Printer Friendly

Modification of physician behavior by performance feedback.

The health care reform debate brought to the forefront issues of cost control that have been dealt with extensively in the literature for a decade or more.[1-4] A 1983 review of strategies for cost containment for laboratory testing detailed five interventional strategies that might be applied: education, feedback, cost awareness, rationing, and market-oriented financial incentives/risk sharing plans.[1] All the strategies were directed at decreased utilization for diagnostic tests and procedures. The author concluded that education had some limited effect, but its utility for future impact remains uncertain. Cost awareness, using decision theory, generally lacks appeal for the nonacademic. Financial incentives would be negligible because of the fee-for-service nature of medical practice in 1983. Unless concern for cost is given a higher priority, cost will not have an impact on the ordering of diagnostic tests. In 1983, there were disincentives to constrain testing because hospital revenues would decline. In the past decade, managed care has achieved a higher penetration in the marketplace. The usual agreements between hospitals and third-party payers are now on a case rate, a per diem, or an "at risk" basis. This shift in reimbursement may create an opportunity for reexamination of strategies not demonstrated to be effective in the past. The purpose of the literature review reported here is to assess the efficacy and sustainability of strategies for behavior modification.

Physician Demographics

Physician factors, such as sensitivity of demographics and the utility of active versus passive participation, need examination. In a study of commonly used blood tests in the intensive care unit, it was found that resident physicians significantly underestimated laboratory charges in comparison to ICU attending physicians.[5] Attending physicians' estimates were 29.5 percent higher. Eighty-five percent of physicians felt that too many blood tests were being performed, and they also perceived that they paid insufficient attention to the risk of anemia and to cost issues. Physicians felt test ordering in the ICU was excessive, and they were open to change.

Management of two common conditions, cystitis and vaginitis, were examined in a small randomized study in a teaching hospital with a retrospective audit of the physician's records of the condition.6 Physicians were asked to set criteria for management of the condition, an audit was performed, and physicians were allowed to modify the criteria and assign weights to each criterion. At that time, prospective monitoring of physicians' adherence to the criteria was conducted. All physicians enjoyed the task of completing their own criteria lists. For senior physicians, lists tended to be concise, while junior faculty produced complicated protocols. All were surprised at their poor performance in the retrospective audit when measured by their own criteria. Although few changed criteria, performance for both conditions improved at 6 and 14 months without further contact. This demonstrates that physicians are interested in improvement and active participation.

The issue of whether younger and older physicians differ in utilization was examined by Weingarten et al.[7] The patients of older physicians had longer lengths of stay and higher costs. Further detailed analysis revealed that older physicians cared for older and more severely ill patients than younger physicians. There was little correlation between resource utilization an physician age.


Coupling educational intervention with automated feedback might increase opportunities to modify physician behavior and reduce charges.[8] A randomized controlled study with hospital information system feedback of charges for participating physicians was performed. Although considerable resources were required to implement these systems, they were easily managed once in place. A hypothesis was that a newsletter attached to the feedback would enhance the effect. The goal was to reduce total prescribing charges for outpatients and to improve resident knowledge of charges for therapeutics with minimal additional resources. Computer feedback of prescription charges previously resulted in a modest improvement of resident knowledge of charges, along with a reduction in prescription charges. Residents in the intervention group indicated that they read most or all the newsletters. Their performance on post-tests indicated otherwise. Perhaps the newsletter was not read or believed. The authors postulate that there are a number of influences on prescribing behavior and that among the most powerful is pharmaceutical representatives. Counter detailing efforts may be necessary to achieve reductions in prescription charges.

In a study of 1,921 patients 30 years or older with chest pain, Lee et al.[9] measured the effect of passive introduction of triage recommendations in the emergency department and daily after admission to the hospital. Length of stay and total charges revealed no effect of guidance on patients at low risk for myocardial infarction.

Automated peer comparison is a technique to quantify physicians performance relationships to their peers.[1] Gehlbach et al.[4] employed a randomized study of family medicine residents and faculty. Members of the experimental group received monthly reports for nine months identifying drugs they prescribed by brand, with estimates of costs and the savings that could be realized by use of generic drugs. A control group received no feedback. Both groups were monitored for 12 months after all feedback had ceased. Although no change occurred in the control group, the percentage of generic prescriptions for the experimental physicians was 14 percent at baseline, 67 percent during the period of feedback, and 54 percent for the follow-up period. This intervention was effective and sustainable. It also suggested that maintenance of feedback is necessary for sustained improvement.

