An Online Survey of Healthcare Professionals to Understand the Use of Clinical Prediction Models in Practice

Healthcare professionals value clinical prediction models but highlight the need for user-friendly and well-integrated systems to enhance their practical application.

Summary and findings

This study explored healthcare professionals' experiences with clinical prediction models (CPMs), focusing on their perceived usefulness and barriers to implementation. A total of 157 healthcare professionals participated in an online survey. The median usefulness rating of CPMs was 8 out of 10.

Abstract

<title>Abstract</title> <p> <bold>Introduction</bold> Clinical prediction models (CPMs) are increasingly promoted as decision-support tools that can improve patient care. However, their real-world utilisation remains variable, and little is known about the factors influencing their adoption across specialties. This study aimed to explore healthcare professionals’ experiences with CPMs, including their perceived usefulness, preferred formats, and the barriers and facilitators to their implementation. <bold>Methods</bold> A cross-sectional survey of healthcare professionals using an online questionnaire using REDCap software was distributed via social media and professional mailing lists, worldwide. <bold/> Healthcare professionals from a range of clinical and allied health specialties were invited to complete the online survey. The survey captured demographic data, CPM usage, perceptions of usefulness (via Likert scale), preferred access methods and presentation formats, and included free-text questions for qualitative analysis. Quantitative data were summarised descriptively, and qualitative responses were analysed thematically. <bold>Results</bold> A total of 157 respondents completed the survey, of whom 62% were doctors and 87% were based in the UK. Most respondents (82%) were aware of CPMs, and 71% used them in clinical practice, rating their usefulness at a median of 8 out of 10. The most used models predicted age-related events (e.g. CHA2DS2-VASc, QRISK), while obstetrics and gynaecology (O&G)-specific models were underutilised despite O&G being the most represented specialty. Preferred access methods included web-based tools and mobile applications, and participants favoured percentage or heat map formats for risk presentation. Barriers included poor integration into workflows, limited awareness, regulatory uncertainty, and inconsistent use across teams. Three themes were developed from qualitative free-text responses; why clinicians believe CPMs are clinically useful, how to integrate CPMs into everyday systems and the need for continuous professional education about CPMs. <bold>Conclusions</bold> Healthcare professionals value CPMs but emphasise the need for models that are transparent, user-friendly, and integrated into clinical systems. As AI-driven approaches become more common, concerns around interpretability and inconsistent regulatory classification—such as whether CPMs are considered medical devices—must be addressed. To ensure credibility and usability, future models should adhere to TRIPOD or TRIPOD-AI reporting standards, undergo external validation in independent datasets, and be evaluated for clinical impact. Coordinated efforts by researchers, journals, funders, and regulators are essential to move from model development to meaningful implementation in practice. </p>

Key findings

• Not reported in abstract.

Limitations

• Not reported in abstract.

Elsewhere in the PE 22-28 corpus