Background: Homelessness is often underrecognized in the healthcare setting, preventing quality discharge planning, care coordination, and connection to community resources. Once found, effective documentation of homelessness in the electronic medical record (EMR) can inform future providers. We investigated how prior homeless documentation in the EMR influenced the identification of patient homelessness in subsequent encounters.
Methods: The Homeless Management Information System (HMIS) is a database of individuals that have stayed at shelters or transitional housing in Washington, DC. Full names and dates of birth for adults in HMIS from 9/1/2019-2/29/2020 were searched in our EMR to locate matching patient records. Inclusion criteria were having had at least 1 ED visit without admission or inpatient admission at our hospital during the 6-month period. Charts were reviewed for documentation of homelessness for up to the first two ED visits and first two inpatient admissions per patient during the 6-month period. We then examined the persistence of correct homeless identification among two consecutive encounters: ED visit to ED visit, ED visit to admission, admission to admission, or admission to ED visit. Univariable logistic regression was used to analyze the association between correct identification of homelessness at the first encounter with correct identification at the second. For second encounters, we examined the difference in identification rate between ED visits and admissions using Fischer’s exact test.
Results: Using the HMIS, we identified 702 patients experiencing homelessness; from their hospital records we reviewed 813 ED visits and 220 admissions. Among patients for whom homelessness was correctly identified during a sentinel ED visit, it was again identified in 63.86% of 83 ED visit second encounters compared to 90.91% of 22 admission second encounters (p=0.0177). There was a significant association between correct identification of homelessness in the sentinel ED visit and the ED visit second encounter (β=9.717, p< 0.0001) or admission second encounter (β=7.778, p=0.0208). Among patients for whom homelessness was correctly identified during a sentinel admission, it was again identified in 81.25% of 16 admission second encounters compared to 54.05% of 37 ED visit second encounters (p=0.0728). There was a significant association between the identification of homelessness in the sentinel admission and the admission second encounter (β=13, p=0.0318) or ED visit second encounter (β=12.94, p=0.0403).
Conclusions: These data suggest that homelessness is more often correctly documented during hospital admissions than ED visits. Correct identification in the EMR during a sentinel ED visit or admission was associated with correct identification during a subsequent encounter, however this was less often true when the subsequent encounter was in the ED. Whether homelessness was documented during a sentinel admission or sentinel ED visit, both resulted in successful identification in only about 60% of second ED visits. This suggests that ED providers are less likely to locate existing homeless status information in the EMR. This may reflect that our health system’s EMR has separate applications for ED and admission documentation, with inpatient providers more commonly accessing both applications. These findings highlight the need for effective screening and documentation of housing status in an EMR location easily accessed by all providers.