Tree-based subgroup discovery using electronic health record data: heterogeneity of treatment effects for DTG-containing therapies.

The rich longitudinal individual level data available from electronic health records (EHRs) can be used to examine treatment effect heterogeneity. However, estimating treatment effects using EHR data poses several challenges, including time-varying confounding, repeated and temporally non-aligned measurements of covariates, treatment assignments and outcomes, and loss-to-follow-up due to dropout. Here, we develop the subgroup discovery for longitudinal data algorithm, a tree-based algorithm for discovering subgroups with heterogeneous treatment effects using longitudinal data by combining the generalized interaction tree algorithm, a general data-driven method for subgroup discovery, with longitudinal targeted maximum likelihood estimation. We apply the algorithm to EHR data to discover subgroups of people living with human immunodeficiency virus who are at higher risk of weight gain when receiving dolutegravir (DTG)-containing antiretroviral therapies (ARTs) versus when receiving non-DTG-containing ARTs.

Enabling Adherence to Treatment (EAT): a pilot study of a combination intervention to improve HIV treatment outcomes among street-connected individuals in western Kenya.

BACKGROUND: Street-connected individuals (SCI) in Kenya experience barriers to accessing HIV care. This pilot study provides proof-of-concept for Enabling Adherence to Treatment (EAT), a combination intervention providing modified directly observed therapy (mDOT), daily meals, and peer navigation services to SCI living with HIV or requiring therapy for other conditions (e.g. tuberculosis). The goal of the EAT intervention was to improve engagement in HIV care and viral suppression among SCI living with HIV in an urban setting in Kenya. METHODS: This pilot study used a single group, pre/post-test design, and enrolled a convenience sample of self-identified SCI of any age. Participants were able to access free hot meals, peer navigation services, and mDOT 6 days per week. We carried out descriptive statistics to characterize participants' engagement in EAT and HIV treatment outcomes. We used McNemar's chi-square test to calculate unadjusted differences in HIV outcomes pre- and post-intervention among participants enrolled in HIV care prior to EAT. We compared unadjusted time to initiation of antiretroviral therapy (ART) and first episode of viral load (VL) suppression among participants enrolled in HIV care prior to EAT vs. concurrently with EAT using the Wilcoxon rank sum test. Statistical significance was defined as p < 0.05. We calculated total, fixed, and variable costs of the intervention. RESULTS: Between July 2018 and February 2020, EAT enrolled 87 participants: 46 (53%) female and 75 (86%) living with HIV. At baseline, 60 out of 75 participants living with HIV (80%) had previously enrolled in HIV care. Out of 60, 56 (93%) had initiated ART, 44 (73%) were active in care, and 25 (42%) were virally suppressed (VL < 1000 copies/mL) at their last VL measure in the 19 months before EAT. After 19 months of follow-up, all 75 participants living with HIV had enrolled in HIV care and initiated ART, 65 (87%) were active in care, and 44 (59%) were virally suppressed at their last VL measure. Among the participants who were enrolled in HIV care before EAT, there was a significant increase in the proportion who were active in HIV care and virally suppressed at their last VL measure during EAT enrollment compared to before EAT enrollment. Participants who enrolled in HIV care concurrently with EAT had a significantly shorter time to initiation of ART and first episode of viral suppression compared to participants who enrolled in HIV care prior to EAT. The total cost of the intervention over 19 months was USD $57,448.64. Fixed costs were USD $3623.04 and variable costs were USD $63.75/month/participant. CONCLUSIONS: This pilot study provided proof of concept that EAT, a combination intervention providing mDOT, food, and peer navigation services, was feasible to implement and may support engagement in HIV care and achievement of viral suppression among SCI living with HIV in an urban setting in Kenya. Future work should focus on controlled trials of EAT, assessments of feasibility in other contexts, and cost-effectiveness studies.

Population-Based Estimates and Predictors of Child and Adolescent Linkage to HIV Care or Death in Western Kenya.

