valentineghanem-bit/sti-hiv-syndemic-ghana-261districts

STI and HIV syndemic burden: geospatial co-clustering and behavioural determinants in Ghana - 260 health districts | 2022 # STI and HIV Syndemic Burden — Geospatial Co-clustering and Behavioural Determinants in Ghana **Author:** Valentine Golden Ghanem | Ghana COCOBOD Cocoa Clinic, Accra, Ghana **ORCID:** 0009-0002-8332-0220 **Affiliation:** Ghana COCOBOD Cocoa Clinic, Accra, Ghana **Reporting standard:** STROBE **Date:** May 2026 **Status:** Manuscript in preparation --- ## 1. Abstract This study quantifies the syndemic co-burden of sexually transmitted infections (STIs) and HIV across Ghana's 261 districts using geospatial co-clustering and machine learning. A **Syndemic Burden Index (SBI)** was constructed to identify districts where STI and HIV burdens spatially overlap. XGBoost with SHAP interpretability identifies the leading behavioural and structural determinants of co-burden. Bivariate LISA reveals strong spatial co-clustering (Moran's I = 0.564), with 54 confirmed High-High co-burden districts. The analysis integrates Ghana DHS, WHO GHO, and Ghana Statistical Service data to produce district-level syndemic risk maps for programmatic targeting. --- ## 2. Research Question & Aims - **Primary:** Quantify spatial co-clustering of STI and HIV burden across Ghana's 261 districts. - **Secondary:** (a) Construct and validate the Syndemic Burden Index (SBI); (b) identify behavioural and structural determinants using XGBoost + SHAP; (c) typologise districts via K-means clustering; (d) estimate spatial-lag regression to quantify associations corrected for spatial dependence. --- ## 3. Methods Summary | Method | Tool | Purpose | |--------|------|---------| | Small-area estimation | Custom | District-level STI/HIV burden interpolation | | Syndemic Burden Index (SBI) | Custom | Composite co-burden scoring | | Global Moran's I | esda / libpysal | Spatial autocorrelation (HIV, STI separately) | | Bivariate LISA | esda | HIV × STI co-clustering | | K-means (k=4) | scikit-learn | District risk typology | | XGBoost + SHAP | xgboost / shap | Risk prediction and driver identification …