Antihypertensive Pharmacotherapy Gaps in Nigeria: A Predictive Machine Learning Analysis of Treatment Uptake Amid Macroeconomic Shock, Using NDHS 2023-24

BackgroundHypertension is the most important modifiable cardiovascular risk factor worldwide, and the prevalence is increasing in sub-Saharan Africa. Nigerias newly released 2023-24 Demographic and Health Survey (NDHS) provides the first opportunity to explore national-level non-uptake of antihypertensive treatment using a machine-learning cascade framework, although survey fieldwork was conducted amid the unprecedented shock of fuel subsidy removal and currency devaluation in Nigeria. ObjectivesTo identify correlates of antihypertensive treatment non-uptake (Gap 2) among the diagnosed adults in Nigeria; test if the non-uptake varied based on when the survey was conducted during this macroeconomic shock; and compare four predictive algorithms. MethodsWe used the 2023-24 Nigeria Demographic and Health Survey (NDHS) data (2,975 women diagnosed with hypertension and 529 men diagnosed with hypertension aged 15-49) to fit survey-weighted logistic regression models separately by sex and evaluated sex-by-predictor interactions in a single pooled model. We then compared the survey-weighted logistic regression model with the elastic net, random forest, and gradient boosting (XGBoost) models, all properly weighted, using held-out test-set AUC. ResultsThe timing of fieldwork interviews was a significant risk factor for women (OR=1.14 per month; 95% CI: 1.06-1.22; p<0.001) but not for men, and this sex difference was confirmed in the formal interaction test (p=0.016). This study tested the interaction of sex with each geopolitical zone, with the strongest result being a sex-reversed pattern in zones protective for women, showing these zones were significant risk factors for men (ORs 2.28-5.77); interaction testing (all p[&le;]0.007) confirmed this. No interaction was observed between diabetes and sex (p=0.170); instead, diabetes was protective for women (OR=0.34, p<0.001). For pooled samples, wealth was associated with non-uptake (OR=0.59; p=0.012), and this appeared significant only for women (OR=0.62; p=0.026), in keeping with the richest-vs-poorest estimate in Table 2. All three ML algorithms had similar AUC for women (0.61-0.63) and lower, near-chance performance for men (0.56-0.59), reflecting the much smaller number of diagnosed men. O_TBL View this table: org.highwire.dtl.DTLVardef@1c8e097org.highwire.dtl.DTLVardef@1110caforg.highwire.dtl.DTLVardef@57c204org.highwire.dtl.DTLVardef@182ac61org.highwire.dtl.DTLVardef@31613_HPS_FORMAT_FIGEXP M_TBL O_FLOATNOTable 2.C_FLOATNO O_TABLECAPTIONSurvey-weighted logistic regression, women, main model (n=2,975) C_TABLECAPTION C_TBL ConclusionsThere is a large disparity in the rate of treatment and management of HTN across sexes, geopolitical zones, and macro-economic environments, making a single uniform national intervention model impractical. In contrast, an earlier study in Africa reported that similar algorithmic complexity was useful in predicting cardiovascular risk; this did not happen in the current study. The proposed programs are to build programs by sex and zone, integrate hypertension screenings into chronic disease and/or maternal health trigger points, and better understand the resilience of the pharmacy supply chain.