Development and validation of a machine learning based early warning scoring system for high altitude polycythemia.

BACKGROUND

High-altitude polycythemia (HAPC) lacks a lifestyle-focused risk-stratification tool among lifelong high-altitude residents. Here we aimed to develop and validate a novel machine-learning predictive scoring system for HAPC using readily modifiable lifestyle variables in this population.

METHODS

In a high altitude cohort (≥4,500 m, n = 1,089), 82 candidate variables were reduced to seven lifestyle predictors via LASSO, Logistic regression, XGBoost and random forest models were trained and compared (10 fold cross validation).

RESULTS

Logistic regression achieved the best balance (AUC 0.848, sensitivity 0.81, specificity 0.79). Low SpO2 (< 83%), male sex, age ≥50 year, smoking, hypertension, higher body mass index (BMI) and lower tea consumption were independent predictors.

CONCLUSION

This score equips frontline health workers in extremely high-altitude, resource-scarce settings to rapidly pinpoint high-risk residents and initiate low-cost lifestyle interventions, thereby curbing the incidence of chronic altitude-related illnesses, easing local medical burdens, and improving overall quality of life for native high-altitude populations.

TRIAL REGISTRATION: ChiCTR2100047945.