Racial/Ethnic and Income Disparities in the Prevalence of Comorbidities That Are Associated With Risk for Severe COVID-19 Among Adults Receiving HIV Care, United States, 2014–2019

John K. Weiser, MD, MPH; Yunfeng Tie, PhD; Linda Beer, PhD; Robyn Neblett Fanfair, MD, MPH; Roy Luke Shouse, MD, MPH


J Acquir Immune Defic Syndr. 2021;86(3):297-304. 

In This Article


For several reasons, our primary outcome likely underestimates the increased risk of severe illness from COVID-19 among adults receiving HIV care. First, to measure comorbidities, we used only diagnosis data instead of supplementing with medical record information that can identify some conditions, even when not diagnosed, eg, laboratory data for CKD and height and weight for obesity. Our estimate of diagnosed obesity (12.3%) is lower than a previous MMP estimate that calculated the prevalence of body mass index ≥30 kg/m2 using height and weight documented in the medical record (19% of men and 42% of women were obese).[46] Because some conditions (eg, COPD) can only be identified using diagnosis data, for consistency, we measured all comorbidities in the same way to avoid information bias of the prevalence of comorbidities, relative to each other.

Second, we included comorbidities known to increase the risk of severe illness as of July 15, 2020, and there may have been subsequent additions to the list. We did not include diagnosed conditions that might be associated with the risk for severe illness, but for which there is currently insufficient evidence to support inclusion on the CDC list of conditions that definitely increase the risk,[16] including poorly controlled HIV, which affects one in 3 adults receiving HIV care.

Third, we did not include older age as a risk factor for severe illness from COVID-19. Because the relationship between age and risk is continuous,[33] we did not include an age cut-off when formulating the primary outcome.

For these reasons, our estimate that 37.9% of adults receiving HIV care had ≥1 diagnosed comorbidity should be considered a minimum estimate of disease burden. Underestimation of risk could have biased our estimates of PDs across race/ethnicities, income levels, and types of health insurance. However, because several of the comorbidities we assessed are more prevalent among Blacks than Whites with HIV,[32,46,47] underestimation of their prevalence would likely have biased differences toward the null.

Strengths of our study included population-based sampling and weighting designed to produce nationally representative estimates of a wide range of sociodemographic and clinical characteristics of adults with diagnosed HIV, of which those receiving HIV care is a subset. The large sample size allows reasonably precise estimates of subgroup characteristics. Although not all sampled persons participated, results were adjusted for nonresponse using the standard methodology.[48,49] Even with suboptimal response rates, there is still value in results obtained from unbiased sampling methods.[50]