Multi-Output Gradient Boosting for Interpretable Multi-Disease Prediction Using Minimal-Feature Survey Data

This research is about Chronic diseases including diabetes mellitus, cardiovascular disease (CVD) and chronic kidney disease (CKD) have strong pathophysiological interdependencies, but most machine learning predictors either consider them separately or require longitudinal electronic health records unavailable in resourceconstrained settings. This work helps fill an important gap by presenting some lightweight and interpretable multidisease frameworks that is works for population-level screening based on minimal feature survey data. We developed a multi-output gradient boosting classifier trained with 70692 respondents from the Behavioral Risk Factor Surveillance System (BRFSS) using only 19 demographic and lifestyle features, SHapley Additive exPlanations (SHAP) for model interpretability and rigorous external validation on an independent cohort. The framework was able to simultaneously predict diabetes (area under the receiver operating characteristic curve [AUC] = 0.831, 95% CI: 0.826-0.837) and CVD (AUC = 0.809, 95% CI: 0.803-0.817) and CKD risk proxy (AUC = 0.962, 95% CI: 0.960-0.964) agents with an exceptional stability during external validation (|∆AUC| < 0.0 SHAP analysis resulted in rules clinician friendly, such as 'Age under 65 years increases CKD risk by +1.262 SHAP units, which were confirmed through structured physician's feedback forms. The results show the practicality of implementing clinically relevant multi-disease prediction based on resource-efficient ensemble techniques, independent of EHR systems and GPU-intensive models, which is crucial for the deployment of such approaches in primary care settings where longitudinal health records are very rare to access.

Authors

Institutions

Publication Details

Journal
International Journal of Drug Delivery Technology
Published
2026-11-16
DOI
https://doi.org/10.25258/ijddt.16.80s.83
Primary Topic
Artificial Intelligence in Healthcare
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Multi-Output Gradient Boosting for Interpretable Multi-Disease Prediction Using Minimal-Feature Survey Data

Sajja Tulasi Krishna, K.R. Manjula Devi
International Journal of Drug Delivery Technology
Artificial Intelligence in Healthcare
article

Multi-Output Gradient Boosting for Interpretable Multi-Disease Prediction Using Minimal-Feature Survey Data

Sajja Tulasi Krishna, K.R. Manjula Devi
article en

Abstract

This research is about Chronic diseases including diabetes mellitus, cardiovascular disease (CVD) and chronic kidney disease (CKD) have strong pathophysiological interdependencies, but most machine learning predictors either consider them separately or require longitudinal electronic health records unavailable in resourceconstrained settings. This work helps fill an important gap by presenting some lightweight and interpretable multidisease frameworks that is works for population-level screening based on minimal feature survey data. We developed a multi-output gradient boosting classifier trained with 70692 respondents from the Behavioral Risk Factor Surveillance System (BRFSS) using only 19 demographic and lifestyle features, SHapley Additive exPlanations (SHAP) for model interpretability and rigorous external validation on an independent cohort. The framework was able to simultaneously predict diabetes (area under the receiver operating characteristic curve [AUC] = 0.831, 95% CI: 0.826-0.837) and CVD (AUC = 0.809, 95% CI: 0.803-0.817) and CKD risk proxy (AUC = 0.962, 95% CI: 0.960-0.964) agents with an exceptional stability during external validation (|∆AUC| < 0.0 SHAP analysis resulted in rules clinician friendly, such as 'Age under 65 years increases CKD risk by +1.262 SHAP units, which were confirmed through structured physician's feedback forms. The results show the practicality of implementing clinically relevant multi-disease prediction based on resource-efficient ensemble techniques, independent of EHR systems and GPU-intensive models, which is crucial for the deployment of such approaches in primary care settings where longitudinal health records are very rare to access.

International Journal of Drug Delivery TechnologyVol. 16(80s)
Koneru Lakshmaiah Education Foundation (IN)
Good health and well-being
Artificial Intelligence in Healthcare
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon