Hospital Peer Review and the National Practitioner Data Bank: Clinical Privileges Action Reports

Context The National Practitioner Data Bank (NPDB) is believed to be an important source of information for peer review activities by the majority of those who use it. However, concern has been raised that hospitals may be underreporting physicians with performance problems to the NPDB.

Objective To examine variation in clinical privileges action reporting by hospitals to the NPDB, changes in reporting over time, and the association of hospital characteristics with reporting.

Design Retrospective cohort study of privileges action reports to the NPDB between 1991 and 1995, linked with the 1992 and 1995 databases from the Annual Survey of Hospitals conducted by the American Hospital Association.

Setting and Participants A total of 4743 short-term, nonfederal, general medical/surgical hospitals throughout the United States that were continuously open between 1991-1995 and registered with the NPDB.

Main Outcome Measures (1) Reporting of 1 or more privileges actions during the 5-year study period and (2) privileges action reporting rates (numbers of actions reported per 100,000 admissions).

Results Study hospitals reported 3328 privileges actions between 1991 and 1995; 34.2% reported 1 or more actions during the period. The range of privileges action reporting rates for these hospitals was 0.40 to 52.27 per 100,000 admissions, with an overall rate of 2.36 per 100,000 admissions. The proportion of hospitals reporting an action decreased from 11.6% in 1991 to 10.0% in 1995 (P=.008). After adjustment for other factors, urban hospitals had significantly higher reporting than rural hospitals (adjusted odds ratio [OR], 1.21 [95% confidence interval {CI}, 1.02-1.43]), while members of the Council of Teaching Hospitals of the Association of American Medical Colleges had significantly lower reporting than nonmembers (adjusted OR, 0.54 [95% CI, 0.40-0.73]). There were notable regional differences in reporting, with the east south Central region having the lowest rate per 100,000 admissions (1.49 [95% CI, 1.33-1.65]).

Conclusions The results of this study indicate a low and declining level of hospital privileges action reporting to the NPDB. Several potential explanations exist, 1 of which is that the information reported to the NPDB is incomplete.

The National Practitioner Data Bank (NPDB) serves as a central repository of information about health care providers' malpractice payments, adverse licensure actions, professional membership restrictions, and adverse hospital privileging actions.¹ Hospitals are required to query the NPDB for all new staff appointments and at least once every 2 years for all existing medical staff. They are also required to report actions that affect the clinical privileges of their medical staff. Reportable actions include reduction, restriction, suspension, or revocation of clinical privileges for at least 31 days; the voluntary resignation of clinical privileges either while peer review of a potential quality concern is taking place or in lieu of the peer review process; and the denial of clinical privileges to a new or existing medical staff member when peer review judgment is involved.²

Two surveys found that the majority of those entities using the NPDB rated it as an important source of information for peer review activities.^3,4 The utility of the NPDB, however, is dependent on its ability to provide complete and accurate information to its users.

A 1995 Office of Inspector General report raised concern that there may be underreporting by hospitals of physicians with performance problems to the NPDB.⁵ The report found that over 3 years, about 75% of all hospitals, some of them facilities with more than 300 beds, had not reported a single privileges action to the NPDB. The report also found great variation in privileges action reporting from state to state. Finally, the number of privileges action reports submitted by hospitals was less than half the number of reports of licensure actions submitted by state licensing boards during the same 3-year period. If there is underreporting of privileges actions to the NPDB, this would undermine the NPDB's effectiveness as a quality-improvement tool.

In this study, we used 5 years of data from the NPDB linked to information from the American Hospital Association (AHA) to further examine (1) reporting of hospital privileges actions to the NPDB, (2) associations between hospital characteristics and reporting, and (3) changes in reporting over time. We hypothesized that physicians would be reticent to take reportable privileges actions against their peers, resulting in a decrease in reporting over the study period.

Hospitals listed as open by the AHA for 1 or more years between 1991 and 1995 were hand linked to hospitals registered with the NPDB from its inception through 1995. Study hospitals were short-term, nonfederal, and general medical/surgical facilities open throughout the study period.

