The goal of this project is to develop predictive models to identify Veterans at risk of homelessness. This is a priority area and there is a commitment by VA leadership to end Veteran homelessness by 2015. To fulfill this commitment, it is necessary to risk-stratify Veterans and offer preventive services in a timely manner. Existing methods for risk stratification are based solely on administrative data. Those 'at-risk' of homelessness, especially for the first time, are a major focus for prevention efforts. Early warning indicators are currentl only inferred from known risk factors for homelessness that can be gleaned from administrative data. References to early warning indicators of risk in the free text of clinical narratives writte by VA providers may precede the formal identification of Veterans as being homeless. A new screening questionnaire (clinical reminder) is now in phased implementation in the VA to ask each Veteran about current homeless status and concern for short-term future homelessness. There is a need to develop robust algorithms to predict those at risk of homelessness, to assist in validating the results of the clinical reminder and determine whether Veterans are receiving appropriate services (needs-service match). It is also important to address gaps in our understanding of new homelessness among OEF/OIF/OND Veterans, in whom the risk factors for homelessness are not fully elucidated, including females and those with disabilities. Under the hypothesis that using text (unstructured data) from the clinical narrative will support and complement available administrative (structured) data in the risk stratification of Veterans for homelessness, the specific aims of this proposal are to (1) Use text data to improve the accuracy of determination of homelessness status of Veterans (2) Develop and apply predictive models for homelessness and homelessness outcomes in Veterans. This project will build on the informatics methods that the PI, co-investigators, and research team have developed under current VA HSR&D funding and used across multiple research projects to extract information from the free text contained in clinical narratives. The project will use VA Informatics and Computing Infrastructure (VINCI) as a platform to access national VA data from the Central Data Warehouse (CDW) including clinical text notes, mental health notes, primary care notes, social work notes and homeless health care notes from cohorts of Veterans. Using reference standard data sets of known homeless Veterans and test sets of Veterans whose homelessness status is unknown, the electronic algorithms will be evaluated for their performance in the risk- stratification Veterans for homelessness. In collaboration with our local station VA homeless program coordinator and national partners such as National Center on Homelessness Among Veterans, the research team will perform a pilot prospective validation of these algorithms to support prevention of homelessness among Veterans. The validated electronic algorithms will also be used to perform epidemiologic research on longitudinal cohorts of Veterans to improve our understanding of homelessness among emerging OEF/OIF/OND Veterans.
Homelessness is both a cause and consequence of ill health. The VA has committed to ending homelessness among Veterans by 2015. To fulfill this commitment, the VA will have to provide appropriate and timely services and resources to manage and prevent homelessness. This will require a coordinated effort by all stakeholders to reliably identify Veterans who are currently homeless, as well as those at risk and match services to their respective needs. This proposal aims to build prediction models to identify Veterans at risk of homelessness and connect them to needed services. The project will extract relevant information from available structured data and the free text of VA electronic medical records to develop automated algorithms to identify those at risk. The implementation of validated research results will positively impact the housing status of Veterans and thereby contribute to improved health outcomes.
