Selection of Subjects
During April and early May 1999, all consecutive patients aged 18 years and older with clinic medical appointments were invited to participate in our study. One of 4 medical office assistants explained the study to each patient. The participants could allow or not allow study findings to be shared with their care providers after the clinic visit. Trained readers were available for those patients unable to read the survey. This service was used on 3 occasions. Participants received a $5 coupon to a local grocery store. Of the 589 patients approached, 68 refused and 21 were missed, for an enrollment of 500 patients (85%), representing 19% of the patients seen at the clinic in 1999. Of those refusing participation, 18 were not interested, 17 were too sick, 23 were too busy, and 10 cited other reasons. There were no significant differences in age or sex between the study participants and those who refused. The mean age for those who refused was 40 years (standard deviation [SD]=11.2) versus 38 years (SD=12.1) for study participants, (t(588)=1.88). Of those who refused 59% were women; 68% of the study participants were women (c2=2.42; df=1).
Data Collection
Patients. The patients completed a questionnaire before being seen by their health care providers. For patients who agreed, providers were alerted when survey results indicated the presence of 1 or more mental disorders.
Study Instrument. The study instruments included the recently validated Primary Care Evaluation of Mental Disorders (PRIME-MD) Patient Health Questionnaire (PHQ)30; the 20-item Medical Outcomes Study Short Form (SF-20), a validated31 tool to assess functional status; and other questions described below. The PRIME-MD PHQ is a self-report version of the original PRIME-MD32 that allows researchers to assess the presence of 7 psychiatric disorders. Like the original PRIME-MD, the PHQ assessed threshold disorders (major depression, panic disorder, other anxiety disorder, bulimia) and subthreshold disorders (minor depression, binge-eating disorder, probable alcohol abuse, somatoform disorder). Because providers were blind to patient response, diagnosis of somatoform disorder was not included. Some patients were classified as symptom screen positive because they indicated distress but failed to meet criteria for a subthreshold or a threshold diagnosis. Patients who screened positive reported depressed mood or low interest on more than half the days, a panic attack in the previous 4 weeks, feeling nervous more than half the days, often feeling “you cannot control what or how much you eat,” or being “bothered a lot” by more than 6 of the 13 PHQ physical symptoms. The SF-20 measures functional status in 6 dimensions. We defined the term “disability days” as the number of days in the past 3 months patients were kept from usual activities because of not feeling well. Health care use was defined by the number of separate times during the past 3 months that patients went to a medical physician in an office, clinic, or emergency room because of not feeling well, not counting the present visit.
Other questions added to the questionnaire included a 20-item list of current physical illnesses (medical comorbidity)24 and demographic questions. A question to assess patient preferences for service design asked: “In the future, if you desire mental health care, please check your top 2 preferred designs:” (A) your medical provider and mental health provider work in the same setting and communicate with one another about your care; (B) communication between providers with service at separate settings; (C) providers do not communicate with service at the same setting; and (D) providers do not communicate, and service is provided at separate settings (alternatives B, C, and D are abbreviated).
Statistical Analyses
We examined descriptive data for the sample. To test the hypothesis that poor, uninsured primary care patients will have higher rates of mental illness compared with a general primary care sample, we compared prevalence data for psychiatric disorders in the study sample against a representative primary care PHQ sample of 3000 patients. Since the PHQ study is from a different population with different sociodemographic and medical characteristics and different sampling techniques, we could not directly compare its data with statistical tests to discern differences in the populations. To determine if there were more problems with functional status and disability days and higher health care use for patients with more psychiatric symptoms, we classified patients into 3 groups based on psychiatric diagnoses: symptom screen negative, screen positive/subthreshold, and threshold groups. Originally the screen-positive and subthreshold groups were analyzed separately. However, because of nearly identical means they were combined for subsequent analyses. Analyses of covariance were used to examine differences in functioning and disability days. The covariates we used were personal income and number of physical health problems. We performed a chi-square analysis to determine if the psychiatric patient groups differed in the proportion with 3 or more physician visits in the past 3 months.