BACKGROUND: Population-level estimates of linkage to HIV care among children and adolescents (CAs) can facilitate progress toward 95-95-95 goals. SETTING: This study was conducted in Bunyala, Chulaimbo, and Teso North subcounties, Western Kenya. METHODS: Linkage to care was defined among CAs diagnosed with HIV through Academic Model Providing Access to Healthcare (AMPATH)'s home-based counseling and testing initiative (HBCT) by merging HBCT and AMPATH Medical Record System data. Using follow-up data from Bunyala, we examined factors associated with linkage or death, using weighted multinomial logistic regression to account for selection bias from double-sampled visits. Based on the estimated model, we imputed the trajectory for each person in 3 subcounties until a simulated linkage or death occurred or until the end of 8 years when an individual was simulated to be censored. RESULTS: Of 720 CAs in the analytic sample, 68% were between 0 and 9 years and 59% were female. Probability of linkage among CAs in the combined 3 subcounties was 48%-49% at 2 years and 64%-78% at 8 years while probability of death was 13% at 2 years and 19% at 8 years. Single or double orphanhood predicted linkage (adjusted odds ratio [aOR]: 2.66, 95% confidence interval [CI]: 1.33 to 5.32) and death (aOR: 9.85 [95% CI: 2.21 to 44.01]). Having a mother known to be HIV-positive also predicted linkage (aOR = 1.94, 95% CI: 0.97 to 3.86) and death (aOR: 14.49, 95% CI: 3.32 to 63.19). CONCLUSION: HIV testers/counselors should continue to ensure linkage among orphans and CAs with mothers known to be HIV-positive and also to support other CAs to link to HIV care.

Implementation outcomes and strategies for delivering evidence-based hypertension interventions in lower-middle-income countries: Evidence from a multi-country consortium for hypertension control.

Guidance on contextually tailored implementation strategies for the prevention, treatment, and control of hypertension is limited in lower-middle income countries (Lower-MIC). To address this limitation, we compiled implementation strategies and accompanying outcomes of evidence-based hypertension interventions currently being implemented in five Lower-MIC. The Global Research on Implementation and Translation Science (GRIT) Coordinating Center (CC) (GRIT-CC) engaged its global network sites at Ghana, Guatemala, India, Kenya, and Vietnam. Purposively sampled implementation science experts completed an electronic survey assessing implementation outcomes, in addition to implementation strategies used in their ongoing hypertension interventions from among 73 strategies within the Expert Recommendations for Implementing Change (ERIC). Experts rated the strategies based on highest priority to their interventions. We analyzed the data by sorting implementation strategies utilized by sites into one of the nine domains in ERIC and summarized the data using frequencies, proportions, and means. Seventeen implementation experts (52.9% men) participated in the exercise. Of Proctor's implementation outcomes identified across sites, all outcomes except for appropriateness were broadly assessed by three or more countries. Overall, 59 out of 73 (81%) strategies were being utilized in the five countries. The highest priority implementation strategies utilized across all five countries focused on evaluative and iterative strategies (e.g., identification of context specific barriers and facilitators) to delivery of patient- and community-level interventions, while the lowest priority was use of financial and infrastructure change strategies. More capacity building strategies (developing stakeholder interrelationships, training and educating stakeholders, and supporting clinicians) were incorporated into interventions implemented in India and Vietnam than Ghana, Kenya, and Guatemala. Although robust implementation strategies are being used in Lower -MICs, there is minimum use of financial and infrastructure change strategies. Our study contributes to the growing literature that demonstrates the use of Expert Recommendations for Implementing Change (ERIC) implementation strategies to deliver evidence-based hypertension interventions in Lower-MICs and will inform future cross-country data harmonization activities in resource-constrained settings.

Development, Assessment, and Outcomes of a Community-Based Model of Antiretroviral Care in Western Kenya Through a Cluster-Randomized Control Trial.

