Methods

Top Methods Results Discussion Author & Article Info References

Study Samples

Physicians

Governors of the American College of Physicians from Arizona, Michigan, Oklahoma, Tennessee, Georgia, Maine, and Connecticut helped recruit a total of 74 members of the College from their respective states. These states were selected because they represented a wide geographic area. Only physicians who spent at least 60% of their time practicing general internal medicine were eligible to participate. Full-time academicians, full-time specialists, full-time researchers, and trainees were excluded. The seven College governors each provided the names of 10 to 15 physicians from their own states who met these eligibility criteria and were thought to be exemplary clinicians. Research personnel sent these physicians (n = 81) a letter explaining that the College governor had recommended them for participation in this project. Telephone follow-up was used to describe the project in detail. Of the 81 physicians recruited, 74 (91%) agreed to participate; these 74 physicians make up our physician sample.

Patients

After participating physicians notified the patients they saw during a specified 2-week period that the patients might be contacted to participate in a voluntary, confidential study of quality of care, the physicians gave study personnel copies of their appointment logs for this period. Research staff randomly selected 11 patients listed in each office's log to participate in the study (n = 814) according to the following procedure, which was devised by the firm that administered the surveys. Research staff divided the total number of patients listed on the log by 11 (the target sample size for each practice) to obtain a number, A. They rounded A up to the next highest integer and went to the Ath number on the random-number table. That random number was divided by 10 to obtain a number, B. The number B was rounded to the next highest integer, and the Bth patient on the log became the first patient in the sample. The second patient was selected by adding A and B to get a number, C. The Cth number on the random-number table was then divided by 10 and rounded to the next highest integer to obtain a number, D. The Dth patient listed on the physician's log became the second patient in the sample. The third patient was selected by adding A and D and repeating the same process. The physicians and their staffs had no role in selecting patients from the logs and were not aware of subsequent patient participation.

Questionnaire

Using the literature on patient perceptions of care and separate focus groups with patients and physicians, we identified a comprehensive list of elements of medical office practice thought to be associated with quality of care in that setting. The 125 elements thought to be important in the delivery of high-quality office-based care covered the following nine domains, which were defined a priori: physician clinical skill, physician interpersonal skill, office support staff, office environment, provision of health-related information, patient involvement in care, nonfinancial access, finances, and coordination of care. The domains are defined in Table 3. In this report, we refer to the 125 elements of care as elements, attributes, or items.

We then constructed a questionnaire that asked both physician and patient respondents to rate, from their own perspectives, the importance of each of the 125 elements to the quality of care delivered in a physician's office (Table 4). Respondents rated each item using a 4-point scale: 1 signified that the respondent considered the item not important; 2 indicated that the item was of medium importance; 3 meant that the item was of high importance; and 4 indicated that the item was essential. Patients and physicians were asked to rate the importance of each item to office-based care in general, not to the performance of the patient's individual physician or to the physician's own practice.

Data Collection and Analysis

An independent survey firm mailed the questionnaire to participants within 1 month of the index visit. Data collection remained open for 5 weeks, and reminders were sent 1 week after the initial mailing.

We used group consensus to cluster items into the domains on the basis of qualitative assessment. When agreement about which domain an item belonged to was not unanimous, we discussed the item as a group until we reached a consensus. During a separate phase of our research that examined patients' reports on the performance of their physicians, we also did a factor analysis that loosely guided the assignment of items to domains. The factor analysis identified two factors. The first was related to the specific skills and style of the physician and thus encompassed the domains of physician clinical skill, physician interpersonal skill, provision of health-related information, and patient involvement in care. The second factor was less directly related to physician skill or style and was thought to encompass the domains of office support staff, office environment, nonfinancial access, coordination of care, and finances. The assignment of items to the nine domains was guided by whether the item belonged to a domain related to the first factor or to the second factor.

We examined the median importance ratings and interquartile ranges of physicians' and patients' ratings for each element of care and then analyzed items aggregated into the nine domains. For each respondent, we determined a domain-specific importance rating by calculating the average rating that the respondent assigned to items in each of the nine domains that the questionnaire addressed. For example, if a respondent assigned importance ratings of 2, 3, 3, 3, 4, and 4 to the six items in the domain of office environment, that person's rating of the importance of the office environment domain would be 3.17 ([2 + 3 + 3 + 3 + 4 + 4]/6). We then determined the median importance ratings assigned to each domain by the patients and by the physicians and used the Mann-Whitney U test to compare the two groups. We also ranked the perceived importance of the domains for patients and physicians by ordering the domains from those with the highest importance score to those with the lowest. We then compared the rank orders of physicians' and patients' ratings of the nine domains using the Spearman rank-order correlation coefficient. All reported P values are two sided.

Results

Top Methods Results Discussion Author & Article Info References

Physician and Patient Characteristics

All 74 physicians who supplied appointment logs were in solo or small-group practice. The mean age was 44 years, 14.9% were women, and 8% belonged to an ethnic minority group. Nine physicians were from Arizona, 9 were from Michigan, 15 were from Oklahoma, 11 were from Tennessee, 12 were from Georgia, 9 were from Maine, and 9 were from Connecticut. Of the 74 physicians, 69 (93%) completed the study questionnaire. Of these 69 physicians, the mean age was 45.5 years, 16 (22%) were women, 5 (7%) belonged to an ethnic minority group, and all were in solo or small-group practice.

The patient response rate was 60% (485 of 815 patients). The characteristics of these patients are shown in Table 1.

Patients' and Physicians' Ratings of the Importance of Individual Elements

Physicians and patients differed substantially in an element-by-element comparison of ratings. Disagreement about the level of importance was statistically significant for 58% of the elements (72 of 125). Table 2 shows data for the 15 elements on which patients' and physicians' ratings greatly differed. Of note, 9 of the 15 largest differences concerned the provision of information to the patient.

