JMIR Publications

ABSTRACT

Background:

Providing patients access to their medical notes has been demonstrated to offer many benefits for patients and providers. This has led to a rapidly expanding national movement, OpenNotes, which provides resources to clinicians who desire to share medical notes with their patients. However, a significant barrier to the widespread adoption of OpenNotes is clinicians' concern that the medical terminology in such notes may confuse patients.

Objective:

Artificial intelligence (AI) software may provide the opportunity to rapidly simplify medical notes to plain language through natural language processing. This offers the potential to resolve concerns over medical terminology and patient confusion. This pilot study assesses patient’s perception of AI-simplified plain language medical notes.

Methods:

Patient’s perception of notes was studied through comprehension questionnaires and guided interviews with subsequent thematic analysis. Study participants were recruited from patients hospitalized at the University of Colorado Hospital. A standardized cardiology patient’s note was generated using a synthetic patient generator. AI software applied natural language processing to produce a simplified version of this note. Patients were randomly assigned to first read either the original template note or the simplified version, then completed a set of seven comprehension assessment questions. Subsequently, patients reviewed the opposite version of the note and participated in a guided interview to discuss their thoughts on these notes. Statistical analysis of comprehension questionnaire scores was performed to assess for differences in comprehension between original (unsimplified) and simplified notes. Guided interview responses were thematically analyzed.

Results:

Twenty patients agreed to participate. The study was found to be underpowered to detect statistical significance of the impact of simplified notes on participant comprehension. Though the mean number of comprehension assessment questions answered correctly was found to be higher in the simplified note group at 4.7 as compared to 3.9 in the unsimplified note group, this was found to be non-significant (p=0.32). Guided interviews found that AI simplified open notes were perceived as desirable and beneficial by participants. Thematic analysis identified that simplified medical notes may (1) be more useable than unsimplified notes, (2) improve the patient-provider relationship, and (3) empower patients through an enhanced understanding of their conditions and management. Participant’s recommendations highlighted the need to reduce lengthy plain-language phrases and to target the level of simplification to each patient’s health literacy.

Conclusions:

Simplified notes were well received by participants, who expressed a desire to have access to such notes for their own medical conditions. This study illustrates the potential for artificial intelligence software to quickly generate plain language medical notes that are useful for patients and their providers. Feedback from participants in this study should be used to improve the simplification of notes. Larger studies should be conducted with heed to the insight gained from this pilot study.