Background: Approximately 40% of first-time adult shelter users have an Emergency Department (ED) visit or hospitalization within one year of shelter entry. Physicians must know patients’ housing statuses to develop appropriate care plans after discharge. Standardized tools for universal screening exist, however the rate of correct identification of homelessness appears to be low and is not widely studied. One barrier to study is the lack of “true” comparison information without surveying patients in real-time.We investigated the rate of correctly identified homelessness documented during ED visits and hospital admissions. We used the Homeless Management Information System (HMIS) to identify individuals who stayed at shelters or transitional housing in Washington, DC during a 6-month period and compared this to Electronic Medical Records (EMRs) during the same time period.
Methods: HMIS data on adults using homeless services between 9/1/2019-2/29/2020 was shared with researchers through a data use agreement. Using names and dates of birth, 5,025 individuals from the HMIS were searched in the EMR. Inclusion criteria were age >17, matching to a record in the EMR and having at least 1 ED visit or inpatient admission within the same 6-month period. Charts were reviewed for documentation of housing status for up to 2 ED visits and 2 inpatient admissions per matched patient, including where it was documented. Simple percentages were calculated and Fischer’s exact test was used to compare rates in the ED versus admissions. This study was approved by the George Washington University IRB.
Results: Out of 5,025 HMIS entries searched, 2,523 (50.2%) had EMR charts, with 702 patients (14%) having had at least 1 ED visit or admission in the 6-month study period. Out of 813 ED visits without admission (586 patients; 227 with second visits), housing insecurity was identified in 37% (303/813). Out of 220 admissions (188 patients; 32 second admissions), housing insecurity was identified in 64% (140/220). The difference in identification of housing insecurity during ED visits versus admissions was significant (p<.0001). Among ED patients whose housing insecurity was correctly documented, documentation was in the physician note (89%), social work note (22%) or nursing note (15%). Among admitted patients whose housing insecurity was correctly documented, documentation was in the physician note (72%), social work note (56%), discharge summary (54%), discharge note (46%), initial ED physician note (37%) or ICD-10 coded using Z59.0 (19%).
Conclusions: Our data show that homelessness was not documented in at least one-third of hospital admissions and two-thirds of ED visits for patients who used shelters or transitional housing. More may have been missed, as other patients may have had housing insecurity without utilizing shelters or transitional housing. Homelessness may be more often correctly documented during admissions due to the more thorough nature of the encounter, additional time available, better documentation, or the need to create more detailed discharge plans that incorporate housing barriers. Some EMRs incorporate screening alerts to improve identification of housing insecurity, however our EMR lacks these. Further study should include impact of using these tools on correctly identifying housing insecurity.