Search

Capture-recapture analysis was applied to medical information and survey self-report in order to estimate the one-year prevalence of substance abuse related health problems among 31861 health plan members and to evaluate how effectively survey and medical information identified substance abuse. The approach used the ongoing nature of data collection in the health plan to create two waves of identification with two indicators for each wave: one based on relevant treatment and one on diagnoses. Using the resulting symmetries in the indicators, we found that the resulting estimates were fairly robust to the exact form of the loglinear model used to estimate prevalence. One year prevalence of substance abuse related health problems based on these estimates was similar for men (7.0%) and women (5.8%) and peaked in mid-life at 7.9% compared with 6.4% for under 35 and 4.7% for 60+. Consistent with differentials in stigma, men with substance abuse problems appeared much more likely to report them (22.1% vs 11.8%) and reporting declined sharply with age, from 25.7% under 35 to 11.2% over 60. Medical identification was almost equally effective for men and women, with about one quarter of each identified during the course of the year, but declined with age: 28% under 35 to 22% over 60. Overall identification rates also differed demographically, but less dramatically than self-report rates. The study suggests that using medical information to monitor substance abuse is promising. ..PAT.-Unpublished Manuscript [ABSTRACT FROM AUTHOR]