees of difference noted between physicians and professional coders using the same guidelines. There was no statistical difference in code selection between resident physicians (with only 2.3 years of coding experience) and attending physicians (with 23.3 years of coding experience). The variance we noted between physicians and professional coders is a significant issue in the specialty of family practice, where the majority of patient care involves the reporting of E & M codes. Future systems used to evaluate chart documentation must demonstrate greater interobserver agreement.

Inadequate Chart Documentation

Auditors often observed that documentation of work performed during the visit appeared in the chart outside the progress note (such as on medication lists or on vital signs sheets), but this information was not referenced according to the audit rules. Physicians might significantly improve documentation by routinely incorporating some reference in the progress note to alternate information in the chart. The future use of electronic medical records may eliminate this problem by allowing chart reviewers to access more of the work actually being performed by medical providers. Alternatively, the HCFA could modify its rules to include this information automatically.

Underreporting of Work Performed

We noted that the documentation often supported a higher code than was originally billed. This tendency toward undercoding may reflect physician fear of audits or regulatory agencies. Physicians also may be uncomfortable or lacking in knowledge of cumbersome coding rules and documentation guidelines, and guessing or routine code assignments may play a role in code selection. The high degree of underreporting in our study makes it unlikely that intentional efforts exist to bill for work not performed.

This is one of the largest physician-directed documentation audits undertaken. It is possible that physician auditors view documentation differently from nonphysician auditors. Our finding that the professional coders agreed with the medical provider’s code selection more often than the faculty or resident physicians for both the 1995 and 1998 documentation guidelines was unexpected. The assumption that the coders would assign lower code levels was not confirmed by our study.

1995 and 1998 Chart Documentation Systems

The 1998 proposed HCFA documentation guidelines appear to provide better agreement with current family physician billing patterns than the 1995 guidelines. The levels of history achieved were very similar between the 1995 and 1998 guidelines. Higher levels of examination were much more difficult to achieve with the 1998 guidelines because of changes in the counting of examination items.

The determination of medical decision making changed in the 1998 guidelines to encompass just 3 types (low complexity, moderate complexity, and high complexity). In addition, the highest level achieved for any of the 3 components of medical decision making was enough to define the level, while 2 of 3 levels of the components were required with the 1995 guidelines to achieve a level of medical decision making. These changes made it easier to reach higher levels of medical decision making using the 1998 guidelines. The shift upward in levels of medical decision making using the 1998 guidelines was offset by the downward shift in the levels of examination.

Conclusions

Using the 1995 and 1998 guidelines, documentation supported the level of code selected in the majority of cases. When auditors disagreed with the code selected by the medical care provider, the documentation supported selection of a higher code than originally billed 4 times as often as it supported a lower code.

National coding systems used to determine levels of physician service should produce substantial interobserver agreement among reviewers and practicing physicians. If this system is used to determine physician fraud and assess penalties, consistency is tantamount. The 1995 and 1998 documentation guidelines fail to produce adequate concurrence among chart auditors for family practice patient office visits.

Acknowledgments

We would like to acknowledge those individuals who significantly assisted with the performance and publication of our study: Shirley Cresswell, Debbie Sanders, Darlene McComb, Sue Dalek, Larry Lalonde, Jan Goldberger, Jacqueline Schultz, Lynn Goldberger, Cyndi Coates, Robert Vitu, Sharon Sawyers, Peter Vasilenko, James Lafleur, John Clements, Kathi Kumar, Sue Davis, Jorge Plasencia, Michael Butman, Mitch Freeman, and John Cavendish.

In 1992, the American Medical Association (AMA) adopted modifications to the Current Procedural Terminology (CPT) codes used for reporting physician evaluation and management (E & M) services.^1-5 Code levels were determined in the majority of visits by the degree of history taking, the physical examination performed, and the complexity of the medical decision making required to provide patient care.^4,6-8 These 3 contributory factors in code determination became known as the key components. The Health Care Financing Administration (HCFA) provided guidelines for the required documentation to support the level of visit coded, and modified the guidelines further in 1995 and 1997.⁹ The 1997 guidelines were not well received by many medical groups,^3,9-12 and the adoption of those guidelines was postponed.⁹ The HCFA released a new proposed set of documentation guidelines in late 1998,^12,13 but at the time of this study those guidelines had not been field tested.

