Data collection for sexually transmitted disease diagnoses: a comparison of self-report, medical record reviews, and state health department reports.

To compare three methods of data collection on case ascertainment of past chlamydia or gonorrhea diagnoses. Data collection for 361 adolescent females between 1998 and 2000 included: 1) face-to-face interviews; 2) computerized and paper medical record reviews; and 3) chlamydia and gonorrhea reports to the state health department. Statistical methods include latent class and composite reference standard analyses. The estimated prevalence of past diagnoses did not differ significantly by data collection method for chlamydia (20.5%, 23.0%, and 19.7% by self-report, medical record reviews, and state health department reports, respectively) or gonorrhea (4.7%, 6.9%, and 5.5%, respectively) during the 2-year study period. The estimated latent class and composite reference standard prevalences for chlamydia were 23.5% and 26.9%, respectively (p=.04 and p < .01 for differences from self-report alone, respectively). For gonorrhea, the estimated latent class and composite reference standard prevalences were 7.8% and 6.9%, respectively (p < .01 for both differences from self-report alone). Kappa scores for self-report compared with the latent class and composite reference standard prevalences ranged from .67 to .80, and the magnitude of under-reporting ranged from 21% to 47%. The similar case ascertainment from the three sources separately and high reliability of self-report, coupled with its feasibility and low cost, suggest that self-report is a viable data collection method for STD diagnoses. However, using multiple sources may be preferable when time and resources permit given that under-reporting by self-report is likely to occur (particularly for gonorrhea) and that greater case ascertainment can be achieved.