Community-Based Early-Stage Chronic Kidney Disease Screening using Explainable Machine Learning for Low-Resource Settings

TL;DR

This study tackles the challenge of early-stage CKD screening in low-resource South Asian settings by developing an explainable ML framework tailored to Bangladeshi populations. It systematically optimizes feature selection from non-invasive inputs and validates performance across three external datasets, achieving balanced accuracies around 90% with full features and around 89% with non-laboratory inputs, while maintaining robust sensitivity. SHAP-based explanations reveal clinically meaningful predictors such as Age_60+y and Hypertension, enhancing interpretability and trust for community adoption. The work demonstrates that compact, accessible feature subsets can outperform larger, lab-heavy sets, offering a practical path toward scalable CKD screening in resource-constrained environments with strong generalizability across diverse populations.

Abstract

Early detection of chronic kidney disease (CKD) is essential for preventing progression to end-stage renal disease. However, existing screening tools - primarily developed using populations from high-income countries - often underperform in Bangladesh and South Asia, where risk profiles differ. Most of these tools rely on simple additive scoring functions and are based on data from patients with advanced-stage CKD. Consequently, they fail to capture complex interactions among risk factors and are limited in predicting early-stage CKD. Our objective was to develop and evaluate an explainable machine learning (ML) framework for community-based early-stage CKD screening for low-resource settings, tailored to the Bangladeshi and South Asian population context. We used a community-based dataset from Bangladesh, the first such CKD dataset in South and South Asia, and evaluated twelve ML classifiers across multiple feature domains. Ten complementary feature selection techniques were applied to identify robust, generalizable predictors. The final models were assessed using 10-fold cross-validation. External validation was conducted on three independent datasets from India, the UAE, and Bangladesh. SHAP (SHapley Additive exPlanations) was used to provide model explainability. An ML model trained on an RFECV-selected feature subset achieved a balanced accuracy of 90.40%, whereas minimal non-pathology-test features demonstrated excellent predictive capability with a balanced accuracy of 89.23%, often outperforming larger or full feature sets. Compared with existing screening tools, the proposed models achieved substantially higher accuracy and sensitivity while requiring fewer and more accessible inputs. External validation confirmed strong generalizability with 78% to 98% sensitivity. SHAP interpretation identified clinically meaningful predictors consistent with established CKD risk factors.

Figures (6)

• Figure 1: A schematic overview of our methodology
• Figure 1: Distribution of best-performing machine learning models identified across all experiments using different feature subsets. The counts reflect how often each classifier achieved the highest balanced accuracy within a given feature configuration.
• Figure 2: Performance comparison of machine learning models trained on three different feature configurations: the full feature set, the best-performing subset excluding pathology features (S2), and the best-performing subset across all feature domains (S1). Error bars represent 95% confidence intervals computed over 10-fold cross validation.
• Figure 3: Confusion matrices illustrating the prediction results for CKD and non-CKD cases using three feature configurations: (a) the full feature set, (b) the best-performing subset without pathology features, and (c) the best-performing subset including all features. Values represent both the counts and percentages of true positives, true negatives, false positives, and false negatives.
• Figure 4: SHAP summary plot illustrating the contribution of the best-performing S1 feature set to the Decision Tree model's predictions for CKD detection.