Increasing concern about inaccurate coding and fraudulent billing has led to audits of physician billing practices.^14,15 The Office of the Inspector General of the Department of Health and Human Resources has stated that approximately 40% of physician payments questioned at audit involved insufficient chart documentation.¹⁵ Huge dollar losses to the Medicare program have been extrapolated from small federally sponsored audits.¹⁵

Studies of billing and coding practices have demonstrated poor agreement between the code level and medical record documentation.^1,16,17 A recent study by Chao and colleagues¹ demonstrated only a 55% concordance rate between the billing codes assigned in the office and those assigned by direct observation.

Our study was undertaken to evaluate the level of agreement among physician and nonphysician auditors with office code selections by practicing family physicians. Chart evaluations were performed by physician reviewers with extensive practice experience (faculty), younger and less-experienced physicians (residents), and trained chart auditors (professional coders). We performed direct comparison of the 1995 and 1998 documentation guidelines.

Methods

Source of Data

Five private family physician offices, including 10 medical providers (9 physicians, 1 family nurse practitioner) volunteered to participate. No documentation training was given to the medical providers before initiating the study.

Consecutive billing sheets or computer sheets describing the level of established patient evaluation and management service billed between July 1998 and January 1999 were selected and placed on top of the corresponding chart. The medical providers were blinded to which charts would be audited.

Supporting documentation for this audit was limited to information recorded in the progress note on the date of service, unless the note directed the reviewer to another section of the chart. For example, medications recorded on a separate page were not included as documentation unless a specific reference was made to that page.

Documentation

The 1995 documentation guidelines were selected because of the poor physician acceptance of the 1997 guidelines. The checklist for the 1995 guidelines was created using a modification of the worksheet of the E/M Documentation Auditors’ Instructions.¹⁸ The checklist for the proposed 1998 guidelines was created from a document released by the AMA in November 1998.¹⁹

On the basis of the documentation, the auditor selected the level for each of the 3 key components for both sets of guidelines. Code selection was based on published coding guidelines from the 1999 CPT.²⁰ Because the proposed 1998 guidelines eliminated the straightforward level of medical decision making, only 3 levels of medical decision making were used to determine the 1998 code selections.

Chart Evaluators

Four faculty physicians, 6 resident physicians, and 6 professional coders volunteered to participate as chart auditors in our study. Auditors attended 2 training sessions on the application of the HCFA chart documentation guidelines conducted by T.Z. An additional pilot audit was performed by the reviewers at one site to gain experience with the work sheet. Auditors were permitted to use copies of the 1995 auditor’s instructions and the AMA summary of the 1998 guidelines to assist them during the audits. Once the work sheets were completed, they were not seen by the other auditors.

Results

Evaluation of the Medical Provider Code Selection

Audits were performed on 1069 charts. Agreement with the medical provider code selection was better for all 3 auditors (faculty, residents, and coders) with the 1998 guidelines. The coders differed significantly (P <.001) from the faculty and resident physician auditors in their greater agreement with the code selected by the medical provider. There was no statistical difference between the faculty and resident physician auditors (P=.604 for 1995; P=.799 for 1998) in their level of agreement with the medical provider’s code selection.

The [Figure] shows the numbers and percentage of each established patient code selection found by the auditors to have sufficient documentation to meet or exceed that level. In 69.5% of established patient visits reviewed, the code 99213 was billed. For this code, faculty noted that the documentation was sufficient for 92.7% (1995) and 91.0% (1998) of the visits, resident physicians for 81.3% (1995) and 83.6% (1998), and professional coders for 82.9% (1995) and 95.6% (1998).

All auditors agreed with the medical provider code selection for only 15.2% (1995) and 29.2% (1998) of the charts reviewed. Concordance among auditors for code determination was 30.4% (1995) and 42.5% (1998). When auditors disagreed with the code selected by the medical provider, the documentation supported a code higher than the one billed 4 times as often as insufficient documentation.

