JMIR Publications

ABSTRACT

Background:

Physicians are hesitant to forgo the opportunity of entering unstructured clinical notes for structured data entry in electronic health records. Does the flexibility of free text also increase informational value?

Objective:

To compare information from unstructured clinical notes harvested and processed by a natural language processing algorithm, to clinician-entered structured data as to their potential utility for automated improvement of patient workflows.

Methods:

EHR of 293’298 patient visists at a Swiss university hospital from 01/2014 to 10/2021 were analyzed. Using emergency department overcrowding as a case in point, we compared supervised natural language processing (NLP)-based symptom clusters from unstructured clinical notes and clinician-entered chief complaints from a structured drop-down menu with the two outcomes hospitalization and high ESI score.

Results:

In 11 of the 12 symptom clusters, the NLP cluster was significant in predicting hospitalization. Eight clusters remained significant even when controlling for the cluster of clinician-determined chief complaints in the model. All 12 NLP symptom clusters were significant in predicting a low ESI score, of which 9 remained significant when controlling for clinician-determined chief complaints. The correlation of NLP clusters with chief complaints was low (-0.04-0.6), indicating a complementarity of information.

Conclusions:

The NLP-derived features and clinicians’ knowledge were complementary in explaining patient outcome heterogeneity. They can provide an efficient approach to patient flow management, for example, in an emergency medicine setting. We further demonstrate the feasibility of creating extensive and precise keyword dictionaries with NLP by medical experts without requiring programming knowledge. Using the dictionary, we could classify short and unstructured clinical texts into diagnostic categories defined by the clinician. Clinical Trial: None