Patients' and Physicians' Ratings of the Importance of the Domains

Patients and physicians agreed strongly about the rank order of the nine domains (Spearman r, 0.80). For example, both groups ranked clinical skills highest and office environment lowest. Provision of information was the only domain for which the rank order substantially differed between groups. Physicians ranked office support staff, physician interpersonal skill, patient involvement, and coordination of care higher than patients did, but these differences were not statistically significant.

Although agreement on rank order (that is, relative importance) was generally strong, several discrepancies were found between the importance ratings (that is, absolute importance) of patients and of physicians. The area of greatest absolute disagreement between ratings was the domain of provision of information (Figure 1). Patients ranked this domain second most important, whereas physicians ranked it sixth most important (median ratings, 3.56 for patients and 2.85 for physicians; P < 0.001). The ratings of the two groups also differed significantly in the domains of physician clinical skill (3.75 for patients and 3.35 for physicians; P < 0.001), nonfinancial access (3.00 for patients and 2.87 for physicians; P < 0.001), and finances (3.00 for patients and 2.80 for physicians; P = 0.006).

View larger version (15K): [in this window] [in a new window]   Figure 1. Ratings of importance by patients and physicians. The vertical bars represent the median importance ratings, the left ends of the bars represent the 25th percentile, and the right ends of the bars represent the 75th percentile. The P values for the comparison of the median ratings in each domain were obtained by using the Mann-Whitney U test. Solid lines = patients; dashed lines = physicians.

Discussion

Top Methods Results Discussion Author & Article Info References

Our data suggest that patients and physicians agree that the clinical skill of the physician is the most crucial element of quality health care, but they also suggest that patients and physicians disagree about the relative importance of several other aspects of care. The most striking finding was that patients placed substantially greater value on effective communication of health-related information than did physicians.

The literature shows that patients value information highly. Robbins and colleagues [6] found that patient satisfaction correlated strongly with the amount of information patients received from their physicians. Patients were most satisfied when medical visits involved health education and discussion about specific therapy. Two separate studies [7, 8] that surveyed patients seen in the emergency department found that patient satisfaction was directly related to the amount of information received. A study of persons infected with human immunodeficiency virus [9] also found that the information patients received from providers strongly affected patients' overall satisfaction. Von Korff and associates [10] showed that providing education to patients with back pain leads to increased satisfaction and more favorable clinical outcomes. However, previous studies have not compared patients' and physicians' appraisals of the importance of effective exchange of information. Our data indicate that physicians may underappreciate that information exchange is extremely important to patients.

Our study had several limitations. First, although the patient response rate (60%) was above average for mailed surveys [11], most respondents were white and in the middle income bracket; their median age was 63 years. We selected patients from appointment logs that did not include any demographic or clinical data; thus, we could not provide information on nonrespondents. Second, all participating physicians were internists; although our results may be generalizable to other generalist providers, they may not apply to all physicians. Physicians were not recruited at random but were selected because they were thought to be exemplary clinicians. Our questionnaire should be applied in other contexts to determine the generalizability of our findings. Patients' views are undoubtedly colored by their experiences with their own physicians, and it would be interesting to compare the importance ratings of physicians with the ratings of their own patients. Our sample size, however, prevented us from doing such analyses. Finally, we may have inadvertently neglected areas that patients or physicians consider important to high-quality care. Inviting open-ended comments from participants might have enriched this investigation.

Despite these limitations, our study provides quantitative measures of the value that patients and physicians place on attributes that comprise high-quality office-based medical care. Our study is unique because the participants were patients and physicians from the same practices and because the findings represented "real-world" practices rather than academic practices. These results have implications for the measurement of patient satisfaction, quality improvement, medical education, and health care reform.

Patient satisfaction is multidimensional, and many existing instruments yield overall and domain-specific ratings without adjusting for the salience of the individual components of questionnaires [4, 12-14]. Our findings suggest that it would make sense if particular domains, such as information exchange, weigh more heavily than others in overall satisfaction ratings. Quality improvement efforts increasingly incorporate patients' perspectives, and providers who know what areas patients value can work to meet expectations or counsel patients so that expectations become more realistic [15-19].

Areas that patients consider highly important, such as receiving information and explanations from physicians, currently receive little emphasis in medical training. Medical educators could use information on aspects of care that patients value to focus on fostering skills that will help trainees to better meet patient expectations.

As primary care moves toward center stage in efforts to improve health care in the United States, issues of quality such as those documented in our report will continue to surface and increase in importance. Limited resources preclude the health care system from supplying patients and providers with everything they desire, but better communication of health-related information might be achieved at little extra cost. Further data, such as those described in our study, should be of value, not only in helping physicians meet patient expectations but also in helping policymakers set priorities for health care.

Dr. Davidoff: Annals of Internal Medicine, American College of Physicians, Sixth Street at Race, Philadelphia, PA 19106.

Dr. Lewis: Division of Health Promotion and Disease Prevention, University of California, Los Angeles, School of Medicine and Public Health, Center for Health Sciences, Room 61-236, Box 951772, Los Angeles, CA 90095.

Dr. Nelson: Dartmouth-Hitchcock Medical Center, 1 Medical Center Drive, Lebanon, NH 03756-0001.

Ms. Nelson: Scientific Policy, American College of Physicians, Sixth Street at Race, Philadelphia, PA 19106.

Dr. Kessler: Institute for Social Research, Survey Research Center, University of Michigan, 426 Thompson Street, PO Box 1248, Ann Arbor, MI 48106-1248.

Dr. Delbanco: Division of General Internal Medicine, Beth Israel Hospital, 330 Brookline Avenue, Boston, MA 19106.