An interpretable Day-1 machine-learning model for COVID-19 mortality and severity prognosis: development in a Pakistani hospital cohort and external evaluation in Chinese and Italian populations
- 1Lahore University of Management Sciences
Abstract
Early risk stratification of COVID-19 patients at admission informs triage and resource allocation, particularly in low- and middle-income settings where intensive-care capacity is constrained. Most published models are not interpretable, rarely externally validated, and developed almost exclusively in high-income cohorts. We developed an interpretable random-forest classifier for three-class COVID-19 severity (Mild: not ventilated, survived; Severe: ventilated, survived; Fatal: died) using only data available within 24 hours of admission in 321 hospitalised PCR-confirmed COVID-19 patients in Pakistan. From 2,278 raw variables we retained 20 predictors by clinical screening and permutation-importance selection, with leakage-controlled multiple imputation by chained equations and SHapley Additive exPlanations (SHAP) for interpretation. A 15-feature variant was evaluated in two independent public cohorts: the Chinese iCTCF cohort (n = 894) and an Italian complete-blood-count cohort (n = 1,218). The development model achieved a macro F1 of 0.55, identifying the Fatal class most reliably (F1 0.67, recall 0.73) and Mild cases least well. SHAP identified a parsimonious set of admission-time predictors (oxygen saturation, lactate dehydrogenase, urea, C-reactive protein, age and radiographic pneumonia), including a raised urea-to-creatinine ratio in fatal cases most consistent with dehydration (median urea 80 vs 50 mg/dL, fatal vs mild; p < 0.001). Externally, the model retained useful discrimination in the Chinese cohort (mortality AUC 0.82 in 719 patients with recorded outcomes; severe-or-worse AUC 0.75; macro F1 0.50) but only modest discrimination in the Italian cohort (AUC 0.62, 95% CI 0.56-0.68), which we treat as a feature-availability stress test rather than a second validation. Cross-cohort SHAP rankings correlated strongly in China (Spearman {rho} = 0.91) and moderately in Italy ({rho} = 0.76), where most laboratory predictors were imputed and carried no information beyond the complete blood count; discrimination tracked how many predictors were actually measured. The admission-time signature converges with established markers of hypoperfusion and inflammation.
Author summaryWhen patients arrive at hospital with COVID-19, risk must be assessed immediately to direct limited resources to those in need, particularly where intensive-care capacity is scarce, as in Pakistan. Most existing computer-based tools come from developed countries, act as "black boxes", and are rarely tested elsewhere. We built a model using first-day admission data from 321 patients at a hospital in Lahore, Pakistan. It identifies patients who later died best, and milder cases least well. The strongest markers were blood oxygen, lactate dehydrogenase, urea, C-reactive protein, age and chest-imaging evidence of pneumonia, together pointing to a coherent picture of pulmonary impairment, inflammation and circulatory stress. In fatal cases urea was markedly raised while creatinine rose only modestly, a pattern that most likely reflects dehydration. We then tested the same model on independent Chinese and Italian cohorts: it transferred well to China but lost accuracy in Italy, where only a basic blood count was available. The findings support an interpretable Day-1 prognostic tool whose useful range is set largely by which laboratory tests are available rather than by geography. Because these markers reflect general inflammation rather than anything unique to COVID-19, the approach may extend to other pneumonias.
