Systematic AI Support for Decision-Making in the Healthcare Sector: Obstacles and Success Factors

Abstract

Background

Currently, health care is expert-centric, especially with regard to decision-making. Innovations such as artificial intelligence (AI) or interconnected electronic health records (EHRs) suffer from low adoption rates. In the rare cases of technically successful implementation, they often result in inefficient or error-prone processes.

Aim & Methods

This paper explores the state of the art in AI-based digital decision support systems (DDSSs). To overcome the low adoption rates, we propose a systematic strategy for bringing DDSS research into clinical practice based on a design science approach. DDSSs can transform health care to be more innovative, patient-centric, accurate and efficient. We contribute by providing a framework for the successful development, evaluation and analysis of systems for AI-based decision-making. This framework is then evaluated using focus group interviews.

Results

Centred around our framework, we define a systematic approach for the use of AI in health care. Our systematic AI support approach highlights essential perspectives on DDSSs for systematic development and analysis. The aim is to develop and promote robust and optimal practices for clinical investigation and evaluation of DDSS in order to encourage their adoption rates. The framework contains the following dimensions: disease, data, technology, user groups, validation, decision and maturity.

Conclusion

DDSSs focusing on only one framework dimension are generally not successful; therefore, we propose to consider each framework dimension during analysis, design, implementation and evaluation so as to raise the number of DDSSs used in clinical practice.

Public Interest Summary

The digital transformation of the healthcare sector creates the potential for the sector to be more accurate, efficient and patient-centric using AI, or so-called digital decision support systems. In this research, we explore why these systems are needed and how they can be successfully implemented in clinical practice. For this, we propose a systematic approach based on our conceptual framework. Against this background, we present our vision for further advancing these technologies. We see our systematic AI support as a primary driver, with the possibility to facilitate the much-needed breakthrough of decision support systems in health care.