Parrino targeted physicians in the top 50 percentiles for antibiotics expenditures.[10] They were notified of their relationship to peers and were compared to a control group with no feedback. Over a one-year period, no significant reduction in expenditures was noted. Thirty percent of the attending physicians were responsible for 80 percent of antibiotic costs, and 60 percent of top percentile physicians were members of the medical, versus surgical, cohort. Future efforts will need to be targeted toward outlier physicians.

Cognitive feedback consists of demonstration of the weights participants give predictive information (cues) in the formation of judgments. A computer calculates weights from responses to a series of cases and presents a display that compares learners' weights to optimal weights. Cognitive feedback has been shown to enhance students' and physicians' diagnostic performance. Wigton et al.[11] studied a student health facility in regard to the likelihood of a positive streptococcal pharyngeal culture. The rate of positive cultures during the period of investigation was 5-7 percent. The average of physician estimates was 24 percent, but was rapidly corrected to a rate of 6.5 percent. Achievement rose during the first two months and then returned to the one-month level after six months. Students were studied in addition to the physicians. Like physicians, students rapidly lowered their high inaccurate base rate. The students reached a higher level of congruence that was durable. Unfortunately, pharyngeal exudate, while not a predictive cue in the rule, continued to be used by physicians. The student cue weighing closely matched prescribed weighting, and pharyngeal inflammation fell from 24 to 3 percent as a weighted factor. Physicians also failed to utilize lack of cough as a positive predictor. The students had a better congruence for this parameter. This suggests that very young practitioners (students) may be more adaptable to new information than more senior practitioners.

Poses et al.[12] conducted a similar study of streptococcal pharyngitis. The efficacy of a state-of-the-art lecture regarding diagnosis and management of pharyngitis, including an introduction to the clinical prediction rule for streptococcal pharyngitis, was examined. The authors compared active cognitive feedback to no intervention in the two groups that participated in the didactic presentation. Both groups improved to some extent, but improvement after computerized cognitive feedback warrants further investigation of other more complex arenas.

In 1985, Studnicki and colleagues reported on the impact of a cybemetic system of feedback to physicians on inappropriate hospital use.[2,3] Their articles were based on studies derived from the Western Maryland Review Organization (WMRO). The appropriateness of hospital admissions and each day of stay were determined using an appropriate admission evaluation protocol with 27 criteria. Criteria were in three categories: physician services, nursing services, and patient condition factors. Every Medicare admission and every day of Medicare inpatient stay were reviewed immediately after the patient's discharge by a WMRO reviewer. A 10 percent sample was reviewed a second time.

Every two weeks, physicians who admitted Medicare patients received a comprehensive report of reviewers' findings. The first report listed the medical record of the physician's Medicare cases; the second reported aggregate findings of the physician's cases reviewed during the current two-week period, i.e., the days of inpatient stay, the failed days of those cases, and the aggregate failed rate. Comparable information was displayed for all reviewed Medicare cases of physicians practicing in a physician's hospital and for all six WMRO hospitals. The reasons for the physician's failed days were reported. The mean percentages of total failed days for all physicians at the hospital in question and for all hospitals were compared. The total number of admission failures specific to the physician versus all physicians' Medicare admissions in his or her hospital and for all WMRO hospitals were compared.

In the absence of any intervention, the levels of inappropriate days increased with patient volume, and there was greater variance around the average of low-volume physicians because of small numbers. The response to the cybernetic control program was influenced by volume. Overall, physicians with fewer than 95 patient days during the 12-week study period did not improve their group average performance. The likelihood that the individual physician would improve performance was directly related to his or her volume of patients. Less than 20 percent of the low-volume physicians improved their rate of inappropriate hospital days during the study period, while nearly 68 percent of the high-volume physicians improved. The characteristics of this system are that physicians are included in a continuous and consistent feedback communication system that provides professionally relevant information.


The most pervasive conclusion from this literature review is that there is no well-documented methodology for modification of physician ordering or admission behavior in the direction of cost containment. There are indications that physicians are sensitive to the issues and desire a participatory role. The impact of age or experience on behavior malleability is inconclusive. Perhaps, the differences relate to the extremes of age difference in the more favorable study of students.

