Performance Evaluation of a Clinical Decision Support System (MAMAOPE AI) for Tuberculosis and Pneumonia Diagnosis

This study evaluated the performance, usability, and clinical relevance of the MAMAOPE Clinical Decision Support System (CDSS)—an artificial intelligence–enabled diagnostic tool designed to support frontline healthcare workers in the differential diagnosis of tuberculosis (TB) and pneumonia in low-resource settings in Uganda

Tuberculosis and pneumonia remain leading causes of morbidity and mortality in low- and middle-income countries, particularly among children and people living with HIV, where overlapping symptoms and limited access to advanced diagnostics exacerbate diagnostic delays. The MAMAOPE CDSS was developed to address these gaps by leveraging machine-learning algorithms informed by national and international clinical guidelines to analyze routinely collected patient data, including symptoms, vital signs, and oxygen saturation, and to generate ranked differential diagnoses in real time.

A sequential explanatory mixed-methods study was conducted between April and May 2025 across five public and private health facilities in Kampala. The quantitative component assessed diagnostic accuracy among 168 patients presenting with TB or pneumonia symptoms, comparing CDSS outputs against clinician reference diagnoses. The qualitative component explored usability, feasibility, and acceptability through in-depth interviews with frontline clinicians.

The CDSS demonstrated high sensitivity and negative predictive value for both TB (89.5% sensitivity; 97.9% NPV) and pneumonia (92.3% sensitivity; 95.2% NPV), indicating strong potential as a screening and triage tool. Diagnostic agreement with clinicians was moderate overall but improved substantially when only first-ranked CDSS diagnoses were considered. Lower oxygen saturation was a significant predictor of concordance between CDSS and clinician diagnoses. Qualitative findings revealed that clinicians perceived the system as fast, intuitive, and well aligned with clinical reasoning and national guidelines, particularly valuing its role in supporting task-sharing, improving workflow efficiency, and promoting rational prescribing.

The study concludes that the MAMAOPE CDSS can meaningfully enhance early identification of TB and pneumonia in resource-constrained settings. However, limitations related to diagnostic specificity, connectivity dependence, and restricted disease scope highlight the need for further refinement, broader validation, and integration with additional diagnostic data sources.