Objective: The objective of this study was to examine primary care physicians' propensity to assess their elderly patients for depression using data from videotapes and patient and physician surveys.

Study design: An observational study was informed by surveys of 389 patients and 33 physicians, and 389 videotapes of their clinical interactions. Secondary quantitative analyses used video data scored by the Assessment of Doctor-Elderly Patient Transactions system regarding depression assessment. A random-effects logit model was used to analyze the effects of patient health, competing demands, and racial and gender concordance on physicians' propensity to assess elderly patients for depression.

Results: Physicians assessed depression in only 14% of the visits. The use of formal depression assessment tools occurred only 3 times. White patients were almost 7 times more likely than nonwhite patients to be assessed for depression (odds ratio [OR], 6.9; P < 0.01). Depression assessment was less likely if the patient functioned better emotionally (OR, 0.95; P < 0.01). The propensity of depression assessment was higher in visits that covered multiple topics (OR, 1.3; P < 0.01) contrary to the notion of competing demands crowding out mental health services. Unexpectedly, depression assessment was less likely to occur in gender and racially concordant patient-physician dyads.

Conclusions: Primary care physicians assessed their elderly patients for depression infrequently. Reducing the number of topics covered in visits and matching patients and physicians based on race and gender may be counterproductive to depression detection. Informed by videotapes and surveys, our findings offer new insights on the actual care process and present conclusions that are different from studies based on administrative or survey data alone.