Integrated rapid diagnostics and AI prediction accelerate targeted antibiotic therapy in a resource-limited setting
Investigators evaluated a prospective cohort of 410 adults with suspected bacterial infections at a Pakistani hospital to assess an integrated diagnostic strategy combining rapid testing, molecular resistance profiling, and an AI prediction model. The study reports that this approach significantly reduced time to pathogen identification and effective therapy initiation, increased appropriate antibiotic use from 54.2% to 78.3%, and lowered hospital stay and mortality compared to conventional susceptibility testing. Genotypic and phenotypic resistance results demonstrated substantial concordance, while the AI model achieved an AUC of 0.89. These findings indicate that pairing rapid diagnostics with predictive analytics can strengthen antimicrobial stewardship, offering a scalable framework for laboratories operating in resource-constrained settings.
The original study
Integrated evaluation of rapid diagnostic testing, genotypic-phenotypic resistance profiling, and AI-Assisted prediction models for antimicrobial stewardship and clinical outcomes in a resource-limited setting.
- Authors
- Hammad M, Arif R, Fardoos S, Shakoor K, Nasir A
- Journal
- PloS one
- Type
- Journal Article, Observational Study
- PMID
- 42623344
Original abstract
INTRODUCTION: Antimicrobial resistance (AMR) remains a major global health concern, necessitating timely and accurate diagnostic approaches to guide appropriate therapy. Conventional antibiotic susceptibility testing (AST) is often associated with delays that may compromise clinical outcomes. OBJECTIVE: To evaluate the impact of integrating rapid diagnostic testing, genotypic resistance profiling, and artificial intelligence (AI)-based prediction models on antimicrobial stewardship and clinical outcomes. METHODS: A prospective observational study was conducted at Lady Reading Hospital, MTI, Peshawar, Pakistan from 13/02/2023-18/04/2025, including 410 adult patients with suspected bacterial infections. Participants underwent conventional AST, rapid diagnostic testing, and molecular detection of resistance genes. An AI-based model was developed using clinical and laboratory parameters to predict antimicrobial resistance. Key outcomes included time to effective therapy, antibiotic appropriateness, length of hospital stay, and mortality. Statistical analysis included comparative tests, logistic regression, and receiver operating characteristic (ROC) curve analysis and was performed using SPSS Version 26.0 and R version 4.5.2. RESULTS: Rapid diagnostics significantly reduced time to pathogen identification (10.4 vs 48.6 hours, p < 0.001) and initiation of effective therapy (20.3 vs 55.8 hours, p < 0.001). Appropriate antibiotic use improved from 54.2% to 78.3% (p < 0.001), while broad-spectrum antibiotic use declined significantly. The integrated approach was associated with reduced hospital stay (8.6 vs 12.9 days, p < 0.001) and lower mortality (14.0% vs 22.7%, p = 0.02). Genotypic-phenotypic concordance was substantial (κ = 0.61-0.69). The AI model demonstrated strong predictive performance (AUC = 0.89). CONCLUSION: An integrated diagnostic and predictive approach was associated with improvements in antimicrobial stewardship and favorable clinical outcomes, suggesting its potential as a cost-effective and scalable strategy for addressing antimicrobial resistance (AMR), particularly in resource-limited settings.