Impact of coronavirus disease 2019-related clinic closures on HIV incidence in young adult MSM and transgender women in Kenya

Introduction: Little is known about the impact that the COVID-19 pandemic had on risk of HIV acquisition in sub-Saharan Africa. We assessed the impact of COVID-19-related clinic closures on HIV incidence in a cohort of gay, bisexual, and other men who have sex with men (MSM) and transgender women in Kenya. Methods: MSM and transgender women enrolled in a prospective, multicentre cohort study were followed quarterly for HIV testing, behaviour assessments, and risk. We estimated the HIV incidence rate and its 95% credible intervals (CrI) among participants who were HIV-negative before COVID-19-related clinic closure, comparing incidence rate and risk factors associated with HIV acquisition before vs. after clinic reopening, using a Bayesian Poisson model with weakly informative priors. Results: A total of 690 (87%) participants returned for follow-up after clinic reopening (total person-years 664.3 during clinic closure and 1013.3 after clinic reopening). HIV incidence rate declined from 2.05/100 person-years (95% CrI = 1.22–3.26, n = 14) during clinic closures to 0.96/100 person-years (95% CrI = 0.41–2.07, n = 10) after clinic reopening (IRR = 0.47, 95% CrI = 0.20–1.01). The proportion of participants reporting hazardous alcohol use and several sexual risk behaviours was higher during clinic closures than after clinic reopening. In multivariable analysis adjusting for study site and participant characteristics, HIV incidence was lower after clinic reopening (IRR 0.57, 95% CrI = 0.23–1.33). Independent risk factors for HIV acquisition included receptive anal intercourse (IRR 1.94, 95% CrI = 0.88–4.80) and perceived risk of HIV (IRR 3.03, 95% CRI = 1.40–6.24). Conclusion: HIV incidence during COVID-19-related clinic closures was moderately increased and reduced after COVID-19 restrictions were eased. Ensuring access to services for key populations is important during public health emergencies.


Introduction
The emergence of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the virus that causes coronavirus disease 2019 (COVID-19), in Kenya since March 2020 [1] led to interruptions in HIV prevention and treatment services, particularly those targeted toward men who have sex with men (MSM) and sex workers [2][3][4].A recent mathematical model predicted that disruption in HIV prevention, testing, and treatment services could lead to an acute increase in HIV incidence and HIV-related deaths over a period of up to 1 year in sub-Saharan Africa [5].To our knowledge, no data are available that can evaluate the impact of COVID-19 restrictions and related interruptions in HIV prevention services on HIV acquisition risk among MSM and transgender women in sub-Saharan Africa.We assessed differences in HIV incidence among HIV-negative MSM and transgender women participating in a multicentre cohort study in Kisumu, Nairobi, and coastal Kenya.

Study setting and participants
The study was conducted at the Anza Mapema clinic in Kisumu [6], the Sex Workers Outreach Programme (SWOP) City clinic in Nairobi, the KEMRI clinics at Mtwapa and Malindi (both in coastal Kenya).
Criteria for study eligibility included: assigned to male gender at birth and identifying as male or transgender female, aged 18-29 years, resident of Kisumu, Nairobi or Coast, HIV-negative, willing to provide informed consent, and had anal intercourse with a man in the past 3 months.Volunteers were assessed for eligibility between September 2019 and May 2021.
In response to COVID-19 lockdown measures, all study sites were closed by 23 March 2020.Sites began to reopen as determined by local approval: 20 and 21 October 2020 (Nairobi, and Kisumu), and 9 November 2020 (Malindi).The Mtwapa clinic remained closed for administrative reasons; research procedures for Mtwapa participants finally resumed on 26 October 2021 at a nearby community-based organisation (HAPA Kenya, Nyali, Mombasa) [7].During the lockdown period, participants had access to HIV prevention services through community-based programmes.

Study procedures
Individuals were followed quarterly, with HIV counselling and testing at each study visit.At enrolment and all subsequent study visits, participants completed an audio computer-assisted self-interview (ACASI) in English, Swahili, or Dholuo for sociodemographics, sexual behaviours, intimate partner violence, depressive symptoms [Patient Health Questionnaire 9 (PHQ-9)], alcohol use [Alcohol Use Disorder Identification Test (AUDIT)], use of substances other than alcohol and tobacco [Drug Abuse Screening Test 10 (DAST-10)], childhood abuse, sexual stigma (abridged China MSM Stigma Scale) [8], intimate partner violence (IPV), social harms, perceived HIV risks, use of preexposure prophylaxis (PrEP), and use of feminizing hormones.