OBJECTIVE: To develop and assess an alternative care model using community-based groups for people living with HIV and facilitate by lay personnel. METHODS: Geographic locations in the Academic Model Providing Access to Healthcare Kitale clinic catchment were randomized to standard of care versus a community-based care group (ART Co-op). Adults stable on antiretroviral therapy and virally suppressed were eligible. Research Assistant-led ART Co-ops met in the community every 3 months. Participants were seen in the HIV clinic only if referred. CD4 count and viral load were measured in clinic at enrollment and after 12 months. Retention, viral suppression, and clinic utilization were compared between groups using χ2, Fisher exact, and Wilcoxon rank sum tests. RESULTS: At 12 months, there were no significant differences in mean CD4 count or viral load suppression. There was a significant difference in patient retention in assigned study group between the intervention and control group (81.6% vs 98.6%; P < 0.001), with a number of intervention patients withdrawing because of stigma, relocation, pregnancy, and work conflicts. All participants, however, were retained in an HIV care program for the study duration. The median number of clinic visits was lower for the intervention group than that for the control group (0 vs 3; P < 0.001). CONCLUSIONS: Individuals retained in a community-based HIV care model had clinical outcomes equivalent to those receiving clinic-based care. This innovative model of HIV care addresses the problems of insufficient health care personnel and patient retention barriers, including time, distance, and cost to attend clinic, and has the potential for wider implementation.

Population-based estimates of engagement in HIV care and mortality using double-sampling methods following home-based counseling and testing in western Kenya.

INTRODUCTION: Data on engagement in HIV care from population-based samples in sub-Saharan Africa are limited. The objective of this study was to use double-sampling methods to estimate linkage to HIV care, ART initiation, and mortality among all adults diagnosed with HIV by a comprehensive home-based counseling and testing (HBCT) program in western Kenya. METHODS: HBCT was conducted door-to-door from December 2009 to April 2011 in three sub-counties of western Kenya by AMPATH (Academic Model Providing Access to Healthcare). For those identified as HIV-positive, data were merged with electronic medical records to determine engagement with HIV care. A randomly-drawn follow-up sample of 120 adults identified via HBCT who had not linked to care as of June 2015 in Bunyala sub-county were visited by trained fieldworkers to ascertain HIV care engagement and vital status. Double-sampled data were used to generate, via multinomial regression, predicted probabilities of engagement in care and mortality among those whose status could not be ascertained by matching with the electronic medical records in the three catchments. RESULTS: Incorporating information from the double-sampling yielded estimates of prospective linkage to HIV care that ranged from 40-45%. Mortality estimates of those who did not engage in care following HBCT ranged from 12-16%. Among those who linked to care following HBCT, between 72-81% initiated ART. DISCUSSION: In settings without universal national identifiers, rates of linkage to care from community-based programs may be subject to substantial underestimation. Follow-up samples of those with missing information can be used to partially correct this bias, as has been demonstrated previously for mortality among those who were lost-to-care programs. There is a need for harmonized data systems across health systems and programs.

Evaluating strategies to improve HIV care outcomes in Kenya: a modelling study.

BACKGROUND: With expanded access to antiretroviral therapy (ART) in sub-Saharan Africa, HIV mortality has decreased, yet life-years are still lost to AIDS. Strengthening of treatment programmes is a priority. We examined the state of an HIV care programme in Kenya and assessed interventions to improve the impact of ART programmes on population health. METHODS: We created an individual-based mathematical model to describe the HIV epidemic and the experiences of care among adults infected with HIV in Kenya. We calibrated the model to a longitudinal dataset from the Academic Model Providing Access To Healthcare (known as AMPATH) programme describing the routes into care, losses from care, and clinical outcomes. We simulated the cost and effect of interventions at different stages of HIV care, including improvements to diagnosis, linkage to care, retention and adherence of ART, immediate ART eligibility, and a universal test-and-treat strategy. FINDINGS: We estimate that, of people dying from AIDS between 2010 and 2030, most will have initiated treatment (61%), but many will never have been diagnosed (25%) or will have been diagnosed but never started ART (14%). Many interventions targeting a single stage of the health-care cascade were likely to be cost-effective, but any individual intervention averted only a small percentage of deaths because the effect is attenuated by other weaknesses in care. However, a combination of five interventions (including improved linkage, point-of-care CD4 testing, voluntary counselling and testing with point-of-care CD4, and outreach to improve retention in pre-ART care and on-ART) would have a much larger impact, averting 1·10 million disability-adjusted life-years (DALYs) and 25% of expected new infections and would probably be cost-effective (US$571 per DALY averted). This strategy would improve health more efficiently than a universal test-and-treat intervention if there were no accompanying improvements to care ($1760 per DALY averted). INTERPRETATION: When resources are limited, combinations of interventions to improve care should be prioritised over high-cost strategies such as universal test-and-treat strategy, especially if this is not accompanied by improvements to the care cascade. International guidance on ART should reflect alternative routes to programme strengthening and encourage country programmes to evaluate the costs and population-health impact in addition to the clinical benefits of immediate initiation. FUNDING: Bill & Melinda Gates Foundation, United States Agency for International Development, National Institutes of Health.

