The diffusion of innovation in AIDS treatment: zidovudine use in two New Jersey cohorts

This study investigates patterns of utilization of zidovudine (ZDV) by gender, race, risk group, and other respondent characteristics following approval of this treatment.

Longitudinal observational data were used on a demographically diverse population participating in New Jersey's Medicaid waiver program for persons with symptomatic HIV disease.

Claims data were merged with administrative data on demographic characteristics, risk group, and functional status. Periods of ZDV utilization were determined by analysis of pharmacy claims.

The proportion of respondents ever using ZDV (treatment incidence) and the proportion of time on ZDV among users (treatment persistence) were analyzed for a cohort enrolling in 1987 and 1988, and for a cohort enrolling in 1989 and 1990, with follow-up of utilization through August 1992. For each cohort, bivariate analyses were used to compare incidence and persistence by patient subgroup; logistic regression was used to investigate the predictors of incidence in a multivariate model; and OLS regression was used to analyze proportion of time on ZDV among those with any ZDV use.

For the 1987-1988 cohort, substantial race, gender, and risk group differences in utilization were observed, even though all participants in this Medicaid population had financial coverage for ZDV treatment. Treatment incidence was significantly lower for blacks than for others in bivariate comparison (45 percent versus 63 percent had any use of ZDV) and in a logistic regression controlling for a variety of demographic and health status indicators (relative risk .46, CI .31 a variety of demographic and health status indicators (relative risk .46, CI .31 to .69). Treatment persistence differences were also substantial in the 1987-1988 cohort: among ZDV users, women, blacks, and injection drug users (IDUs) had significantly less persistence in use, and the gender and risk group differences were significantly in a multivariate model. In the 1989-1990 cohort, however, both incidence and persistence of treatment converged: no significant differences were observed across demographic groups.

Less-advantaged subgroups lagged in access to this new therapy, suggesting the presence of nonfinancial barriers to care. However, these initial differences subsequently converged.

Socioeconomic differences have been observed in access to newly introduced treatments for a variety of diseases, reflecting nonfinancial as well as financial barriers to care. Such differences may or may not disappear as use of therapies becomes institutionalized. Monitoring patterns of treatment initiation as well as persistence of treatment over time, using merged data from claims and administrative files, can provide important information on the diffusion of treatments and the extent to which initial disparities are or are not reduced over time.