Measuring concurrency: an empirical study of different methods in a large population-based survey and evaluation of the UNAIDS guidelines.

BACKGROUND: Recent UNAIDS guidelines recommend measuring concurrency 6 months before the interview date, based on overlapping partnership dates. This has theoretical advantages, but little is known about how well it can be measured in practice.
METHODS: The assumptions underlying the UNAIDS measure were tested using data from a sexual behaviour survey conducted in rural northern Malawi. All resident adults aged 15-59 were eligible. Questions included self-reported concurrency and dates for all marital and nonmarital partnerships in the past 12 months.
RESULTS: A total of 6796 women and 5253 men were interviewed, 83 and 72% of those eligible, respectively. Since few women reported multiple partners, detailed analysis was restricted to men. Overall 19.2% [95% confidence interval (CI) 18.1-20.2] of men self-reported concurrent relationships in the past year (almost all of those with more than one partner). Using overlapping dates the estimate was 16.7% (15.7-17.7). Excluding partnerships which tied on dates (making overlap uncertain) or restricting the analysis to the three most recent partners gave similar results. The UNAIDS 6-month measure was 12.0% (11.1-12.9), and current concurrency was 11.5% (10.6-12.4). The difference between dates-based and self-reported 12-month measures was much larger for unmarried men: 11.1% (9.7-12.4) self-reported; 7.1% (6.9-8.2) on dates. Polygyny (15% of married men) and the longer duration of relationships stabilized the estimates for married men. Nonmarital partnerships were under-reported, particularly those starting longer ago.
CONCLUSIONS: The difficulties of recall of dates for relationships, and under-reporting of partners lead to underestimation of concurrency using date-based measures. Self-reported concurrency is much easier to measure and appears more complete.