Identifying causes of death among people experiencing homelessness: Definitions and data-sources matter

Abstract

Objectives
This presentation reports analysis from a linked data study into underlying causes of death among people with experience of homelessness in Wales. Findings are compared with similar UK based studies to highlight the potential distortions in evidence relating to homelessness arising from differing research designs.

Methods
Data were accessed via the Secure Anonymised Information Linkage Databank. Records of people who applied for assistance from a local authority homeless service in Wales (January 2012 to March 2017) were linked to death records (up to January 2020). The primary outcome was underlying cause of death using the ICD-10 classification system. A literature search was conducted to identify UK based studies on cause of death among people defined as homeless which used administrative data. The underlying cause of death for people who sought assistance from the local authority service was re-categorised to allow comparison with identified studies.

Results
232 people who had used the local authority homeless service in the past were found to have died. The leading underlying cause of death was accidents (33%), followed by cancer (13%) and heart disease (13%). The literature search identified 4 UK based studies meeting the search criteria, and whose findings were published at a suitable level of detail to enable comparison. Underlying cause of death varied across studies, particularly deaths related to drugs and alcohol, with greater external causes found in studies focusing mainly on rough sleepers and people using shelters. Differences in data sources and definitions of homelessness are explored in relation to findings across studies.

Conclusion
The way researchers define and identify people experiencing homelessness can significantly impact the outcomes of studies on this population. Clearly specifying which experiences of homelessness are included in (linked) administrative data studies may help address the distorted framing that arises when research focuses only on extreme cases.