Hospital characteristics were obtained from the AHA's database of US hospitals. For almost all variables, we used information from the 1995 AHA database to describe the hospitals in this study. The percentage of active and associate staff who were board certified was no longer included in the 1995 survey, so we used data from the only other year of the AHA database we had available (1992). Data on hospital characteristics were linked to information that the NPDB maintains on each privileges action taken by hospitals, including the date of the action and the length of the action.

We identified whether a hospital had submitted a privileges action report in each of the study years, and calculated a rate of privileges action reports per 100,000 admissions for each hospital. The number of admissions in each hospital over 5 years was calculated using 1992 and 1995 AHA data. We assumed relative constancy in number of admissions, and applied 1992 admission figures to the years 1991-1993 and the 1995 admissions to the years 1994-1995. Reports that had been submitted initially and later rescinded were not included.

We used individual hospital-based variables to calculate both the percentage of hospitals reporting any privileges action and the rate of privileges actions per 100,000 admissions over the study period for groups of hospitals. This aggregate privileges action rate sums the number of actions for all hospitals with a particular characteristic (eg, rural hospitals) over the 5 years, divides by the total number of admissions for those same hospitals, and expresses the rate per 100,000 admissions.

We were interested in the effect of a number of hospital characteristics on hospital privileges action reporting: (1) governance (for-profit; nongovernment, not-for-profit; government, nonfederal [from here on referred to as state and local]), (2) accreditation by the Joint Commission on Accreditation of Healthcare Organizations (JCAHO), (3) membership in the Council of Teaching Hospitals of the Association of American Medical Colleges (AAMC), (4) urban/rural location (urban defined as metropolitan statistical area counties, rural defined as all other counties), (5) number of beds (<100, 100-299, ≥300), (6) osteopathic vs nonosteopathic, (7) whether the hospital was contract managed, (8) whether a nursing home was part of the hospital, (9) percentage of board-certified physicians on staff, and (10) region of the country, defined by the standard Census Bureau divisions (Table 1).

We examined variation in privileges action reporting rates for individual hospitals, the change in reporting for all study hospitals over the study period, and differences in reporting between groups of hospitals, stratified by number of beds. χ² Tests were used to assess the statistical significance of differences between dichotomous outcome variables. A jackknife procedure was used to estimate the confidence intervals around the aggregate privileges action rates and to determine the statistical significance of differences in rates between hospital groups.

Multiple logistic regression for the dichotomous outcome variable of whether the hospital had reported any privileges actions and multiple linear regression for the outcome variable of the number of privileges actions (natural log) were used to identify the independent effect of each hospital characteristic on reporting while controlling for other hospital characteristics and factors, including number of admissions.

Of the 6903 hospitals listed as open for 1 or more years between 1991 and 1995 by the AHA, 6754 (97.8%) were linked with hospitals registered with the NPDB. Of these 6754 hospitals, 6009 (89.0%) were open for all 5 study years. A total of 4743 facilities were short-term, nonfederal, general medical/surgical hospitals that were included in our analyses.

The majority of the 4743 study hospitals were nongovernment, not-for-profit, and accredited by JCAHO (Table 1). They were about equally distributed between urban and rural areas. Nearly 30% had nursing homes attached.

The 4743 study hospitals reported 3328 clinical privileges actions over the 5-year study period; about a third of the hospitals (34.2%) reported at least 1 action over the study period. The range of the privileges action rates for individual hospitals that had taken actions was between 0.40 and 52.27 per 100,000 admissions. The overall privileges action rate for the study hospitals in aggregate was 2.36 per 100,000 admissions.

There was a decrease both in the percentage of hospitals reporting at least 1 privileges action and in the rate of privileges actions over the study period (Table 2). These findings were generally consistent across hospitals with differing characteristics. While the proportion of hospitals taking privileges actions decreased over time, the proportion of actions that were permanent or indefinite increased significantly, from 79.6% in 1991 to 85.1% in 1995 (P=.01).

Urban hospitals and hospitals accredited by JCAHO were more likely to have reported 1 or more privileges actions and had higher rates of reported actions per 100,000 admissions than their counterparts for nearly all bed size categories (Table 3). State and local hospitals were least likely of the 3 hospital ownership types to have reported actions and had the lowest rates of reporting an action in almost all bed size categories. The majority of hospitals that were members of the Council of Teaching Hospitals of the AAMC had 300 or more beds. Within this bed size category, hospital members of the Council of Teaching Hospitals of the AAMC had lower rates of reported privileges actions and were less likely to have reported an action than nonmember hospitals.