Implementation and operational research: evaluating outcomes of patients lost to follow-up in a large comprehensive care treatment program in western Kenya.

BACKGROUND: Academic Model Providing Access To Healthcare (AMPATH) program provides comprehensive HIV care and treatment services. Approximately, 30% of patients have become lost to follow-up (LTFU). We sought to actively trace and identify outcomes for a sample of these patients. METHODS: LTFU was defined as missing a scheduled visit by ≥3 months. A randomly selected sample of 17% of patients identified as LTFU between January 2009 and June 2011 was generated, with sample stratification on age, antiretroviral therapy (ART) status at last visit, and facility. Chart reviews were conducted followed by active tracing. Tracing was completed by trained HIV-positive outreach workers July 2011 to February 2012. Outcomes were compared between adults and children and by ART status. RESULTS: Of 14,811 LTFU patients, 2540 were randomly selected for tracing (2179 adults, 1071 on ART). The chart reviews indicated that 326 (12.8%) patients were not actually LTFU. Outcomes for 71% of sampled patients were determined including 85% of those physically traced. Of those with known outcomes, 21% had died, whereas 29% had disengaged from care for various reasons. The remaining patients had moved away (n = 458, 25%) or were still receiving HIV care (n = 443 total, 25%). CONCLUSIONS: Our findings demonstrate the feasibility of a large-scale sampling-based approach. A significant proportion of patients were found not to be LTFU, and further, high numbers of patients who were LTFU could not be located. Over a quarter of patients disengaged from care for various reasons including access challenges and familial influences.

Linkage to and engagement in HIV care in western Kenya: an observational study using population-based estimates from home-based counselling and testing.

BACKGROUND: Few population-based studies exist on the HIV care continuum in sub-Saharan Africa. We aimed to describe engagement in care in all adults with an existing diagnosis of HIV and to assess the time to and predictors of linkage and engagement in adults newly diagnosed via home-based counselling and testing (HBCT) in a high-prevalence setting in western Kenya. METHODS: Data were derived from AMPATH (Academic Model Providing Access to Healthcare), which has provided HIV care in western Kenya since 2001 and the HBCT programme, which has been operating since 2007. After a widespread HBCT programme in Bunyala subcounty from December, 2009, to February, 2011, we reviewed electronic medical records to identify uptake of care in individuals (aged 13 years or older) with previously known (self-reported) infection and new (identified at HBCT) HIV diagnoses as of June 1, 2014. We defined engagement in HIV care as an initial encounter with an HIV care provider. We used Cox regression analysis to examine the predictors of engagement in care for newly diagnosed individuals. FINDINGS: Of the 3482 adults with HIV identified at HBCT, 2122 (61%) had previously been diagnosed with HIV, of whom 1778 (84%) had had at least one clinical encounter within AMPATH. 993 (73%) of the 1360 individuals with new diagnoses at HBCT were registered in the electronic medical records, although only 209 (15%) had seen a clinician over a median of 3·4 years since diagnosis. The median time to engagement in the newly diagnosed individuals was 60 days (IQR 10­411). INTERPRETATION: Creative and innovative strategies are needed to support people to engage with care when they are newly diagnosed with HIV through population-based case-finding initiatives. FUNDING: US President's Emergency Plan for AIDS Relief (PEPFAR), Abbott Laboratories, the Purpleville Foundation, the Global Business Coalition, the US National Institute of Mental Health, and the Bill & Melinda Gates Foundation.