Elegant techniques, such as cognitive feedback with modeling of physician behavior and cybernetic systems, may be worth additional efforts. The studies of Studnicki were published in 1985, and no substantive amplification of the work was detected in an extensive literature search. Targeting of specific subsets of physicians may be the most cost-effective intervention. Economic incentives to modify behavior need to be reexamined in the context of the shift from fee-for-service reimbursement to managed care reimbursement. Physicians who have been actively screened in the selection process for learning skills and other explicit characteristics demonstrate other implicit characteristics that preclude further behavior modification to serve the common good.


[1.] Grossman, R. "A Review of Physician Cost-Containment Strategies for Laboratory Testing." Medical Care 21(8): 783-802, Aug. 1983. [2.] Studnicki, J., and others. "Impact of a Cybernetic System of Feedback To Physicians on Inappropriate Hospital Use." Journal of Medical Education 60(6):454-60, June 1985. [3.] Studnicki, J., add Stevens, C. "Cybernetic Appropriateness Review: Does It Change Physician Hospital Utilization Patterns?" Evaluation and Program Planning 8:195-205, 1985. [4.] Gehlbach, S., and others. "Improving Drug Prescribing in a Primary Care Practice." Medical Care 22(3):193-201, March 1984. [5.] Cook, D. "Physicians' Perceptions of Laboratory Costs in the Intensive Care Unit." Clinical Investigation Medicine 15(5):476-81, May 1992. [6.] Norton, P., and Dempsey, L. "Self-Audit: Its Effect on Quality of Care." Journal of Family Practice 21(4):289-91, Oct. 1985. [7.] Weingarten, S., and others. "Do Older Internists Use More Hospital Resources than Younger Internists for Patients Hospitalized with Chest Pain? A Study of Patients Hospitalized in the Coronary Care and Intermediate Care Units." Critical Care Medicine 247:762-7, 1992. [8.] Hershey, C., and others. "The Effect of Computerized Feedback Coupled with a Newsletter upon Outpatient Prescribing Charges." Medical Care 26(l):88-94, Jan. 1988. [9.] Lee, T., and others. "Failure of Information as an Intervention to Modify Clinical Management. A Time-Series Trial on Patients with Acute Chest Pain." Annals of Internal Medicine 122(6):434-7, March 15, 1995. [10.] Parrino, T. "The Nonvalue of Retrospective Peer Comparison Feedback in Containing Hospital Antibiotic Costs." American Journal of Medicine 86(4):442-8, April 1989. [11.] Wigton, R., and others. "Teaching Old Dogs New Tricks: Using Cognitive Feedback to Improve Physicians' Diagnostic Judgments on Simulated Cases." Academic Medicine 65(9 Suppl):s5-6, Sept. 1990. [12.] Poses, R., and others. "Controlled Trial Using Computerized Feedback to Improve Physicians' Diagnostic Judgments." Academic. Medicine 67(5):345-7, May 1992.

Michael P. Corder, MD, is Director of Clinical Resources, Heritage Medical Systems, Bakersfield, Calif., and Clinical Professor of Medicine, UCLA. He may be reached at 4570 California Ave., Bakersfield, Calif. 93309, 805/327-4411. FAX 805/327-2517.
COPYRIGHT 1996 American College of Physician Executives
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1996, Gale Group. All rights reserved. Gale Group is a Thomson Corporation Company.

Article Details
Printer friendly Cite/link Email Feedback
Author:Corder, Michael P.
Publication:Physician Executive
Date:Apr 1, 1996
Previous Article:From the factory floor to the doctor's office.
Next Article:The changing role of physicians in hospital governance.

Related Articles
Advantages of local UR.
Performance appraisal as a modifier of physician behavior.
Defining quality physicians.
Reducing hospital lengths of stay.
Medical staff handling of concerns with physicians' performance.
Evaluation of physician practice using risk-adjusted commercial databases.
How to shape positive relationships in medical practices and hospitals. (Part 2: Conflict Management).
The levers of influence. (Positively Influencing Physicians).
Pre-certification: lessons from diagnostic imaging. (Positively Influencing Physicians).
Why become a Certified Physician Executive? (Career Management).

Terms of use | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters