Comprehensive Summary
This study, presented by Wasser et al., examines the detection of transportation barriers in ophthalmology care using natural language processing (NLP). Researchers analyzed deidentified demographic information and free text from electronic health records, training an NLP model to identify transportation-related keywords (e.g., “transportation,” “ride”). In validation, the model achieved precision of 0.860, recall of 0.960, and an F1-score of 0.778, demonstrating strong alignment with expert review. When applied to 1,801,572 clinical notes, the algorithm identified 726 patients (0.6%) with transportation issues. Patients with transportation barriers were more likely to be older (adjusted OR = 3.01) and less likely to identify as Asian (adjusted OR = 0.03). No significant sex differences were found. The model’s higher precision than recall suggests under-detection, meaning results likely represent a lower bound of true transportation needs. Overall, NLP shows promise for uncovering social determinants of health such as transportation access in clinical populations.
Outcomes and Implications
This research is important as transportation insecurities are a major reason patients miss appointments or delay care. Specifically in the field of ophthalmology, missed visits can lead to vision loss that negatively impacts patients’ quality of life. The results of this study can be used in other fields of medicine, as researchers can use advanced analytical tools, such as NLP, to discover social barriers to health care hidden in medical records. These types of questions are often not asked in structured questionnaires and instead show up in free-text notes, meaning that this information is often lost in documentation. If multiple areas of medicine start using the NLP analytical tools, other social barriers that restrict equitable care could be revealed and, therefore, changed to improve overall care to patients.