INTRODUCTION: Optimal use of HIV testing resources accelerates progress towards ending HIV as a global threat. In Kenya, current testing practices yield a 2.8% positivity rate for new diagnoses reported through the national HIV electronic medical record (EMR) system. Increasingly, researchers have explored the potential for machine learning to improve the identification of people with undiagnosed HIV for referral for HIV testing. However, few studies have used routinely collected programme data as the basis for implementing a real-time clinical decision support system to improve HIV screening. In this study, we applied machine learning to routine programme data from Kenya's EMR to predict the probability that an individual seeking care is undiagnosed HIV positive and should be prioritized for testing. METHODS: We combined de-identified individual-level EMR data from 167,509 individuals without a previous HIV diagnosis who were tested between June and November 2022. We included demographics, clinical histories and HIV-relevant behavioural practices with open-source data that describes population-level behavioural practices as other variables in the model. We used multiple imputations to address high rates of missing data, selecting the optimal technique based on out-of-sample error. We generated a stratified 60-20-20 train-validate-test split to assess model generalizability. We trained four machine learning algorithms including logistic regression, Random Forest, AdaBoost and XGBoost. Models were evaluated using Area Under the Precision-Recall Curve (AUCPR), a metric that is well-suited to cases of class imbalance such as this, in which there are far more negative test results than positive. RESULTS: All model types demonstrated predictive performance on the test set with AUCPR that exceeded the current positivity rate. XGBoost generated the greatest AUCPR, 10.5 times greater than the rate of positive test results. CONCLUSIONS: Our study demonstrated that machine learning applied to routine HIV testing data may be used as a clinical decision support tool to refer persons for HIV testing. The resulting model could be integrated in the screening form of an EMR and used as a real-time decision support tool to inform testing decisions. Although issues of data quality and missing data remained, these challenges could be addressed using sound data preparation techniques.

Early combination antiretroviral therapy (cART), as recommended in WHO's universal test-and-treat (UTT) policy, is associated with improved linkage to care, retention, and virologic suppression in controlled studies. We aimed to describe UTT uptake and effect on twelve-month non-retention and initial virologic non-suppression (VnS) among HIV infected adults starting cART in routine HIV program in Kenya. Individual-level HIV service delivery data from 38 health facilities, each representing 38 of the 47 counties in Kenya were analysed. Adults (>15 years) initiating cART between the second-half of 2015 (2015HY2) and the first-half of 2018 (2018HY1) were followed up for twelve months. UTT was defined based on time from an HIV diagnosis to cART initiation and was categorized as same-day, 1-14 days, 15-90 days, and 91+ days. Non-retention was defined as individuals lost-to-follow-up or reported dead by the end of the follow up period. Initial VnS was defined based on the first available viral load test with >400 copies/ml. Hierarchical mixed-effects survival and generalised linear regression models were used to assess the effect of UTT on non-retention and VnS, respectively. Of 8592 individuals analysed, majority (n = 5864 [68.2%]) were female. Same-day HIV diagnosis and cART initiation increased from 15.3% (2015HY2) to 52.2% (2018HY1). The overall non-retention rate was 2.8 (95% CI: 2.6-2.9) per 100 person-months. When compared to individuals initiated cART 91+ days after a HIV diagnosis, those initiated cART on the same day of a HIV diagnosis had the highest rate of non-retention (same-day vs. 91+ days; aHR, 1.7 [95% CI: 1.5-2.0], p<0.001). Of those included in the analysis, 5986 (69.6%) had a first viral load test done at a median of 6.3 (IQR, 5.6-7.6) months after cART initiation. Of these, 835 (13.9%) had VnS. There was no association between UTT and VnS (same-day vs. 91+ days; aRR, 1.0 [95% CI: 0.9-1.2], p = 0.664). Our findings demonstrate substantial uptake of the UTT policy but poor twelve-month retention and lack of an association with initial VnS from routine HIV settings in Kenya. These findings warrant consideration for multi-pronged program interventions alongside UTT policy for maximum intended benefits in Kenya.

SETTING: Drug susceptibility testing (DST) is recommended in Kenya to identify multidrug-resistant tuberculosis (MDR-TB) in persons registered for tuberculosis (TB) retreatment. DST is performed at a central laboratory with a two-step growth-based process and a regional laboratory with a simultaneous molecular- and growth-based process. OBJECTIVE: To compare proportions of retreatment cases who underwent DST and turnaround times for hospitals referring to the central vs. regional laboratory. DESIGN: Cases were persons registered for TB retreatment from 1 January 2012 to 31 December 2013. Records of 11 hospitals and 7 hospitals referring patients to the regional and central laboratories, respectively, were reviewed. RESULTS: Respectively 238/432 (55%) and 88/355 (25%) cases from hospitals referring to the regional and central laboratories underwent DST. The mean time from case registration to receipt of DST results and initiation of MDR-TB treatment was quicker in hospitals referring to the regional laboratory. The time required for the transportation of specimens, specimen testing and receipt of DST results at hospitals was shorter for the regional laboratory (P < 0.05). CONCLUSION: Testing was faster and more complete at hospitals referring to the regional laboratory. A comprehensive review of MDR-TB detection in Kenya is required to increase the proportion of cases receiving DST.