Viral suppression rates in a safety-net HIV clinic in San Francisco destabilized during COVID-19 : AIDS

Secondary Logo

Journal Logo

Research Letters

Viral suppression rates in a safety-net HIV clinic in San Francisco destabilized during COVID-19

Spinelli, Matthew A.a; Hickey, Matthew D.a; Glidden, David V.b; Nguyen, Janet Q.a; Oskarsson, Jon J.a; Havlir, Dianea; Gandhi, Monicab

Author Information
AIDS 34(15):p 2328-2331, December 1, 2020. | DOI: 10.1097/QAD.0000000000002677



The COVID-19 pandemic is expected to hinder US End the HIV Epidemic (EHE) goals [1] via disruptions in care. Transition to phone or video visits in HIV clinics during shelter-in-place ordinances could prove particularly difficult for vulnerable populations who rely on clinic-based social support services and for whom telemedicine is less accessible.

Ward 86 is a safety-net clinic in San Francisco serving people with HIV (PWH) on publicly funded insurance; our population has a high prevalence of mental illness, substance use and unstable housing [2]. On 17 March 2020, the city put a moratorium on routine medical care appointments [3] one day following the San Francisco shelter-in-place ordinance [4]. Ward 86 immediately instituted a telemedicine model for primary care. We examined trends in viral nonsuppression and retention-in-care for PWH after the San Francisco shelter-in-place ordinance in our large, urban clinic.

Materials and methods

All clinic visits were transitioned to phone except when in-person visits were requested by providers or patients. Clinic management facilitated viral load testing via rapid laboratory visits, with at least quarterly monitoring recommended during shelter-in-place. This analysis compares viral loads and retention-in-care, defined as no-shows for scheduled in-person/telephone visits, before and during shelter-in-place (1 December 2019–29 February 2020 vs. 1 April 2020–30 April 2020) [5]. To identify factors associated with nonsuppressed viral load at least 200 copies/ml and no-show visits, we used mixed effects logistic regression accounting for clustering across visits and lab draws. Factors for which the association with outcomes differed pre/postshelter-in-place are presented with separate effect estimates by time-interval (interaction P < 0.01) [6].

Propensity score analysis for primary viral load outcome

As an alternate modelling strategy for the viral load outcome, we first constructed a propensity score model to estimate each patient's probability of having a viral load measured pre/during shelter-in-place, adjusting for sex, race/ethnicity, primary language (English vs. non-English) and homeless status (all binary/categorical); while age and CD4+ cell count were included using cubic splines. After propensity scores were stratified into quintiles, we checked for evidence of interaction between the exposure and propensity scores; balance for key confounders was examined across these five strata. We then calculated the marginal odds ratio for viral nonsuppression using postestimation commands. Finally, to test the robustness of these findings, we conducted a trimmed analysis restricted to the overlap between propensity scores in the pre/during-shelter-in-place groups.


Ward 86 averaged 1836 visits/month over 2019. Overall, 344 of 1766 (19%) individuals had unsuppressed viral loads at any point and 16% were homeless. After the transition to telemedicine, 54% of scheduled visits were telephone visits, although homeless individuals were offered telehealth for only 32% of visits (P < 0.001). Pre-shelter-in-place, 1287 of 4153 visits (31%) were no-shows compared with 599 of 1997 (30%) post-shelter-in-place. Fewer no-shows occurred for telephone vs. in-person visits [adjusted odds ratio (aOR) 0.56; 95% confidence interval (95% CI) = 0.36–0.86; Table 1]. Younger individuals (age < 35 years) were more likely to no-show pre-COVID-19 (aOR 1.57; 95% CI = 1.28–1.93), but not during shelter-in-place (aOR 1.11; 95% CI = 0.82–1.51). Homeless individuals had fewer no-shows during shelter-in-place (aOR 0.64; 95% CI = 0.48–0.85) but not pre-COVID-19 (aOR 1.15; 95% CI = 0.95–1.32).

