Abstract
Purpose :
Unstructured text data in electronic medical records and online patient fora represents a rich, underutilised resource to advance patient-centred care. Computational ontologies permit structured annotation of unstructured data to facilitate analysis.
Methods :
We extracted unstructured text from online patient fora hosted by oliviasvision.org and uveitis.org, divided into training and test sets. We built an ontology for ocular immune mediated inflammatory diseases (IMIDs), ocIMIDo, in Protégé. The foundation combined clinical guidelines, domain expertise, relevant classes from 5 ontologies (ORDO, HPO, UBERON, PATO, DOID), and text-mined and manually curated patient-preferred synonyms identified in the training data, using a tf-idf machine learning approach, with validation in the test set. OcIMIDo includes classes relating to anatomy, clinical phenotype, disease activity status, complications, investigations, interventions and functional impacts. We annotated forum data using ocIMIDo and Python, and performed natural language sentiment analysis on each post.
Results :
OcIMIDo contains 612 classes, 1563 relationships, 2602 annotations, 946 cross-references, and 180 synonyms from oliviasvision.org. We analysed 9031 forum posts including 1903 unique threads and 7128 replies. OcIMIDo annotation revealed frequent mention of different ocular IMIDs, treatments and complications. Sentiment analysis of posts was generally positive (median 0.12, IQR 0.01-0.24), but ranged from very negative (-1.00) to very positive (+1.00). In multivariable logistic regression, the odds of a post expressing positive sentiment were significantly reduced in first posts as compared to replies (OR 0.72, 95%CI 0.63-0.83,p<0.001), and in posts mentioning anterior uveitis (OR 0.80, 95%CI 0.65-0.98, p=0.032), steroid therapy (OR 0.79, 95%CI 0.65-0.96, p=0.018) or biologic therapy (OR 0.82, 95%CI 0.68-0.98, p=0.027), but increased in posts mentioning more severe subtypes of inflammatory eye disease (OR 1.58, 95%CI 1.17-2.12, p=0.003).
Conclusions :
We report development of a novel, open access ontology for ocular IMID which can be used to annotate unstructured big data. Sentiment analysis of online posts, whilst relatively insensitive, provided new insights into the impact of ocular IMIDs and their treatment on patients.
This is a 2020 ARVO Annual Meeting abstract.