A Tool For Predicting Loss-to-follow-up Among People Living With Hiv At Busia Border

Abstract

Human Immuno-Deficiency Virus (HIV) and Acquired Immune Deficiency Syndrome (AIDS) are a global emergency. Infection with HIV can lead to poor health, loss of lives, increased number of orphans and reduced national productivity. In Kenya and Uganda, National AIDS & STI Control Programme (NASCOP) and the Uganda AIDS Commission (UAC) respectively are promoting combination of approaches for HIV prevention with the key populations. Studies have revealed that 1.5 million people live with HIV with a 5.9% adult HIV prevalence. There are an estimated 78,000 new HIV infections with 36,000 aids related deaths and 59% adults on antiretroviral treatment (AVERT 2016). Complex socio-cultural, economic, and health-system factors inhibit excellent patient retention. Better tracking, enhanced social support, and regular adherence counselling in addressing stigma, and alternative healing options are needed. Intervention strategies aimed at changing clinic routines and improving patient–provider communication could address many of the identified barriers (Tiruneh et al. 2016). The objective of the study is to develop a tool to predict possible loss-to-follow-up among mobile people living with HIV/AIDS enrolled in care and treatment at the Busia border. Adherence to ARV drugs and retention in care and treatment programs of a sample of PLHIVs in the Northern Transport Corridor of Kenya and Uganda particularly in Busia cross-border site was assessed to determine the factors leading to loss-to-follow-up among them. A design science methodology was adopted in the design, development, testing, implementation and validation of the data mining and analytics tool for predicting possible loss-to-follow-up among people living with HIV. The tool was piloted at two cross-border participating facilities close to the border crossing point. Missing variables speaking to nationality and cross-border and crosscounty/district mobility characteristics were collected through a community and facility profile form. Exiting data from daily activity registers were filtered based on main dependent LTFU predictor variables resulting from reviewed literature on factors, used to inform design of the tool and as the training dataset. New PLHIV were enrolled and their data run through the decision tree predictor for results. Metrics were used to assess the differences in tracking and management of PLHIVs before and after implementation of the tool. Model evaluation metrics was used to test the accuracy, efficacy and utility of the tool. Study findings inform future effective interventions on tracing and linking back mobile PLHIVs displaying transnational service access characteristics to adhere and be retained in care and treatment as an effort to achieving the global UNAIDS 90-90-90 targets.