Forecasting climate-associated non-tuberculous mycobacteria (NTM) infections in the UK using international surveillance data and machine learning

Abstract

Nontuberculous mycobacteria (NTM) cause skin and lung infections, have high mortality rates, and are resistant to a range of antibiotics and water treatment methods. As NTM reside in environmental reservoirs, they are sensitive to environmental conditions. The suitability of their environmental reservoirs can increase as a result of climate change, subsequently increasing environmental exposure and infection rates. NTM infections are not generally notifiable, including in the UK, but sustained increases have been observed in regions that report NTM infection rates. To assess the burden of NTM infections in the UK under projected climate change, we examined the relationship between climate variables and available NTM surveillance data internationally. Statistically significant increases were found in regions where NTM infections are notifiable, which were positively associated with increased precipitation and temperatures. A random forest regressor was trained using supervised learning from international NTM surveillance data and linked climate variables. The random forest model was applied to UK climate projections, projecting a 6.2% increase in NTM infection rates over the next 10 years, with notable regional variation. Our random forest model predicts that the forecasted impacts of climate change in the UK, including increasing temperatures and frequency of heavy rainfall, will lead to increases in NTM infection rates. Robust surveillance in the future is necessary to increase data available to train models, increasing our predictive power in forecasting climate-associated NTM trends. Our results highlight a novel aspect of how climate change will impact health outcomes in the UK.