Abstract
The rapid development of artificial intelligence, including large language models (LLMs), creates new opportunities for prediction modeling research in healthcare. LLMs can understand and process large amounts of text, which may help automate text-related research tasks that are otherwise time-consuming and resource-intensive. This is particularly relevant for tasks such as extracting information from electronic health records (EHRs) and conducting systematic reviews. For extraction of information form EHRs, LLMs can be used to automatically identify patient cohorts and predictors from EHRs, although extraction errors remain and their added value for prediction models may depend on the specific context. In systematic reviews, LLMs appear most useful for tasks such as literature searching and screening. However, for both applications, LLM evaluation has mainly focused on how accurately the LLM performs a task, rather than on its impact on the research process or outcomes in practice. In addition, standardized evaluations that can support implementation in healthcare research are still limited. LLM performance can also vary depending on factors such as data quality, research topic, and the specific task. Therefore, although substantial progress has been made in using LLMs to automate prediction research, future research should focus more on their practical impact. Evaluating whether LLMs actually make research more efficient, reliable, or effective in real-world settings will be important for their responsible implementation in research practice.
| Original language | English |
|---|---|
| Awarding Institution |
|
| Supervisors/Advisors |
|
| Award date | 3 Nov 2026 |
| Publisher | |
| Print ISBNs | 978-94-6537-954-8 |
| DOIs | |
| Publication status | Published - 3 Nov 2026 |
Keywords
- large language models
- natural language processing
- artificial intelligence
- electronic health records
- prediction modeling
- prognosis research
- systematic reviews
- prioritized screening
- automation
- reporting guidelines
Fingerprint
Dive into the research topics of 'Large language models to advance prediction modeling research in healthcare'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver