Subjects were recruited from private internal medicine offices or from a family medicine clinic. Stratified sampling techniques were used to oversample patients with depressive symptoms21
; this was done to facilitate other studies focusing on patients with depressive conditions. The recruitment process, described in detail in this section, is shown in the flow chart ().
The 129 subjects from the private offices were included in a previously published study.22
The private offices were those of three board-certified general internists, clinical faculty at the University of Rochester School of Medicine and Dentistry who maintained full-time private practices. Their offices drew on a patient population that came predominantly from middle-class neighborhoods in the city of Rochester, and from the mostly middle-class and upper-middle-class adjacent suburbs. Two of the practices shared office space, and the third had an adjacent office; all three shared on-call responsibilities for the combined practices. Each of the two office groupings was joined by a physician assistant part way through the subject recruitment period.
Patients were screened 2 or 3 1/2 days per week with a self-report depression symptom scale, the Center for Epidemiologic Studies Depression Scale (CES-D),23
to allow oversampling of patients with depressive symptoms and syndromes. All patients aged 60 years and over who appeared for visits in these offices during these times were eligible for screening, with the rare exception of patients unable to complete the questionnaires owing to gross impairments in communication skills or language barriers (n
= 9). With informed consent (procedures approved by the University of Rochester Medical Center’s Research Subjects Review Board), patients completed the CES-D in the office. Screening was performed at first by office staff (March–May 1994), during which time 83 (79%) of 105 patients eligible for screening completed the CES-D. During the majority of the enrollment period (May 1994 – June 1995), screening was performed by research personnel, and 424 (87%) of 485 eligible patients completed the CES-D. The lower screening completion rate by office staff as compared with research staff was due to potentially eligible patients being missed, rather than a higher patient refusal rate.
Screening was conducted in similar fashion by research personnel two half-days per week from July 1995 to June 1996 at the University of Rochester Family Medicine Center at Highland Hospital. This free-standing clinic, located near Highland Hospital, is the home of the University’s Family Medicine Department and residency program, and serves a predominantly urban and poorer population than the private offices. Clinicians include faculty attending physicians in family medicine, family medicine residents, and family medicine nurse practitioners. During the enrollment period, 436 patients aged 60 years or older visited the Family Medicine Center. Twenty-nine of these were ineligible for screening, either because of gross communication impairments (14 were mentally retarded—the Family Medicine Center serves several group homes for mentally retarded residents; 1 had advanced Alzheimer’s disease) or because of a language barrier without available family or other translator (n = 14). Of the remaining 407 eligible subjects, 3 were missed, 53 refused screening, and a total of 351 (86%) completed the CES-D.
Selected screened patients were approached by telephone for informed consent to complete an in-depth interview including the Structured Clinical Interview for DSM-III-R (SCID),24
within 4 weeks of their primary care visit. To enrich the study group with persons with significant depressive symptoms and syndromes, the sample was stratified on the CES-D: all patients scoring>21 were approached for SCID interview, and a random sample of those scoring ≤21 were approached, up to a maximum of three SCID interviews per week. The SCID interview was held either in the patient’s home (n
= 198), at our research offices at the University of Rochester Medical Center (n
= 23), or at another location (n
At the private offices, 53 patients scored above the CES-D cutoff, and the SCID interview was completed with 26 (49%) of them; the remainder either refused the interview outright or stated they were unable to schedule the interview within the 4-week time frame. Of those scoring below the cutoff, 252 were selected randomly and approached for SCID interview, and 104 (41%) of these did complete this in-depth assessment. To allow group comparisons of medical illness severity, medical charts were reviewed by a physician-investigator (JML) on a random sample of 50 patients (25 above and 25 below the CES-D cutoff) who were approached but did not complete the SCID interview, and the Cumulative Illness Rating Scale (CIRS),25