Related
- Development and validation of an interpretable machine learning model for early prediction of mortality in ICU patients with chronic obstructive pulmonary disease and heart failure
- Development and validation of an interpretable machine learning model for predicting in-hospital mortality among critically ill patients with liver cirrhosis
- Development and external validation of interpretable machine learning models for 7-, 14-, and 28-day mortality in sepsis-associated liver injury
- An interpretable machine learning model for predicting intensive care unit mortality risk in elderly patients with concomitant cardiogenic shock and sepsis: Development and external validation using two large critical care databases
- Development and external evaluation of an interpretable machine-learning model for early prediction of organ failure in higher-risk acute pancreatitis patients: A multicentre cohort study
BibTeX
@misc{jaffer2026interpretable,
title = {{An interpretable Day-1 machine-learning model for COVID-19 mortality and severity prognosis: development in a Pakistani hospital cohort and external evaluation in Chinese and Italian populations}},
author = {T. Jaffer and S. Aslam and N. Javaid and A. Khaliq and E. Tariq and T. Mustafa and M. A. Zaffar and S. Mirza},
howpublished = {medRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.01.26364553},
url = {https://www.medrxiv.org/content/10.64898/2026.10.01.26364553v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - An interpretable Day-1 machine-learning model for COVID-19 mortality and severity prognosis: development in a Pakistani hospital cohort and external evaluation in Chinese and Italian populations
AU - T. Jaffer
AU - S. Aslam
AU - N. Javaid
AU - A. Khaliq
AU - E. Tariq
AU - T. Mustafa
AU - M. A. Zaffar
AU - S. Mirza
PY - 2026
DA - 2026/10/01
PB - medRxiv
DO - 10.64898/2026.10.01.26364553
UR - https://www.medrxiv.org/content/10.64898/2026.10.01.26364553v1
AB - Early risk stratification of COVID-19 patients at admission informs triage and resource allocation, particularly in low- and middle-income settings where intensive-care capacity is constrained. Most published models are not interpretable, rarely externally validated, and developed almost exclusively in high-income cohorts. We developed an interpretable random-forest classifier for three-class COVID-19 severity (Mild: not ventilated, survived; Severe: ventilated, survived; Fatal: died) using only data available within 24 hours of admission in 321 hospitalised PCR-confirmed COVID-19 patients in Pakistan. From 2,278 raw variables we retained 20 predictors by clinical screening and permutation-importance selection, with leakage-controlled multiple imputation by chained equations and SHapley Additive exPlanations (SHAP) for interpretation. A 15-feature variant was evaluated in two independent public cohorts: the Chinese iCTCF cohort (n = 894) and an Italian complete-blood-count cohort (n = 1,218). The development model achieved a macro F1 of 0.55, identifying the Fatal class most reliably (F1 0.67, recall 0.73) and Mild cases least well. SHAP identified a parsimonious set of admission-time predictors (oxygen saturation, lactate dehydrogenase, urea, C-reactive protein, age and radiographic pneumonia), including a raised urea-to-creatinine ratio in fatal cases most consistent with dehydration (median urea 80 vs 50 mg/dL, fatal vs mild; p < 0.001). Externally, the model retained useful discrimination in the Chinese cohort (mortality AUC 0.82 in 719 patients with recorded outcomes; severe-or-worse AUC 0.75; macro F1 0.50) but only modest discrimination in the Italian cohort (AUC 0.62, 95% CI 0.56-0.68), which we treat as a feature-availability stress test rather than a second validation. Cross-cohort SHAP rankings correlated strongly in China (Spearman {rho} = 0.91) and moderately in Italy ({rho} = 0.76), where most laboratory predictors were imputed and carried no information beyond the complete blood count; discrimination tracked how many predictors were actually measured. The admission-time signature converges with established markers of hypoperfusion and inflammation. Author summaryWhen patients arrive at hospital with COVID-19, risk must be assessed immediately to direct limited resources to those in need, particularly where intensive-care capacity is scarce, as in Pakistan. Most existing computer-based tools come from developed countries, act as "black boxes", and are rarely tested elsewhere. We built a model using first-day admission data from 321 patients at a hospital in Lahore, Pakistan. It identifies patients who later died best, and milder cases least well. The strongest markers were blood oxygen, lactate dehydrogenase, urea, C-reactive protein, age and chest-imaging evidence of pneumonia, together pointing to a coherent picture of pulmonary impairment, inflammation and circulatory stress. In fatal cases urea was markedly raised while creatinine rose only modestly, a pattern that most likely reflects dehydration. We then tested the same model on independent Chinese and Italian cohorts: it transferred well to China but lost accuracy in Italy, where only a basic blood count was available. The findings support an interpretable Day-1 prognostic tool whose useful range is set largely by which laboratory tests are available rather than by geography. Because these markers reflect general inflammation rather than anything unique to COVID-19, the approach may extend to other pneumonias.
ER -