Scholarlink

Aim <p>Multimorbidity is complex due to interactions between diseases and treatments, resulting in varied patterns and outcomes. Machine learning enables the identification of patterns in large datasets. This scoping review aims to explore current research on machine learning for multimorbidity, focusing on models, input data, disease types, applications, and algorithm maturity.</p> Subject and methods <p>The Preferred Reporting Items for Systematic Reviews and Meta-analyses extension for Scoping Reviews (PRISMA-ScR) was used. Searches were conducted in PubMed, EMBASE, IEEE, Web of Science, and Scopus. Articles were assessed by two researchers. Inclusion required analysis of more than one chronic disease or a single condition at risk of another.</p> Results <p>40 studies met the inclusion criteria. Ensemble models (45%) and neural networks (28%) were the most applied algorithms. The most frequently applied input data were socio-demographic (21%), clinical diagnosis (17%), laboratory (16%), and functional data (13%). Diabetes and cardiovascular diseases were the most common co-occurring chronic conditions. Diabetes, cardiovascular, respiratory, and mental diseases were the most included conditions. Most studies focused on diagnosis (58%), with fewer addressing treatment (28%), and prognosis (15%). Most studies only performed internal validation (88%), whereas studies performing external validation were scarce (10%).</p> Conclusion <p>This scoping review on machine learning revealed a predominant emphasis on diagnostic aspect, alongside with frequent utilization of ensemble models and neural networks. Treatment and prognosis applications were limited, and only few studies evaluate algorithm performance in clinical practice. Future research should aim to enhance the maturity of these algorithms.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Application of machine learning in multimorbidity research: A scoping review

  • 期刊论文

  • 发表日期：

    2025年9月3日

查看全文
  • Danny Jeganathan Anthonimuthu,
  • Anne-Maj Knudsen,
  • Ole Kristian Hejlesen,
  • Ann-Dorthe Zwisler,
  • Flemming Witt Udsen

摘要

Aim

Multimorbidity is complex due to interactions between diseases and treatments, resulting in varied patterns and outcomes. Machine learning enables the identification of patterns in large datasets. This scoping review aims to explore current research on machine learning for multimorbidity, focusing on models, input data, disease types, applications, and algorithm maturity.

Subject and methods

The Preferred Reporting Items for Systematic Reviews and Meta-analyses extension for Scoping Reviews (PRISMA-ScR) was used. Searches were conducted in PubMed, EMBASE, IEEE, Web of Science, and Scopus. Articles were assessed by two researchers. Inclusion required analysis of more than one chronic disease or a single condition at risk of another.

Results

40 studies met the inclusion criteria. Ensemble models (45%) and neural networks (28%) were the most applied algorithms. The most frequently applied input data were socio-demographic (21%), clinical diagnosis (17%), laboratory (16%), and functional data (13%). Diabetes and cardiovascular diseases were the most common co-occurring chronic conditions. Diabetes, cardiovascular, respiratory, and mental diseases were the most included conditions. Most studies focused on diagnosis (58%), with fewer addressing treatment (28%), and prognosis (15%). Most studies only performed internal validation (88%), whereas studies performing external validation were scarce (10%).

Conclusion

This scoping review on machine learning revealed a predominant emphasis on diagnostic aspect, alongside with frequent utilization of ensemble models and neural networks. Treatment and prognosis applications were limited, and only few studies evaluate algorithm performance in clinical practice. Future research should aim to enhance the maturity of these algorithms.