Measures
Outcomes Follow-up period: Follow-up was divided into two periods: during COVID-19-related clinic closure, defined as time from date of site closure until the first visit that a participant made following reopening of the study site, and after COVID-19-related clinic closure, defined as all visits that occurred during follow-up after the initial visit following reopening of the study site.
Retained: A participant was defined as retained in the cohort if they returned to the clinic following study resumption.

HIV infection:
The estimated date of infection was calculated as described previously [9].

Covariates
Detailed information on covariates is provided in supplemental material, http://links.lww.com/QAD/D44.In short, data collected at enrolment included: age, gender identity, ever married to a woman, education, employment status, religion, and childhood abuse.We also obtained at each three-monthly visit: gender of last sexual partner (male or female), last male sexual partner category (regular, casual, paying or paid and partner numbers), condom use for receptive or insertive anal intercourse (RAI, IAI) (yes, no, no RAI, no IAI), and perceived risk of acquiring HIV (no chance at all to small chance, moderate chance to great chance).
The following were dichotomized (yes or no): RAI, IAI, receiving payment for sex, paying for sex, group sex, circumcision status, any intimate partner violence, other violence, PrEP use in the past 3 months, and AUDIT score at least 8.
The following information was collected at enrolment and/or half-yearly: moderate-to-severe depressive symptoms in the past 2 weeks (defined as having PHQ-9 score of 10-27), substance use disorder in past year (defined as having a DAST-10 score !3), childhood abuse and current use of feminizing hormones.Sexual stigma was collected at enrolment and half-yearly and measured using a continuous score ranging from 0 to 33.

Statistical analysis
We compared demographic and behavioural characteristics at enrolment of participants who did versus those who did not return for follow-up after COVID-19-related clinic closure using Pearson's x 2 test for categorical variables and Mann-Whitney U test for continuous variables.
Follow-up time began at the date of site closure and continued until a positive HIV test result, loss to followup, withdrawal from study participation or 20 December 2022, whichever occurred first.Data collected on the first visit upon which the participant returned to the clinic after it reopened were assumed to represent behaviours during clinic closure.In all longitudinal analysis, certain information (e.g.PHQ-9) collected at baseline and halfyearly thereafter were carried forward to their subsequent quarterly visits.
Due to the few incident cases, we suspected that parameter estimates from standard regression techniques could be uncertain [10].To minimize this bias, we used a penalized regression approach whereby uncertain estimates from the data are pulled towards more realistic ones with the use of prior distributions [11].Therefore, to estimate the HIV incidence rate, incidence rate ratios (IRR) and the 95% credible intervals (CrI), we fitted a Poisson model using a Bayesian approach.
We calculated the proportion of individuals reporting any RAI, condomless RAI, any IAI, condomless IAI, PrEP use during the past 3 months and those with a PHQ-9 score of 10-27, AUDIT score at least 8 and DAST-10 score at least 3.We modelled these endpoints during clinic closure and after reopening.We calculated odds ratios (OR) and their 95% confidence intervals (CI) comparing the odds of having an endpoint for the two timepoints and tested for changes between timepoints using a Wald x 2 test.
The KEMRI Ethics Review Committee approved the study.

Results
Most (70.8%) of the 794 cohort participants were 18-24 years old, and about 1 in 10 (12.8%) identified as transgender women.Overall, 690 (86.9%) returned to follow-up after clinic reopening (details in Supplemental Table 1, http://links.lww.com/QAD/D43).Compared with participants who were lost to follow-up, participants who returned were less likely to have been recruited from Mtwapa (P < 0.001), to be circumcised (P ¼ 0.009), and to have reported IPV (P ¼ 0.009), and more likely to have harmful or hazardous drinking by AUDIT score (P ¼ 0.033) or to report a moderate-to-great perceived risk of HIV acquisition (P < 0.001).Supplemental Table 2, http://links.lww.com/QAD/D43presents cohort characteristics by site.One in seven (14%) participants had a PHQ-9 score of at least 10, indicating moderate-to-severe depressive symptoms.Over a third (37%) of participants had an AUDIT score of at least 8, suggesting hazardous alcohol use.Three in 10 (30%) participants reported a DAST-10 score of at least 3, indicating problematic substance use in the past year.Participants reported a median stigma score of 7 (IQR 3-13), and two-thirds (64%) reported any childhood abuse.