Evaluation of the 1995 and 1998 Documentation Guidelines

The visits billed as 99213 by the medical provider yielded the largest amount of data in comparing the documentation systems. Few differences were noted between the 1995 and 1998 documentation systems for the levels of history. The change in the counting of examination items in the 1998 guidelines produced lower levels of examinations. For all auditors, there was a 34-fold drop (68.6% vs 2%) in detailed examinations from the 1995 guidelines to the 1998 guidelines. Higher levels of medical decision making were achieved using the 1998 guidelines. Moderate complexity medical decision making was noted 6 times more often using the 1998 guidelines.

Discussion

Code Selection

The selection of office evaluation and management codes appears to be subjective, with significant degrees of difference noted between physicians and professional coders using the same guidelines. There was no statistical difference in code selection between resident physicians (with only 2.3 years of coding experience) and attending physicians (with 23.3 years of coding experience). The variance we noted between physicians and professional coders is a significant issue in the specialty of family practice, where the majority of patient care involves the reporting of E & M codes. Future systems used to evaluate chart documentation must demonstrate greater interobserver agreement.

Inadequate Chart Documentation

Auditors often observed that documentation of work performed during the visit appeared in the chart outside the progress note (such as on medication lists or on vital signs sheets), but this information was not referenced according to the audit rules. Physicians might significantly improve documentation by routinely incorporating some reference in the progress note to alternate information in the chart. The future use of electronic medical records may eliminate this problem by allowing chart reviewers to access more of the work actually being performed by medical providers. Alternatively, the HCFA could modify its rules to include this information automatically.

Underreporting of Work Performed

We noted that the documentation often supported a higher code than was originally billed. This tendency toward undercoding may reflect physician fear of audits or regulatory agencies. Physicians also may be uncomfortable or lacking in knowledge of cumbersome coding rules and documentation guidelines, and guessing or routine code assignments may play a role in code selection. The high degree of underreporting in our study makes it unlikely that intentional efforts exist to bill for work not performed.

This is one of the largest physician-directed documentation audits undertaken. It is possible that physician auditors view documentation differently from nonphysician auditors. Our finding that the professional coders agreed with the medical provider’s code selection more often than the faculty or resident physicians for both the 1995 and 1998 documentation guidelines was unexpected. The assumption that the coders would assign lower code levels was not confirmed by our study.

1995 and 1998 Chart Documentation Systems

The 1998 proposed HCFA documentation guidelines appear to provide better agreement with current family physician billing patterns than the 1995 guidelines. The levels of history achieved were very similar between the 1995 and 1998 guidelines. Higher levels of examination were much more difficult to achieve with the 1998 guidelines because of changes in the counting of examination items.

The determination of medical decision making changed in the 1998 guidelines to encompass just 3 types (low complexity, moderate complexity, and high complexity). In addition, the highest level achieved for any of the 3 components of medical decision making was enough to define the level, while 2 of 3 levels of the components were required with the 1995 guidelines to achieve a level of medical decision making. These changes made it easier to reach higher levels of medical decision making using the 1998 guidelines. The shift upward in levels of medical decision making using the 1998 guidelines was offset by the downward shift in the levels of examination.

Conclusions

Using the 1995 and 1998 guidelines, documentation supported the level of code selected in the majority of cases. When auditors disagreed with the code selected by the medical care provider, the documentation supported selection of a higher code than originally billed 4 times as often as it supported a lower code.

National coding systems used to determine levels of physician service should produce substantial interobserver agreement among reviewers and practicing physicians. If this system is used to determine physician fraud and assess penalties, consistency is tantamount. The 1995 and 1998 documentation guidelines fail to produce adequate concurrence among chart auditors for family practice patient office visits.

Acknowledgments

We would like to acknowledge those individuals who significantly assisted with the performance and publication of our study: Shirley Cresswell, Debbie Sanders, Darlene McComb, Sue Dalek, Larry Lalonde, Jan Goldberger, Jacqueline Schultz, Lynn Goldberger, Cyndi Coates, Robert Vitu, Sharon Sawyers, Peter Vasilenko, James Lafleur, John Clements, Kathi Kumar, Sue Davis, Jorge Plasencia, Michael Butman, Mitch Freeman, and John Cavendish.