Background: A recent study found that 10% of emergency department (ED) patients experiencing homelessness had their housing status documented by the healthcare team. Identification of homelessness may inform care decisions and direct patients toward community resources, both of which may improve health outcomes. Standardized screening tools for homelessness exist but are not commonly used. We investigated patient characteristics associated with missed identification of homelessness to inform the development and implementation of effective screening strategies.
Methods: We used the Homeless Management Information System (HMIS), a database of information on individuals who have stayed at shelters or transitional housing in Washington, D.C. to identify homeless individuals. Full names and dates of birth from adults in HMIS between 9/1/2019-2/29/2020 were searched in our Electronic Medical Record (EMR) for matches. Inclusion criteria were having had at least 1 ED visit or inpatient admission during the 6-month period. Charts were reviewed for documentation of homelessness for up to 2 ED visits and 2 inpatient admissions per patient within the 6-month period. Instances of missed identification of homelessness were recorded. We used multivariable logistic regression to identify patient characteristics independently correlated with missed identification of homelessness, including age, gender, race, primary diagnosis, type of insurance, and prior or current use of shelters or transitional housing. We report adjusted Odds Ratios (aOR). This study was approved by the George Washington University IRB.
Results: Out of 5,025 adults in HMIS, 2,523 (50.2%) had hospital records, and 14% (702 patients) met inclusion criteria. Housing insecurity was correctly identified in 37% of ED visits (303/813 visits) and 64% of admissions (140/220 admissions). In the final multivariable model with visits nested within patients, significant independent predictors of missed homelessness during ED visits included prior or current use of transitional housing (aOR 1.92 [95% CI 1.07-3.44], p=.03), Black or African American race (aOR 2.13 [1.33-3.40] p=.0016) and having DC Alliance insurance (aOR 1.65 [1.14-2.40], p=.009). Homeless patients aged 60 or older were more likely to be correctly identified (aOR 0.40 [0.26-0.61], p<.0001), as were those with primary psychiatric diagnoses (aOR 0.24 [0.13-0.42], p<.0001). During hospital admissions, missed homelessness was significantly associated with having DC Alliance insurance (aOR 2.76 [1.29 – 5.93], p=.009); correct identification was associated with a primary psychiatric diagnosis (aOR 0.14 [0.04-0.49], p=.002).
Conclusions: Homelessness was not documented in a large percentage of patients during ED visits and hospital admissions. Characteristics associated with missed documentation were different for ED visits and admissions, but it is unclear why having DC Alliance insurance was associated with this for both. DC Alliance covers District residents < 200%FPL who are not eligible for Medicaid or Medicare. Use of transitional housing was associated with missed identification in the ED, which may reflect that healthcare teams are better at picking up housing insecurity when present in its most severe forms, highlighting the need to use standardized screening tools. Having a primary psychiatric diagnosis was associated with correct identification in both settings, which may indicate that components of the psychiatric interview include housing screening questions.