Development and validation of a novel Quality of Life ontology (QoLo), used with natural language processing and online questionnaires to lower existing barriers to developing Patient Reported Outcome Measures (PROMs)

Abstract

Purpose : Traditional PROM item identification entails qualitative research and thematic content analysis (e.g. NVivo), typically of single-centre focus group transcripts. The resources and expertise required create barriers to PROM development, and the PROM may lack external generalisability. Our study aimed to a) develop a new ontology for ophthalmic Quality of Life(QoL) to enable annotation of unstructured text data; b) Explore online questionnaires as an alternative text data source; and c) to validate these approaches to PROM item identification, through comparison to a ‘Gold Standard’ (focus group/interview transcript data, n=41), taking uveitis as our exemplar.

Methods : We developed QoLo(v1.0) using Protégé, using items in 11 QoL domains from 6 Rasch-validated, published ophthalmic PROMs (Table 1). We recruited participants via 10 uveitis charities and 7 Facebook groups, to complete online questionnaires (Qualtrics). These explored impacts of uveitis and treatment, by QoL domain. We text-mined this dataset to extend QoLo, adding new uveitis-relevant items, and patient-preferred synonyms, using a natural language processing approach and statistical ranking. We validated this approach versus the Gold Standard.

Results : QoLo(v1.0) contained 765 unique classes within 11 QoL domains (Table 1, Figure 1). Participants(n=58) resided in 5 continents(81% UK-based), 81% were female, with mean age 55(sd 11.2, range 31-80)years. They reported anterior(47%), intermediate(9%), and posterior/pan-uveitis(44%), 48% were undifferentiated. 69% had received systemic immunosuppression and 22% intravitreal implants. Two authors reviewed 1641 text items identified through statistical ranking (taking <60mins), selecting new classes and synonyms. The extended QoLo annotated 504/796(63%) classes in the questionnaire data, and 399/796(50%) in the transcript data, revealing significant overlap in QoL items across eye diseases.

Conclusions : QoLo, a novel ontology, is hosted and open-access on GitHub, and offers an accelerated, low-cost approach for annotating text data. We illustrate application to identify new uveitis PROM content from online questionnaires, helping address key barriers and generalisability issues for PROM development across ophthalmology.

This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.

 

Table 1

 

Figure 1 QoL domains contained in the ontology