Table 1 - Factors associated with unsuppressed viral load and no-show visits before and after shelter-in-place/COVID-19a.
Factor Unsuppressed viral load adjusted odds ratio; 95% Confidence Interval No-show visitAdjusted odds ratio; 95% confidence interval
Post-COVID-19 vs. pre-COVID-19 1.31; 1.08–1.53 0.91; 0.77–1.09
Age under 35a 1.29; 1.11–1.51 1.57; 1.28–1.93 (Pre-COVID-19)1.11; 0.82–1.51 (Post-COVID-19)
Female vs. male birth sex 0.94; 0.77–1.15 0.99; 0.80–1.21
Race/ethnicity vs. white
 Black 1.60; 1.33–1.91 1.14; 0.94–1.38
 Latin 1.04; 0.63–1.34 1.06; 0.88–1.27
 Asian 0.92; 0.63–1.34 1.16; 0.82–1.64
 Other 0.96; 0.78–1.19 0.97; 0.77–1.24
Homeless housing statusa 2.27; 1.91–2.71 (Pre-COVID-19)3.36; 2.74–4.12 (Post-COVID-19) 1.15; 0.95–1.32 (Pre-COVID-19)0.64; 0.48–0.85 (Post-COVID-19)
Telephone vs. in-person visits (post-COVID-19 only) 0.56; 0.36–0.86
aEach factor was tested for an interaction with the pre/post COVID-19 time interval indicator. Adjusted odds ratios and 95% confidence intervals are presented separately for before and during COVID-19 time intervals if the test of interaction P-value was <0.1 [6].

For the viral load outcome, the odds of viral nonsuppression were 31% higher during shelter-in-place [(aOR 1.31; 95% CI = 1.08–1.53) than before COVID-19 in adjusted analyses. Viral nonsuppression was also higher among homeless individuals during-COVID-19 (aOR 3.36; 95% CI = 2.74–4.12) vs. pre-COVID-19 (aOR 2.27; 95% CI = 1.91–2.71). Age less than 35 years and black vs. white race were each associated with higher odds of viral nonsuppression (aOR 1.29; 95% CI = 1.11–1.51 and aOR 1.60; 95% CI = 1.3–1.91, respectively), but these associations did not differ pre/post COVID-19 (P = 0.49 and 0.93, respectively; Table 1). In the alternate modelling strategy using propensity scores, the association between shelter-in-place conditions and higher odds of viral nonsuppression was nearly identical (aOR 1.32; 95% CI = 1.14–1.51). Limiting the analysis to propensity scores within the region of overlap also had a minimal impact on results (aOR 1.31; 95% CI = 1.14–1.51; Supplemental Appendix,


After institution of a shelter-in-place ordinance for COVID-19 in San Francisco, viral suppression rates fell substantially compared with pre-COVID-19 in a large HIV clinic serving vulnerable populations. The odds of viral nonsuppression are now 31% higher than before the pandemic. This destabilization occurred despite our population attending telemedicine visits at a higher rate than expected, given the 60% drop in ambulatory care visit volume nationwide [7]. Telehealth visits, while offering greater patient convenience, may lead to less access to clinic-based social support services essential to achieving viral suppression among vulnerable groups [8].

Concomitant with this explanation, homeless individuals at Ward 86 had higher odds of unsuppressed viral loads post-COVID-19 vs. pre-COVID-19, despite higher visit attendance. The disproportionate economic impact of the shutdown on those with housing instability, coupled with depopulation of San Francisco shelters with COVID-19 outbreaks [9], are expected to destabilize viral suppression, despite ongoing or increased healthcare utilization by this group. Younger individuals had higher retention-in-care post-COVID-19, possibly related to comfort with telemedicine, but not higher viral suppression. Black individuals had persistent, unchanged disparities in viral suppression compared with white individuals, in spite of similar visit volume.

