Understanding Adoption Patterns of AI-based Clinical Decision Support Systems in Kenya’s Healthcare
With the advancement of technology in healthcare, utilizing large language model (LLM)-based clinical decision support systems (CDSS) presents a transformative opportunity. This article delves into the adoption patterns of such a system in private primary healthcare facilities in Kenya, operated by Penda Health, and examines the factors influencing its acceptance among clinicians.
Methods and Analysis
This study employs a mixed-methods approach, integrating quantitative analysis with qualitative insights. Data were collected from consultations conducted between February 1 and October 1, 2024, across Penda Health facilities. The quantitative analysis focused on CDSS metadata, while qualitative data were gathered from 42 staff members, including clinical officers, facility managers, and other healthcare professionals. The data collection methods included journey mapping interviews, user experience interviews, focus groups, and analysis of system usage metrics.
The quantitative data were summarized using descriptive statistics, while qualitative data underwent thematic analysis guided by established technology adoption and change management theories.
Results
During the eight-month observation period, Penda Health recorded 258,106 clinical episodes, with 56,050 (21.7%) augmented by AI Consult. Notably, the use of AI Consult increased from 4% to 47% in 16 institutions over this period. Clinicians provided feedback in 31% of the clinical episodes, with an overwhelming 99.5% of the feedback being positive.
The qualitative analysis revealed five key themes related to clinicians’ experiences with AI Consult:
- Multiple value propositions exist for an AI Consult-style tool.
- Clinicians’ confidence in AI Consult grew over time.
- Usage is influenced by case complexity.
- AI Consult responses are generally trusted and valued, though improvements are recommended.
- The tool is easy to use, yet certain issues require attention.
Diploma
The study highlights essential requirements for the successful implementation of generative AI/LLM-powered CDSS in resource-constrained environments. These include establishing a robust technological infrastructure, localizing to address clinical guidelines, structured change management with clinical experts, and seamless workflow integration. Future development should consider alternatives to actively soliciting CDSS input, acknowledging the risk of underutilization due to potential over-reliance.
For further details and insights, you can access the full study Here.
“`
