artificial intelligence in health games: a systematic review on opportunities, challenges, and future clinical implementation

Background: digital health games integrating artificial intelligence (ai) and medical internet of things (iomt) technologies show transformative potential for enhancing health behaviors and patient self-management. this systematic review evaluates clinical efficacy, implementation challenges, and future development strategies of ai-powered health games.method: following prisma 2020 guidelines, we conducted a mixed-methods systematic review searching pubmed, scopus, and ieee xplore (inception to march 2023). using pre-registered keywords, dual independent screening (κ=0.82) of 2,137 records yielded 45 included studies. these underwent standardized quantitative/qualitative data extraction, quality assessment (robins-i, grade), and synthesis via combined thematic analysis and meta-analysis.results: the analysis demonstrated significant clinical efficacy with a pooled 45.2% improvement in fugl-meyer scores for neurorehabilitation (95% ci 38.1-52.3%; i²=12%) and 30.1% cortisol reduction in mental health applications (p<0.001). critical implementation barriers included algorithmic bias (37.4%), privacy concerns (68% of wearable studies), and heterogeneous evaluation frameworks (i²=89%). a three-component conceptual framework—comprising screening, wearable integration, and patient engagement—was proposed as critical for effective implementation.conclusion: while ai-health games demonstrate significant clinical potential (grade b evidence), key limitations persist, including heterogeneous outcome measures and scarce long-term data. our synthesis highlights the need for standardized evaluation criteria, federated learning to mitigate algorithmic biases, and robust regulatory frameworks. this unfunded study provides foundational evidence for these recommended advancements in digital therapeutics.