Video Annotation in Nurse Home Visits

In home health care (HHC), video recordings can complement electronic health record (EHR) data by capturing essential details typically missed, such as environmental risk factors. This study developed a video annotation methodology for HHC by systematically annotating video data and mapping the resulting annotation codes to standardized clinical terminologies. Four simulated nurse home-visit videos were annotated and mapped to SNOMED CT and other vocabularies within the UMLS Metathesaurus. A total of 97 annotation codes were identified: 66 (68%) at the data level, mostly representing environmental observations (N=52, 78.8%). From these, 17 information-level (17.5%) and 14 knowledge-level (14.5%) codes were derived. Of these 97 annotation codes, 88.7% mapped exactly to SNOMED CT, 6.2% mapped partially to SNOMED CT but aligned better with other UMLS terminologies, 3.1% mapped only to non-SNOMED terminologies, and 2.1% could not be mapped. Among the 92 codes mapped to SNOMED CT, 88.0% were exact matches, 10.9% broader matches, and 1.1% close matches. This pilot study validates the feasibility and practicality of the proposed video annotation methodology in HHC. Future research could expand this methodology to diverse HHC settings and assess the added value of video-based information when integrating with EHR data in predicting patient outcomes.