Speaker
Description
Background: Human Immunodeficiency Virus (HIV) remains a major public health problem. In Ethiopia, particularly the Chagni Town is a development corridor area with main road to the Ethiopian great renaissance dam rout and with more than two years of ongoing conflict may aggravate HIV burden. However, predicting model on death among HIV adult patients is yet not documented.
Objective: To develop and validate prediction model of mortality among adults living with HIV/AIDS in Chagni health enter in Chagni Town, Awi Zone, Amhara Region, Ethiopia.
Methods: A retrospective follow-up study was conducted among 337 HIV infected adults on ART in Chagni health center. Data were accessed from February 28, 2025 to March 19, 2025 from patient records. STATA Version 17 was used for statistical analysis. Discriminative ability and predictive accuracy were determined using area under receiver operating characteristic and model calibration respectively. Youden index value was used to fix the threshold for risk classification. Internal validation of the model was evaluated by bootstrapping technique.
Results: Participants had a mean age of 40.32 years (SD ±11.13); 61.4% were females. Mortality incidence was 1.98 (95% CI: 1.47 - 2.65) per 100-persons years of observation. The determinants were having no history of treatment regimen change (AOR: 5.60; 95% CI: 2.14-14.62), CD4 count below 450cells/ul (AOR: 5.17; 95% CI: 1.88-14.24), no tuberculosis preventive therapy (TPT) (AOR: 14.44; 95% CI: 5.69-36.64), and underweight (AOR: 2.87; 95% CI: 1.20-6.86). The optimal risk threshold was 0.0860; yielding 90.9% sensitivity and 72.7% specificity. Using this cut point, 120(35.6%) patients were classified high risk, among whom 40(33.3%) died. The discriminative probability of the model (AUROC) was 0.92 (95% CI: 0.88- 0.95) and showed good calibration (Hosmer-Lemeshow: p value = 0.3299). After bootstrapping, AUROC was 89.9% (95% CI: 84.7% - 93.7%).
Conclusions: The incidence of mortality was high. The model had strong discrimination and calibration to determine risk of mortality. Lack of TB preventive therapy, absence of regimen change, having low CD4 count and underweight were determinants. External validation needs to be carried out before using the developed model.