a validated measure of overall organ system burden, was completed. (The CIRS was completed in similar fashion for all study participants.) Subjects scoring above the cutoff who completed the SCID assessments did not differ statistically from those not completing the SCID on any of the variables available for comparison (age, gender, CES-D score, CIRS, and visit type). Patients below the cutoff who completed the SCID interview did not differ from those completing the SCID on age, CES-D score, or visit type, but did have a greater proportion of men (45% vs 30%, p
= .017) and a higher score on the CIRS (mean 6.5 vs 4.6, p
At the Family Medicine Center, 49 patients scored above the CES-D cutoff, and the SCID interview was completed with 25 (51%) of them; again, the remainder either refused the interview outright or stated they were unable to schedule the interview within the 4-week time frame. Of those scoring below the cutoff, 143 were selected randomly for SCID interview, and 70 (49%) completed this in-depth assessment. Patients scoring above the cutoff who completed the SCID interview did not differ significantly from those not completing the interview on age, gender, or visit type, but did have a higher score on the CES-D (mean 31.1 vs 27.9, p = .047). Patients below the cutoff who completed the SCID interview did not differ on age, gender, or CES-D score, but did differ significantly on visit type (more likely to be seen for a scheduled follow-up appointment) (Fisher’s Exact Test, p = .007).
The in-depth diagnostic assessment was based on the SCID, a validated and reliable instrument that depends on rater clinical judgment, using all available sources of data including the patient’s report, family report, and medical records, to arrive at Axis I diagnoses. The SCID diagnoses used indicate current disorders, as well as certain fully remitted conditions (major depression and substance use disorders only) and partially remitted major depression (i.e., symptoms improved below the diagnostic threshold for major depression but still present at clinically significant levels). The SCID interview was administered by master’s-level prepared raters trained in the use of the SCID and other study measures by research personnel in the Department of Psychiatry’s Program in Geriatrics and Neuropsychiatry. A physician-investigator (JML) reviewed the primary care record for all subjects, along with all other available outpatient and inpatient medical and psychiatric records, to facilitate completion of study measures.
Study patients were presented by the SCID rater at a weekly consensus conference of program investigators, raters, and other research personnel. Consensus diagnoses were assigned based on the SCID. The SCID does not include a section to assess the diagnosis of dementia; consensus diagnoses of dementia were assigned using DSM-III-R criteria,26
based on patient (and when possible family) responses to specific probe questions about cognitive decline and related functional impairment, score on the Mini-Mental State Examination,27
and all other available clinical and laboratory data. As well, a diagnosis of current minor depressive disorder was assigned based on the SCID data, using the criteria proposed in the appendix to DSM-IV (the diagnostic criteria are identical to those for major depression, except that a minimum of two symptoms rather than five are required, with one symptom necessarily being depressed mood or diminished interest or pleasure).28
For all depressive diagnoses, an inclusive approach was used regarding symptoms. That is, symptoms were counted toward the criteria for depressive diagnoses without attempt to attribute them to “psychiatric” or “medical” causes, as has been recommended by our group and others,29
because of both the need to include the broad range of medical comorbidity in the study of later life depressions, and the arbitrary nature of most decisions regarding attribution. Thus, the diagnosis of secondary or “organic” mood disorder was not used.
Prevalence estimates for specific diagnostic groups were based on weighted combinations of stratum-specific rates because of the stratified sampling strategy, i.e., the weighting reconstructed the proportion of subjects scoring above and below the CES-D cutoff in the parent population for each diagnosis. Ninety-five percent confidence intervals for these prevalence estimates were calculated by using a normal approximation; several of the confidence intervals had lower limits that were negative numbers, but these lower limits are reported as 0% because a prevalence rate cannot be less than 0. Standard deviations for the prevalence estimates were obtained by appropriately weighting the variances for individual rates within each CES-D-defined stratum, which were calculated using the variance of the binomial distribution. Comparison of prevalence rates by gender used approximate Z tests. All reported p values are two-sided because of the number of comparisons performed.