| dc.description.abstract | As the number of people living with HIV in the population who do not know their HIV status continues to decline, as more people are linked to ART, continuing to offer HIV testing in a universal manner becomes inefficient. Finding ways to target HIV testing to persons more likely to be HIV positive, for efficiency, is a global priority. The purpose of this study was to evaluate the useof three strategies to identify sub-populations and granular-geographic areas with higher HIV positive yield to inform efficient targeting of HIV testing among persons >15 years in Homa Bay, Siaya and Kisumu; counties with the highest HIV prevalence and incidence in Kenya. The specific objectives were to evaluate the use of a HIV predictive risk-score algorithm, geospatial analysis of new HIV diagnoses, and mapping of HIV testing uptake. Using a hospital-based retrospective cohort study design, aHIV predictive risk-score screening algorithm was developed using univariable and multivariable analyses of outpatient data, comprising 19,458 persons >15 years tested for HIV from September 2017–May 2018 from five purposively selected health facilities in Homa Bay, Siaya and Kisumu Counties. Using a community-based retrospective cohort study design, the use of geospatial analysis to assess geospatial patterns of new HIV diagnoses, and the use of mapping HIV testing uptake, were evaluated. Community home-based data comprised 365,798 clients aged >15 years offered home-based HIV testing as part of a routine public health program from May 2016–July 2017 in Siaya County. Geospatial analysis using Kulldorff’s spatial scan statistic was used to detect geographic clusters (radius <5 kilometers) of new HIV diagnoses. A Geographical Information System program was used to map HIV testing uptake. The results showed that an HIV predictive risk-score screening algorithm developed grouped patients into four risk-score categories: <9, 10–15, 16–29 and >30, with increasing HIV prevalence of 0.6% [95% Confidence Interval (CI): 0.46–0.75], 1.35% (95% CI: 0.85–1.84), 2.65% (95% CI: 1.8–3.51), and 15.15% (95%CI: 9.03–21.27), respectively. External validation of the algorithm produced similar results. The algorithm’s discrimination performance was modest, with an area under the receiver-operating-curve of 0.69 (95% CI: 0.53–0.84). The algorithm accounted for a high proportion (R2 0.89) of the variability of HIV prevalence in the study population. Results from geospatial analysis of new HIV diagnoses showed spatial variation in the distribution of new HIV diagnoses, and nine sub-location clusters in which the number of new HIV diagnoses was significantly (1.56 to 2.64 times) higher than expected were identified. Results from mapping HIV testing uptake found that268,543 (86%) clients were tested for HIV. Of the 43,680 eligible clients not tested, 32,852 (75%) were not found at home and 5,931 (14%) declined testing. Granular geographic areas with low testing uptake, a high proportion of clients not found at home and a high proportion who declined testing, yet with clusters of higher new HIV diagnoses were identified. In conclusion, the following strategies successfully identified sub-populations and granular-geographic areas with higher HIV positive yield that should be targeted in the implementation of HIV testing services: aHIV predictive risk-score screening algorithm that identified patients who are more likely to be HIV-positive; geospatial analysis that identified granular sub-location clusters (<5 kilometers) of higher new HIV diagnoses; and mapping of HIV testing uptake that identified granular-geographic areas with low HIV testing uptake yet higher HIV positive yield. These study findings inform global, national, and county government policies and strategies for targeting HIV testing, for efficient use of resources and maximal epidemiologic impact. | en_US |
