AI-Driven Medical Reporting in Nigerian Hospitals: A Comparative Analysis of GPT-4 Omni and Gemini Advanced

Abstract: This study evaluates and compares the performance of GPT-4 Omni and Gemini Advanced in automating and enhancing medical reporting within resource-constrained Nigerian hospitals, focusing on accuracy, efficiency, and usability. A comparative case study was conducted across three participating Nigerian hospitals. Data sources included anonymized, simulated patient cases derived from the MIMIC-III database and Google Open Health Data, physician interviews (n=15), and, where available, limited access to existing Electronic Medical Record (EMR) data focusing on discharge summaries. Both AI models were assessed using pre-defined metrics: (1) report accuracy (measured by error rate and physician agreement using Cohen's Kappa), (2) time efficiency (percentage reduction in reporting time compared to manual methods), (3) completeness (percentage of key data points included as determined by expert physicians), and (4) physician perceived usability (System Usability Scale - SUS and select questions from the Technology Acceptance Model - TAM). Data analysis involved descriptive statistics, comparative t-tests, and thematic analysis of interview transcripts. Preliminary results indicate a significant reduction in reporting time with both models (GPT-4 Omni: 42%, p<0.05; Gemini Advanced: 38%, p<0.05). GPT-4 Omni demonstrated a higher accuracy rate in generating discharge summaries (92% accuracy, Cohen's Kappa = 0.85) compared to Gemini Advanced (85% accuracy, Cohen's Kappa = 0.78), with both models achieving high levels of agreement with physician experts. A stronger correlation was observed between perceived usability (SUS score) and intention to adopt AI reporting for GPT-4 Omni (r = 0.72, p<0.01) than for Gemini Advanced (r = 0.61, p<0.05). This study provides valuable insights into the potential of AI-driven medical reporting to improve healthcare delivery in resource-limited settings. The findings inform the development of AI solutions tailored to the specific needs of Nigerian hospitals and contribute to a broader understanding of AI adoption in healthcare. Future research should prioritize the seamless integration of these models with existing EMR systems, address data security concerns, and evaluate the long-term impact on patient outcomes and clinician workflow. Keywords: AI in healthcare, Medical reporting, Large language models, GPT-4 Omni, Gemini Advanced, Nigerian hospitals, Resource-constrained settings, Technology adoption, Usability, Accuracy, Time efficiency, Electronic Medical Records (EMR), Digital health, Machine learning, Natural Language Processing (NLP), Healthcare informatics, Health information systems, Developing countries, Technology Acceptance Model (TAM), Clinical documentation, Ethical AI, Bias mitigation, Algorithmic transparency, MIMIC-III, Google Open Health Data, Comparative case study, Healthcare technology, Clinical decision support, Medical AI, Nigeria, Health information technology, Doctor workload, Automated reports, Patient data.