New finance-based model predicts death risk in tuberculosis patients

Tuberculosis (TB) remains a major global killer, especially when combined with poor nutrition or HIV. Traditional survival studies point to risk factors but do not explain the dynamic collapse that leads to early death. To tackle that gap, E.W. Pefura-Yone and co-authors adapted a framework from quantitative finance—the Merton jump-diffusion model—to think of survival as a form of biological solvency. In this view, a patient’s physiological resources can be tracked over time, and death occurs when a stochastic health trajectory crosses a critical failure line. The team applied this idea to a large retrospective cohort of 15,182 TB patients in Cameroon followed over two decades. The patients were largely young adults (median age 33) with a median body mass index (BMI) of 20.7 kg/m2, and 35% were HIV co-infected. By treating adjusted BMI as a proxy for health capital and modeling both gradual trends and sudden shocks to that capital, the researchers aimed to move beyond simple associations and toward a mechanistic picture of who is likely to survive and who is at imminent risk.

The researchers modeled adjusted BMI as a stochastic process that includes individual recovery trends, physiological instability, and sudden clinical shocks—features captured by the Merton jump-diffusion framework. Over a 240-day follow-up, the overall mortality was 7.0%, with 55.1% of deaths occurring within the first 30 days. The analysis identified a critical failure threshold at BMI 17.329 kg/m2. HIV co-infection stood out as a major driver of metabolic instability, significantly increasing physiological volatility and the likelihood of abrupt declines. Statistical checks showed that sudden clinical shocks were necessary to reproduce the observed pattern of early deaths. From this model the team derived a Distance-to-Death (DtD) metric to quantify how close an individual is to the failure threshold. DtD slightly outperformed standard associative extended Cox models in predicting survival (Harrell’s C-index: 0.781 vs. 0.772; p = 0.038). Patients in the highest-risk group had a 16.7% mortality rate versus 1.6% in the most stable group. To support application, the authors developed an interactive digital triage tool.

This work bridges mathematical finance and clinical epidemiology, offering a new, mechanistic way to understand vulnerability in TB. By treating BMI as a measurable reserve and accounting for both slow decline and sudden shocks, the model identifies a concrete BMI threshold and a quantitative Distance-to-Death signal that can flag patients at very high short-term risk. The finding that HIV raises physiological volatility helps explain why co-infected patients are more likely to die soon after diagnosis. Importantly, the approach moves beyond simple risk-factor lists toward a dynamic view of patient trajectories, which is especially useful in resource-limited settings where quick, actionable triage is essential. The interactive digital triage tool derived from the model is intended to help clinicians prioritize life-saving interventions and closer monitoring for those closest to the failure threshold. While the study is retrospective, it demonstrates that finance-derived models can yield clinically relevant metrics and support targeted care in populations with heavy burdens of malnutrition and HIV.