A TWO-PART MODEL FOR THE COSTS OF HEALTH RESOURCE UTILIZATION AMONG INDIVIDUALS WITH HIV/AIDS

Sunday, October 20, 2013
Key Ballroom Foyer (Hilton Baltimore)
Poster Board # P1-35
Applied Health Economics (AHE)

Bohdan Nosyk, Ph.D.1, Viviane D. Lima, PhD2, Guillaume Colley, MSc3, Benita Yip, BSc3, Robert Hogg, PhD4 and Julio Montaner, MD4, (1)Simon Fraser University - Faculty of Health Sciences, Vancouver, BC, Canada, (2)BC Centre for Excellence in HIV-AIDS, Vancouver, BC, Canada, (3)BC Centre for Excellence in HIV/AIDS, Vancouver, BC, Canada, (4)Centre for Excellence in HIV/AIDS Research, Vancouver, BC, Canada
Purpose: Through delayed HIV disease progression, Highly Active Antiretroviral Therapy (HAART) may reduce other direct medical costs, thus at least partially offsetting the costs of therapy.  Recent findings regarding the secondary preventive benefits of HAART have necessitated careful consideration of resource allocation decisions for HIV/AIDS prevention in both developed and developing world settings.  Our objective is to estimate non-HAART direct medical costs at different levels of disease progression and over time in British Columbia (BC), Canada throughout the HAART era.  

Method: We considered the population of individuals identified as HIV-positive and engaged in  HIV care within a set of linked HIV disease registries and health administrative databases, drawn from 1996-2010.    Multivariate analyses controlling for a range of demographic and clinical indicators were executed to identify differences in quarterly, non-HAART direct medical costs across CD4 count strata.  Two-part models, estimated using generalized linear modeling, were constructed to account for excess zeroes and skewness in cost data.  The first stage modeled the probability of having a non-zero cost using a logit model specification, while the second estimated the level of non-zero costs that occurred, using a gamma distribution and log link.  Net effects of CD4 strata on direct non-HAART medical costs were estimated from the two-part models, evaluated at mean values of all covariates.

Result: Our analysis included N=9,018 individuals with a mean follow-up of 8.56 years.  The majority of the sample (80.4%) was male, the mean age at HIV diagnosis was 38.8(SD:11.0), and 45.4% had accessed health services for substance abuse. Compared to person-quarters in which CD4 was >500, non-HAART medical costs were $50(95%CI:$50,$50) greater for CD4:350-500; $185($185,$186) greater for CD4 200-350 and $731($730,$731) greater when CD4<200.  Finally, those with no measured CD4 (prior to linkage to HIV care) had costs $362($361,$362) greater than in periods with CD4>500.  Separate analyses on domain-specific costs found hospitalization and physician billing costs comprised the incremental cost amongst those with no measured CD4.   

Conclusion: Non-HAART direct medical costs increase substantially as a function of disease progression and are higher among those not yet linked to HIV care.  Health economic models comparing HIV treatment and prevention initiatives require detailed data on health resource utilization to inform funding allocation decisions.