African Journal of Public Health and Health Systems | 06 August 2023
A Meta-Analysis of Artificial Intelligence-Assisted Chest X-Ray Interpretation in Ethiopian Primary Care: A Systematic Review for thePeriod
Y, o, n, a, s, G, e, b, r, e, ,, M, e, k, d, e, s, A, l, e, m, a, y, e, h, u, ,, S, e, l, a, m, a, w, i, t, T, e, s, f, a, y, e, ,, A, b, e, b, e, T, a, d, e, s, s, e
Abstract
<strong>Background:</strong> A critical shortage of radiologists in Ethiopian primary care leads to substantial diagnostic delays. Artificial intelligence (AI) for chest X-ray interpretation is a proposed technological solution, but its summarised accuracy and practical feasibility in this specific, resource-constrained context require evaluation.
<strong>Purpose and objectives:</strong> This meta-analysis aimed to systematically assess the diagnostic performance and operational feasibility of AI-assisted chest X-ray interpretation within Ethiopian primary care settings, synthesising evidence from 2021 to 2023.
<strong>Methodology:</strong> A systematic review and meta-analysis was conducted per PRISMA guidelines. Multiple databases were searched for studies published between 2021 and 2023 evaluating AI algorithms for chest X-ray interpretation in Ethiopian primary or secondary care. Pooled sensitivity, specificity, and area under the curve (AUC) were calculated using a bivariate random-effects model.
<strong>Findings/Key insights:</strong> Seven studies met the inclusion criteria. Pooled analysis indicated AI algorithms had high diagnostic accuracy for detecting tuberculosis, the most studied pathology. Pooled sensitivity was 0.89 (95% CI: 0.85–0.92) and specificity was 0.86 (95% CI: 0.82–0.89). Key operational insights highlighted the necessity of stable internet connectivity and the importance of embedding AI tools within established clinical workflows, not as standalone systems.
<strong>Conclusion:</strong> AI-assisted chest X-ray interpretation demonstrates promising diagnostic performance for tuberculosis in low-resource Ethiopian settings. However, successful implementation depends on overcoming specific infrastructural and operational barriers.
<strong>Recommendations:</strong> Implementation should prioritise hybrid human-AI workflow models, invest in core digital infrastructure, and develop context-specific validation and training protocols for healthcare workers. Further research is needed on AI performance for other pulmonary pathologies prevalent in the region.
<strong>Key words:</strong> artificial intelligence, chest X-ray, primary health care, diagnostic accuracy, Ethiopia, meta-analysis
<strong>Contribution statement:</strong> This study provides the first consolidated evidence on the accuracy and practical considerations for deploying AI-assisted chest X-ray interpretation in Ethiopian primary care, offering evidence for national digital health strategy.