AI Diagnostics in Resource-Limited Healthcare Settings of Malawi: A Comparative Analysis

AI diagnostics are increasingly being explored as a potential solution to improve disease diagnosis in resource-limited healthcare settings, particularly in sub-Saharan Africa where access to advanced medical infrastructure and trained personnel is often constrained. A comparative analysis was conducted using a dataset of patient records from two major hospitals in Malawi. Machine learning algorithms were trained on historical diagnostic data to predict disease outcomes based on clinical symptoms and basic laboratory results. The AI models' performance was evaluated against the consensus diagnoses provided by human clinicians, with accuracy measured using a confusion matrix. AI models demonstrated an overall accuracy of 85% in diagnosing malaria and tuberculosis compared to clinician assessments, which varied between 80-90%. Notably, AI outperformed clinicians in cases where laboratory results were inconclusive or limited. This suggests that AI could provide a valuable adjunctive tool for improving diagnostic precision. The findings indicate potential benefits of integrating AI diagnostics into routine healthcare workflows to support resource-limited settings and enhance disease management outcomes. Given the promising preliminary results, it is recommended that further studies be conducted to validate these findings across different populations and contexts. Additionally, targeted training programmes for clinicians should be developed to ensure effective integration of AI diagnostic tools. AI diagnostics, Malawi, resource-limited healthcare, machine learning, disease prediction Model estimation used $\hat{\theta}=argmin_{\theta}\sum_i\ell(y_i,f_\theta(x_i))+\lambda\lVert\theta\rVert_2^2$, with performance evaluated using out-of-sample error.