Homelessness in general is a major issue in the US today. The risk factors of homelessness are myriad, including inadequate income, lack of affordable housing, mental health and substance abuse issues, lack of social support, and nonadherence to treatment/follow-up appointments. Early identification of these factors from clinical documents may help detect or even predict homelessness cases, allowing adequate intervention and prevention measures.
Objective
We demonstrate a semi-automated approach to induce and curate lexical domain knowledge for identification of evidence and risk factors for homelessness found in VA clinical documents. This domain knowledge can be used to support training and evaluation of automated methods such as Natural Language Processing (NLP) systems for detection and prediction of homelessness among veterans. This could serve as a proxy for public health and other surveillance involving homeless individuals. Similar methods could be used to identify other conditions of interest.