INTRODUCTION:
Air pollution affects many health outcomes, but most studies focus on single pollutants, with few investigating the combined effects of multiple pollutants 1-2. Predicting at-risk individuals is particularly challenging in low and middle-income settings with complex exposure–health relationships3. This study investigated the joint effects of air pollutants on cardiovascular and respiratory hospital admissions in Cape Town from January 2011 to October 2016

METHODS:
Respiratory and cardiovascular hospital admission data were obtained from a private hospital group, and PM10, NO2, SO2, and weather data from the Department of of Forestry, Fisheries and the Environment and South African Weather Services. Associations between daily air pollution and hospital admissions were assessed using classification and regression trees and time-series quasi-Poisson regression models, with results reported as rate ratios and 95% confidence intervals (CI).

RESULTS:
Respiratory (n = 47 611) and cardiovascular (n = 45 021) admissions on 1,766 days with complete pollutant data were included. Among respiratory admissions, 49%, 33%, and 18% were aged 0–14, 15–64, and ≥65 years; for cardiovascular admissions, 47% and 46% were 15–64 and ≥65 years. Using pollutant quartiles, 64 unique air pollution mixtures were possible. Classification and regression tree and time-series quasi-Poisson models showed significant joint effects of PM10, NO2, and SO2, with rate ratios of 1.3 (95% CI: 1.3–1.5) for respiratory and 1.2 (95% CI: 1.1–1.4) for cardiovascular admissions.

CONCLUSIONS:
Significant joint effects of PM10, NO2, and SO2 were observed on cardiovascular and respiratory hospital admissions. Building on these findings, the ongoing PhD extends the framework by applying and comparing multiple machine learning algorithms, including advanced ensemble approaches and neural networks, using integrated environmental and health data to predict respiratory symptoms and hospitalisation risk.