A Survival Analysis Approach to Determine the Predictors of Time to Hiv Partner Notification in Kenya.

Abstract

Introduction: Globally, Sub-Saharan Africa (SSA) has registered the highest Human Immunodeficiency Virus (HIV) prevalence with over 50% of seropositive individuals unaware of their HIV status. Assisted Partner Services (APS) is one of safe and cost-effective HIV testing and awareness strategy that has documented increase in uptake of HIV testing by 1.5 times. However, APS is an intense process that requires significant amount of time and many factors contribute to the amount of time spent on this process. Knowledge on time to Partner Notification (PN) and possible predictors is critical in developing individual specific strategies that could ease and quicken notification process and result in timely case finding and initiation to Antiretroviral Therapy (ART) and eventually reduction in onward transmission and HIV prevalence. APS Studies conducted in SSA have not provided a lot of information on time to PN. We therefore carried out this study to determine the possible predictors of time to partner notification among the Kenyan population. Methodology: Secondary data from 1,119 HIV positive adults sampled from a cluster randomized control trial conducted in Kenya from August 2013 to August 2015 was used. The primary study randomized 9 clusters to immediate arm (that implemented APS immediately a partner was named) and the other 9 clusters to delayed arm (that implemented APS 6 weeks after a partner was named) and the outcome of measure was number of partners of an index tested for HIV, identified as HIV infected and linked to HIV care. Data analysis for this study was done using STATA version 14.2. Descriptive analysis was used to report socio-demographic characteristics. Kaplan –Meier (K-M) estimates was used to estimate time to HIV partner notification between groups and the equality of survival functions between groups tested using Wilcoxon test. A shared frailty multiple cox regression model was fitted to determine the possible predictors of time to HIV partner notification. Time varying effects cox regression model was fitted to address the violation of Proportional Hazard (PH) assumption by three variables. Results: Majority of index participants were females (61%) while males were most of the partners notified of exposure to HIV (56%). Index participants were younger (aged 30 years (IQR 25-38)) than their partners aged 31 years (IQR 26-38). There was statistically significant difference in the survival curves between immediate and delayed arms (p<0.001) at 5% significance level. Time to HIV partner notification was statistically associated to intervention arm (method of notification), sex of the partner and sex of the index at 5% significance level. The effects of intervention arm, sex of the partner and sex of the index on time to HIV partner notification varied with time. The intervention arm resulted in an increase in the rate of HIV partner notification at the beginning (HR=23.7) then the effects drops off with time. Conclusion: Sex of partner and index are important predictors of time to HIV partner notification. Effective time to conduct partner notification is within 42 days of being named. Health care provider characteristics could be obtained in future studies because they might have an effect on time to Partner notification.