There were significant regional differences in privileges action reporting, with some of the lowest reporting by hospitals in the east south Central region (ie, Alabama, Kentucky, Mississippi, and Tennessee).

Regression analysis confirmed several of these findings (Table 4). After controlling for other factors, urban hospitals and JCAHO-accredited hospitals were significantly more likely to report actions than their counterparts. Hospitals that were members of the AAMC Council of Teaching Hospitals were significantly less likely to report actions than nonmembers. These findings were essentially the same when we used the total number of privileges actions (natural log) as the dependent variable, although the finding for JCAHO-accredited hospitals was no longer statistically significant. In this regression, hospitals with 300 or more beds reported significantly more privileges actions than hospitals with fewer beds.

This study found evidence of a low and declining level of clinical privileges action reporting by hospitals to the NPDB. The total number of clinical privileges actions reported was small, and decreased over the study period. The variation between individual hospitals in reporting was great. More than 65% of all study hospitals, including more than 250 large hospitals (40.7%) with 300 or more beds, reported no privileges actions during the 5 years of the study. There are a number of potential explanations for this low level of clinical privileges action reporting to the NPDB.

Data from other studies suggest that low level of quality-of-care problems as an explanation for the low level of reporting is unlikely. The Harvard Malpractice Study estimated that 1% of New York's hospitalizations involved adverse events due to negligence (an estimated 27,179 events in New York State during 1984).⁶ This figure was similar to the 0.8% negligence rate found in the California Medical Association's Medical insurance feasibility study.⁷ While many negligent events that occur in hospitals may not prompt clinical privileges action, figures such as these suggest that the average of 47 privileges actions reported by New York hospitals in each of our study years is low.

Underreporting of clinical privileges actions taken is unlikely to fully account for the low level of reporting, as it would suggest that hospitals were failing to comply with legislation requiring clinical privileges action reporting. Failure to comply with this legislation puts hospitals at risk of losing the immunity protection for their peer review processes provided by the Health Care Quality Improvement Act of 1986. On the other hand, there may be room for interpretation of the reporting requirements to the NPDB, leading to variation between hospitals in the types of actions reported. In addition, a related analysis of the association between the strength of state-imposed penalties for not reporting privileges actions and the level of reporting to the NPDB found that the 3 states with strong penalties (>$5000) had significantly higher reporting to the NPDB.⁸ The 1995 Office of Inspector General report also found that at least 1 state licensing board charged with sending privileges reports to the NPDB had made administrative errors leading to underreporting of privileges actions.

Two studies provide evidence that preferential imposition of penalties that did not require reporting is a plausible explanation. A 1994 survey of all short-stay general, rural hospitals in Washington, Alaska, Montana, and Idaho found that 20% had increased the use of peer review decisions that did not require reporting to the NPDB (eg, monitoring professional activities or requiring continuing medical education without restricting privileges, and imposing privileges actions of <31 days) in the prior 2 years.³ A second 1994 survey of 807 hospitals and 76 health maintenance organizations found that in the prior 2 years 9.4% of hospitals and 13.1% of health maintenance organizations reported that practitioners had offered concessions to avoid having a reportable action taken against them.⁴ These studies suggest that the NPDB has had a paradoxical effect on hospital-based peer review activities.

Over the study period, hospital quality improvement programs were shifting emphasis from individual provider review to continuous quality improvement (also known as total quality management). Continuous quality improvement emphasizes quality improvement through examination and revision of systems and processes within a facility rather than targeting individuals and their practices.^9,10 The implementation of continuous quality improvement programs could contribute to a decline in privileges actions either by decreasing the frequency of negligent actions or shifting resources away from peer review activities that result in privileges actions.

Hospitals are not required to report the loss of privileges for physicians who lose them through a licensing action. There were nearly 3 times the number of licensing actions (11,680, not including reinstatements and revisions) than privileges actions over the study period. In addition, licensing actions increased between 1991 and 1995, while privileges actions decreased, raising the possibility that licensure actions were substituting for privileges actions. Despite these suggestive trends, we believe this is an unlikely explanation for the low level of hospital privileges action reporting.