Men who have sex with men in Kisumu, Kenya: comfort in accessing health services and willingness to participate in HIV prevention studies.

Men who have sex with men (MSM) are a crucial and marginalized at risk population for HIV in Africa but are poorly studied. Like other areas of Africa, homosexuality is illegal in Kenya. We assessed MSM comfort in accessing health services and willingness to participate in HIV prevention research in Kisumu, Kenya-an area of high HIV prevalence. We conducted a two-phase formative study with individual interviews (n = 15) and a structured survey (n = 51). Peer contact or snowball method (n = 43, 84.3%) was the primary recruitment strategy used to locate MSM. Exact logistic regression models were used for survey data analysis. Over 60% (32/51) of survey participants were not very comfortable seeking health services from a public hospital. Almost all MSM (49/51; 96.1%) reported willingness to be contacted to participate in future HIV research studies. Efforts to provide facilities that offer safe and confidential health services and health education for MSM is required. Continued community engagement with the MSM population in Kenya is needed to guide best practices for involving them in HIV prevention research.

Baseline findings of an HIV incidence cohort study to prepare for future HIV prevention clinical trials in Kisumu, Kenya.

INTRODUCTION: In an analysis of baseline findings of an HIV incidence cohort study, an assessment was made of HIV prevalence among persons presenting for enrollment and any differences in demographic characteristics between persons not enrolled compared to those enrolled. We also described and compared HIV risk behaviors in males and females enrolled in the study. METHODOLOGY: A computer-assisted survey was administered to collect baseline demographic and HIV risk data from 1,277 men and women aged 18-34 years. Testing for HIV and other sexually transmitted infections (STI) was conducted. Out of 1,277 persons prescreened for eligibility, 625 were enrolled. RESULTS: HIV prevalence of all persons who completed screening was 14.8% (females: 21.1%; males: 8.1%). The odds of being enrolled in the study were higher for persons 18-24 years compared to those 30-34 years of age [adjusted odds ratio (AOR)=2.18, CI=1.13, 4.21] and males compared to females [AOR=2.07, CI=1.43, 2.99]. Among those enrolled in the study, the most prevalent HIV risk behaviors were unprotected sex (49%), alcohol use (45%), and transactional sex (30%) in the last three months. Compared to females, a significantly greater proportion of males reported using any alcohol or recreational drug in the last three months, a history of oral sex, sex with partner other than a spouse or main partner, ever having a blood transfusion, ever being treated for an STI, and having knowledge of their last HIV test result. CONCLUSION: The Kisumu Field Station successfully recruited individuals with HIV risk characteristics for the HIV incidence cohort study.

Baseline Findings of an HIV Incidence Cohort Study to Prepare for Future HIV Prevention Clinical Trials in Kisumu; Kenya

Introduction: In an analysis of baseline findings of an HIV incidence cohort study; an assessment was made of HIV prevalence among persons presenting for enrollment and any differences in demographic characteristics between persons not enrolled compared to those enrolled. We also described and compared HIV risk behaviors in males and females enrolled in the study. Methodology: A computer-assisted survey was administered to collect baseline demographic and HIV risk data from 1;277 men and women aged 18-34 years. Testing for HIV and other sexually transmitted infections (STI) was conducted. Out of 1;277 persons prescreened for eligibility; 625 were enrolled. Results: HIV prevalence of all persons who completed screening was 14.8 (females: 21.1; males: 8.1). The odds of being enrolled in the study were higher for persons 18-24 years compared to those 30-34 years of age [adjusted odds ratio (AOR)=2.18; CI=1.13; 4.21] and males compared to females [AOR=2.07; CI=1.43; 2.99]. Among those enrolled in the study; the most prevalent HIV risk behaviors were unprotected sex (49); alcohol use (45); and transactional sex (30) in the last three months. Compared to females; a significantly greater proportion of males reported using any alcohol or recreational drug in the last three months; a history of oral sex; sex with partner other than a spouse or main partner; ever having a blood transfusion; ever being treated for an STI; and having knowledge of their last HIV test result. Conclusion: The Kisumu Field Station successfully recruited individuals with HIV risk characteristics for the HIV incidence cohort study