ABSTRACT
The estimated burden of chronic disease among people living with HIV (PLWH) varies considerably by data source, due to differences in case definitions, analytic approaches, and underlying patient populations. We evaluated the burden of diabetes (DM) and chronic kidney disease (CKD) in two large data systems that are commonly queried to evaluate health issues affecting HIV care patients: the Medical Monitoring Project (MMP), a nationally representative sample, and the Centers for AIDS Research Network of Integrated Clinical Systems (CNICS), a clinical cohort. In order to reconcile these two data sources, we addressed issues common to observational data, including selection bias, missing data, and development of case definitions. The overall adjusted estimated prevalence of DM and CKD in MMP was 12.7% and 7.6%, respectively, and the overall prevalence of DM and CKD in CNICS was 9.9% and 8.3%, respectively; prevalence estimates increased with age in both data sources. After reconciling the approach to analyzing MMP and CNICS data, sub-group specific prevalence estimates of DM and CKD was generally similar in both data sources. Both data sources suggest a considerable burden of disease among older adults in HIV care. MMP and CNICS can provide reliable data to monitor HIV co-morbidities in the US.
Acknowledgements
The authors would like to acknowledge the staff, advisory boards, and participants of the Medical Monitoring Project: http://www.cdc.gov/hiv/statistics/systems/mmp/resources.html#StudyGroupMembers.
Disclosure statement
GB has received research support from Bristol Myers Squibb and Amgen, Inc. and consulted for Definicare, LLC and Medscape. No other potential conflict of interest was reported by the authors.
Notes
1 Standard error was calculated using Rubin’s Rules (Rubin, Citation1987); the Taylor Series approach was used to obtain variance estimates for MMP, accounting for its survey design (Vizcarra&Sukasih, Citation2013).
2 The following variables were included in the multiple imputation (MI) models: age, sex, race, ethnicity, risk transmission category, region, body mass index, site, average eGFR, average systolic blood pressure, average diastolic blood pressure, CKD status, DM status, HTN status, and dyslipidemia status. The MI models implemented on MMP data also included the following variables that were unavailable in CNICS: dialysis, reported ever using injection drugs, reported ever engaging in male-to-male sex, birth country, and years since HIV diagnosis. The inclusion of these five additional variables did not meaningfully affect the prevalence estimates generated from the multiple imputation models executed on MMP data.