Assessment of strategies for identifying diagnosed cases of systemic lupus erythematosus through self-report.

Abstract

The objective of this work was to assess the optimal way to identify potential systemic lupus erythematosus (SLE) cases in large epidemiologic studies through self-reported information about diagnosis of SLE, symptoms and medications, and to investigate the utility of a criteria checklist sent directly to participants' physicians. We used data collected in 1997 from 53322 participants in a study of African-American women, the Black Women's Health Study, including a lupus screening questionnaire (LSQ) and questions about SLE diagnosis and medications. We confirmed self-reported SLE through medical records and criteria checklists sent to participants' physicians. Among those for whom we received medical records and/or criteria checklists, we compared the predictive value and proportion of missed cases of several algorithms using combinations of self-reported SLE diagnosis, LSQ score and medication use to self-reported SLE diagnosis alone. We obtained a physician checklist or medical chart for 251 individuals who reported SLE, of whom 212 (84%) fulfilled ACR criteria for definite or probable SLE, or had clinical lupus (SLE diagnosis recorded in medical charts plus appropriate medication use). The use of LSQ score or medication use in addition to self-report of SLE tended to decrease the false positive rate but also to reduce the proportion of true cases identified. Checklists of ACR criteria completed by subjects' physicians documented more criteria than medical records. In conclusion, among participants who consented to medical record review, SLE prediction algorithms using questions about lupus symptoms and medications offered slightly higher predictive value for detecting cases than self-reported diagnosis alone, but at the cost of case detection. SLE case confirmation strategies can be complemented by the use of criteria checklists sent directly to participants' physicians.