Hybrid Collaborative Filtering Models for Clinical Search Recommendation

TL;DR

This paper presents a hybrid collaborative filtering model that leverages patient encounter data and search history to improve search term recommendations for clinicians, enhancing efficiency in clinical information retrieval.

Contribution

The study introduces a novel hybrid collaborative filtering approach tailored for clinical search recommendations, outperforming existing methods in accuracy.

Findings

Model outperforms state-of-the-art baselines in top-N recommendation tasks.

Effective use of encounter and search term data improves recommendation relevance.

Demonstrated robustness across multiple datasets.

Abstract

With increasing and extensive use of electronic health records, clinicians are often under time pressure when they need to retrieve important information efficiently among large amounts of patients' health records in clinics. While a search function can be a useful alternative to browsing through a patient's record, it is cumbersome for clinicians to search repeatedly for the same or similar information on similar patients. Under such circumstances, there is a critical need to build effective recommender systems that can generate accurate search term recommendations for clinicians. In this manuscript, we developed a hybrid collaborative filtering model using patients' encounter and search term information to recommend the next search terms for clinicians to retrieve important information fast in clinics. For each patient, the model will recommend terms that either have high co-occurrence frequencies with his/her most recent ICD codes or are highly relevant to the most recent search terms on this patient. We have conducted comprehensive experiments to evaluate the proposed model, and the experimental results demonstrate that our model can outperform all the state-of-the-art baseline methods for top-N search term recommendation on different datasets.