Limitations of our study included its nonrandomized observational pre/post design and the short time intervals analysed. We attempted to account for the possibility that individuals with worse HIV health status might be more likely to present for viral load monitoring by adjusting for markers of disease severity (e.g. CD4+ cell count) in our propensity score model. This analysis did not substantially impact our estimates, although residual, unmeasured confounders could still exist. Studies over longer periods will be needed to understand the full impact of COVID-19 on HIV outcomes.

In conclusion, viral suppression rates fell during COVID-19 in an urban HIV clinic serving publicly insured patients. Telemedicine may facilitate retention-in-care in the context of shelter-in-place for those without a digital divide, but is unlikely to compensate for the loss in clinic-based social services and support for PWH with vulnerabilities. In our population, retention-in-care via telemedicine was not sufficient to keep suppression rates stable. Loss of viral suppression is deleterious to the individual and hinders treatment-as-prevention: the COVID-19 pandemic is threatening the goals of the US EHE initiative. Measures to counteract the effect of COVID-19 on HIV care outcomes are urgently needed [10,11].


This research was supported by US National Institute of Allergy and Infectious Diseases (NIAID) R01AI158013 (M.A.S. and M.G.), and US NIAID P30AI027763 (M.G.). M.A.S., M.D.H. and M.G. conceived the analysis, M.A.S. wrote the first draft of the article. M.A.S. and D.V.G. performed the statistical analysis. J.Q.V. and J.J.O. maintained the clinical database. M.G. and D.H. oversaw clinical research operations. All authors reviewed the manuscript and provided edits.

Conflicts of interest

The authors have no conflicts of interest to report.


1. Fauci AS, Redfield RR, Sigounas G, Weahkee MD, Giroir BP. Ending the HIV epidemic: a plan for the United States. JAMA 2019; 321:844–845.
2. Clemenzi-Allen A, Geng E, Christopoulos K, Hammer H, Buchbinder S, Havlir D, et al. Degree of housing instability shows independent ‘Dose-Response’ with Virologic suppression rates among people living with human immunodeficiency virus. Open Forum Infect Dis 2018; 5:ofy035.
3. Order of the Health Office No. C19-08. Moratorium on routine medical appointments and elective surgery, encouraging remote appointments. 17 March 2020. [Accessed 25 June 2020]
4. Order of the Health Officer No. C19-07. Directing all individuals living in the county to shelter at their place of residence. San Francisco Department of Public Health. 16 March 2020. [Accessed 25 June 2020]
5. Mugavero MJ, Westfall AO, Cole SR, Geng EH, Crane HM, Kitahata MM, et al. Beyond core indicators of retention in HIV care: missed clinic visits are independently associated with all-cause mortality. Clin Infect Dis 2014; 59:1471–1479.
6. Selvin S. Statistical analysis of epidemiologic data. 1996; New York, NY: Oxford University Press, 213–214.
7. Mehrotra A, Chernew M, Linetsky D, Hatch H, Cutler D. The impact of the COVID-19 pandemic on outpatient visits: a rebound emerges. The Commonwealth Fund. [Accessed 10 June 2020]
8. Christopoulos KA, Massey AD, Lopez AM, Geng EH, Johnson MO, Pilcher CD, et al. Taking a half day at a time:’ patient perspectives and the HIV engagement in care continuum. AIDS Patient Care STDS 2013; 27:223–230.
9. Fuller T. Major outbreak in San Francisco shelter underlines danger for the homeless. New York Times. [Accessed 10 June 2020].
10. UNAIDS. The impact of the COVID-19 response on the supply chain, availability and cost of generic antiretroviral medicines for HIV in low- and middle-income countries. UNAIDS 22 June 2020. [Accessed 25 June 2020]
11. UNAIDS. Modelling the extreme—COVID-19 and AIDS-related deaths. 25 May 2020. [Accessed 25 June 2020]

Supplemental Digital Content

Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.