Discussion
Among MSM and transgender women followed up in Kenya, we assessed the impact of COVID-19-related clinic closures on HIV-acquisition risk.The overall HIV incidence was moderately high during COVID-19 restrictions and had an approximately 43% reduction after the clinics reopened.Although a minority (13%) of study participants did not return to follow-up, it is uncertain how lost participants impacted our HIV incidence estimate.Compared with the period after COVID-19 restrictions eased and clinics reopened, the proportion of participants reporting RAI, IAI, and condomless RAI, and IAI, and hazardous alcohol use was substantially higher during restrictions, in the majority (76.5%) of the cohort (Nairobi and Kisumu).
That HIV incidence was higher during clinic closure may be at least in part because of difficulty accessing services provided at the cohort sites, including distribution of condoms and lubricants, HIV counselling and testing, and PrEP, as well as antiretroviral therapy (ART) for individuals who test positive at initial screening or during follow-up.These services were lacking entirely in Mtwapa, the site that remained closed following lifting of the COVID-19 lockdown restrictions.
We found that the risk of HIV acquisition was two-fold higher among participants who reported moderate-togreat perceived risk of HIV than that among participants who reported no-to-small risk.This suggests that despite awareness of their HIV acquisition risk, these participants faced challenges in effectively using HIV prevention tools [12].Although PrEP for HIV prevention was freely available to all participants, only one in four participants reported PrEP use at baseline and PrEP use remained low throughout the study period.Earlier, we reported that among participants receiving PrEP during 2017-2019, only 14.5% had sufficient tenofovir diphosphate in dried blood spots to confer protection [13].Similarly, a more recent study of MSM in a PrEP program in Kisumu found that just 14.6% participants had protective TFV-levels in DBS samples [14].
Our study had several limitations.Firstly, 13% of study participants did not return to follow-up after COVID-19 restrictions, and it is uncertain how this impacted our incidence estimate.Secondly, during the COVID-19 restrictions when the study clinics remained closed, we could not collect research data; hence, it is uncertain how risk behaviours and psychosocial factors changed during this period.Lastly, we imputed certain measures collected at baseline and half-yearly thereafter based on the last observation carried forward, which is likely to have led to an underestimation of changes in these psychosocial factors during the clinic closure period.
In conclusion, in this large multicentre study of MSM and transgender women in Kenya, we documented a moderately high HIV incidence during COVID-19related clinic closures when alcohol use and sexual risk behaviour was increased.HIV incidence reduced to a baseline estimate when clinics reopened.HIV incidence varied by location and was markedly higher in Mtwapa, where services were the most disrupted.Although a reduction in HIV incidence may be in keeping with an overall trend for reducing incidence in these key populations, measures to upscale and ensure uninterrupted HIV prevention and treatment services for MSM and transgender women should be prioritized.Forty-six percent of data collected every 6 months were carried forward to the subsequent quarterly return visit from before COVID visit.
manuscript.All authors have read and provided feedback on manuscript drafts and have approved the final manuscript.
Tatu Pamoja (three site) study group authors: Oscar , pre-exposure prophylaxis use, mental health and substance use by follow-up period Compared with the period after COVID-19 restrictions, the proportion of participants reporting RAI, IAI, AIDS 2024, Vol 38 No 3

a
During COVID-19 restrictions: between 24 March 2020 and the first visit after reopening of study sites (i.e.19 October 2020 for Nairobi, 21 October 2020 for Kisumu, 9 November 2020 for Malindi and 26 October 2021 for Mtwapa).After COVID-19 restrictions: all visits that occurred during follow-up after the initial visit following reopening of the study site.b

Table 1 .
Risk factors for HIV acquisition among 690 Kenyan MSM and transgender women who returned for follow-up, March 2020 to December 2022.