First, the study's finding that the proportion of permanent or indefinite hospital privileges actions increased over the study period suggests that it was physicians with less severe problems whose privileges actions declined over the study period. This group was unlikely to have had licensing actions substituted for their privileges actions. Second, licensing actions are much more severe penalties than loss of hospital privileges. In Washington State, licensing actions take an average of 9 to 12 months from complaint to action (Maryella Jansen, Washington State Medical Quality Assurance Commission, oral communication, February 26, 1999). It is unlikely that hospitals would maintain a physician's privileges for this length of time if he/she had performance problems severe enough to warrant licensing action. If this is the case, it supports the conclusion that hospitals may be underascertaining physicians with performance problems or using penalties that do not require reporting.

The explanation that low level of detection of physicians with performance problems is supported by the data presented above from the Harvard Medical Practice Study⁶ and the higher level of licensing actions that suggest many more quality-of-care problems than are reflected by clinical privileges action reporting. In addition, the literature is replete with articles discussing the many barriers to effective peer review, such as lack of specific qualifications and training of physicians performing peer review, poor agreement between physicians regarding quality of care, personal or professional ties that can bias physician judgments about quality of care, and fear of litigation.^11-16

The evidence from this study cannot be used to definitively identify the causes for the low and declining level of clinical privileges action reporting. Supporting evidence from other sources and the high degree of dissatisfaction with the concept of the NPDB and its operation reported in the early 1990s suggest that underascertainment of physicians with performance problems and the use of penalties that do not require reporting were the most significant contributors to these findings, however.^17-23

This study also found that some hospital characteristics were associated with clinical privileges action reporting levels. Hospitals accredited by JCAHO, which must uphold certain peer review standards to maintain their accreditation, were more likely to report actions than nonaccredited hospitals. This supports the idea that hospitals with some surveillance of their peer review processes may be more likely to take clinical privileges actions against physicians with performance problems. The finding of lower privileges action reporting for member hospitals of the AAMC Council of Teaching Hospitals, on the other hand, could be due to less effective peer review activity or the employment of physicians with fewer performance problems in these teaching hospitals. The latter explanation is supported by 3 recent studies^24-26 suggesting that teaching hospitals may provide a higher quality of care than nonteaching hospitals. Rural hospitals, which had lower action reporting than urban hospitals, may be most likely to experience some of the barriers to effective peer review (eg, difficulty objectively evaluating a peer with whom the reviewer has close personal or professional ties or an economic relationship).^12,27-30 In addition, rural hospitals may allocate fewer resources than urban facilities to peer review. Alternately, rural hospitals may have provided higher-quality care. Previous work in this area has been mixed, with some reporting higher,^31,32 others lower,²⁴ and still others equivalent quality of care.²⁵

This study is limited by our ability to report clinical privileges action rates using only admissions as the denominator, which results in rates that are difficult to interpret. Admissions were used as a surrogate for physician exposure. Is a rate of 5 privileges actions per 100,000 admissions high or low? Is the difference between 5 and 3 privileges actions per 100,000 admissions meaningful? Using physicians as the denominator would have created more easily interpretable results, but was impossible due to the inconsistent nature of the information on active medical staff size reported to the AHA.

Ensuring the highest quality health care system is of paramount importance to the health care profession and the public. Support for rigorous surveillance of the quality of health care providers led Congress to authorize the establishment of the NPDB. This unique resource is one of only a few truly national data sets that can provide information on the quality of care rendered by our country's physicians. Organizations and institutions use it routinely in their day-to-day decisions about credentialing and licensing of individual physicians. This study's finding of a low level of clinical privileges action reporting suggests that the information reported to the NPDB may be incomplete. This is not unexpected, given the barriers to effective peer review that have been reported, including fear of liability and preexisting personal and professional ties between peer reviewers and their colleagues under review. In addition, the decline in privileges action reporting over the study period raises the possibility that the NPDB itself may be serving as a disincentive to effective hospital peer review practices. Given the critical importance of quality of care, and our national interest in fostering the movement of high-quality physicians to places of need, it is axiomatic that some sort of national reporting system captures and makes available data about problematic practitioners. To this end, it is important to develop new strategies for ensuring effective hospital peer review, as well as to find ways to minimize the disincentive that the NPDB may have on peer review activities.