Individuals who had experience as standardized patients or other acting experience played these roles. One actor played Boyd Kelly, and 2 played Louise Williams. Each patient had a scripted presentation and medical and social history. We coached the actors regarding case histories, affect, and behaviors. Coaching focused on maintaining a natural dialogue with the physician in which questions were answered directly but diagnostic information was not offered unless explicitly elicited by the physician. If the physicians mentioned depression, the actors were instructed to be surprised and hesitate in accepting the diagnosis. If the actors felt the physician had developed good rapport and adequately explained the diagnosis, they were to reluctantly agree with the treatment plan. If the physicians had not done these things, the actors were to resist the diagnosis and treatment plan except for returning for a follow-up visit. To ensure that the actors played their roles reliably, we videotaped them being interviewed by 3 physicians as part of their training. In addition, we used a hidden microphone to audiotape their visits with the study physicians. We reviewed approximately 10% of the tapes to ensure that the actors maintained their roles and accurately completed their checklists.
Checklist and scale measures. The standardized patients completed a checklist and rating scales immediately after the visit. The checklist included specific questions about the physician’s interviewing behavior (eg, whether the physician asked about psychosocial stressors and the 9 DSM-IV criteria for major depression, discussed the possibility of depression, and educated the patient about depression and its treatment). The actors also recorded whether the physician prescribed antidepressants, the amount of time the physician spent with them, and when they were scheduled for a follow-up visit. For a global assessment of the physicians’ communication skills, the standardized patients completed a 9-item version of the American Board of Internal Medicine’s Patient Satisfaction scale,62 a single-item about rapport, and an adapted version of Kaplan’s 3-item Participatory Decision-Making scale63 (Table 2). The patient satisfaction and rapport items were rated on a 5-point scale (1 = poor; 5 = excellent). Items in the Participatory Decision-Making scale were rated on a separate 5-point scale (1 = strongly disagree; 5 = strongly agree). The scale score was determined by adding the responses to the items. If responses were missing, the scale was not scored. Cronbach’s a was used to assess the reliability of the scale measures. To assess overall knowledge and communication skills, the standardized patients were asked if they thought the physician had attended the Depression Education Program.
Statistical Analyses
We used the chi-square statistic and Student’s t test to assess preintervention and postintervention differences between intervention and control groups. We developed estimates of the prevalence of depression in their practices by dividing the estimated number of adult patients who have major depression by the estimated number of adult patients seen during a typical week.
If we found significant differences in the postintervention scores, we controlled for potential confounding variables (physician sex, specialty, and suspicion or detection of the standardized patient) using logistic and multiple linear regression for categorical and interval variables, respectively. Because of the small sample size, we did separate analyses for each confounding variable that included an interaction term between the confounding and intervention variable. An interaction term was considered significant if the P value was less than .2. If the interaction was not significant, the interaction term was excluded from the final analysis. If an interaction term was significant, the results were analyzed by level of the confounding variable. All tests were 2-tailed using an a of 0.05.
Results
Forty-nine (88%) of the 56 randomly assigned physicians completed the study. There were no statistically significant demographic differences between intervention and control groups (Table 3). The physicians were predominantly men, family physicians, and in single-speciality or multispecialty groups. Capitation and other forms of managed care accounted for approximately 60% of their practices’ revenue. On average, the physicians felt responsible for recognizing and treating depression.
Study physicians did well on the 54-item postintervention knowledge test. Intervention and control groups did not differ significantly in their mean scores (41.5 ± 5.1 and 39.3 ± 4.8, respectively, t = 1.52, P = .136). In addition, most of the physicians agreed or strongly agreed with the statement “I am very knowledgeable in the use of antidepressants” (64% intervention and 44% control, c2 = 1.84, P = .175).
Intervention physicians performed better than control physicians on many of the behaviors assessed by the 2 unannounced standardized patients (Table 4). For Louise Williams, after controlling for the physicians’ sex, specialty, and suspicion that she was a standardized patient, intervention physicians were more likely to assess 5 or more criteria of major depression, suicidal ideation, and stress at home and to discuss the possibility of depression (Table 4). The intervention group also had significantly higher ratings on patient satisfaction, establishing rapport, and participatory decision making. Intervention physicians and control physicians did not differ in the amount of time they spent with Louise Williams (mean minutes ± standard deviation [SD]: 32.2 ± 10.5 and 28.6 ± 14.2, respectively, t = 0.99, P = .33) or in the percentage who suspected she was a standardized patient (56% of intervention and 42% of control physicians suspected or detected, c2= 0.987, P = .32).