Developing artificial intelligence and machine learning to support primary care research and practice

This thesis was motivated by the potential to use "everyday data", especially that collected in electronic health records (EHRs) as part of healthcare delivery, to improve primary care for clients facing complex clinical and/or social situations. Artificial intelligence (AI) techniques can identify patterns or make predictions with these data, producing information to learn about and inform care delivery. Our first objective was to understand and critique the body of literature on AI and primary care. This was achieved through a scoping review wherein we found the field was at an early stage of maturity, primarily focused on clinical decision support for chronic conditions in high-income countries, with low levels of primary care involvement and model evaluation in real-world settings.\nOur second objective was to demonstrate how AI methods can be applied to problems in descriptive epidemiology. To achieve this, we collaborated with the Alliance for Healthier Communities, which provides team-based primary health care through Community Health Centres (CHCs) across Ontario to clients who experience barriers to regular care. We described sociodemographic, clinical, and healthcare use characteristics of their adult primary care population using EHR data from 2009-2019. We used both simple statistical and unsupervised learning techniques, applied with an epidemiological lens. In addition to substantive findings, we identified potential avenues for future learning initiatives, including the development of decision support tools, and methodological considerations therein.\nOur third objective was to advance interpretable AI methodology that is well-suited for heterogeneous data, and is applicable in clinical epidemiology as well as other settings. To achieve this, we developed a new hybrid feature- and similarity-based model for supervised learning. There are two versions, fit by convex optimization with a sparsity-inducing penalty on the kernel (similarity) portion of the model. We compared our hybrid models with solely feature- and similarity-based approaches using synthetic data and using CHC data to predict future loneliness or social isolation. We also proposed a new strategy for kernel construction with indicator-coded data.\nAltogether, this thesis progressed AI for primary care in general and for a particular health care organization, while making research contributions to epidemiology and to computer science.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.