Illnesses Associated With Increased Length of Stay for Individuals Experiencing Homelessness: A Retrospective Cohort Study of Emergency Department Visits and Hospitalizations.

Background: Individuals experiencing homelessness (IEH) tend to have increased length of stay (LOS) in acute care settings, which negatively impacts health care costs and resource utilization. It is unclear however, what specic factors account for this increased LOS. This study attempts to dene which diagnoses most impact LOS for IEH and if there are differences based on their demographics. Methods: A retrospective cohort study was conducted looking at ICD-10 diagnosis codes and LOS for patients identied as IEH seen in Emergency Departments (ED) and also for those admitted to. Data were stratied based on diagnosis, gender and age. Statistical analysis was conducted to determine which ICD-10 diagnoses were signicantly associated with increased ED and inpatient LOS for IEH compared to housed individuals. Results: Homelessness admissions were associated with increased LOS regardless of gender or age group. The absolute mean difference of LOS between IEH and housed individuals was 1.62 hours [95% CI 1.49 – 1.75] in the ED and 3.02 days [95% CI 2.42-3.62] for inpatients. Males age 18-24 years spent on average 7.12 more days in hospital, and females aged 25-34 spent 7.32 more days in hospital compared to their housed counterparts. Thirty-one diagnoses were associated with increased LOS in EDs for IEH compared to their housed counterparts; maternity concerns and coronary artery disease were associated with signicantly increased inpatient LOS. Conclusion: Homelessness signicantly increases the LOS of individuals within both ED and inpatient settings. We have identied numerous diagnoses that are associated with increased LOS in IE; these inform the prioritization and development of targeted interventions to improve the health of IEH.


Background
Homelessness has signi cant implications for health. Individuals experiencing homelessness (IEH) present with increased disease complications compared to their housed counterparts due to factors such as poor diet, inability to afford medications, and irregular access to primary care [1]. IEH are at higher risk for uncontrolled chronic diseases and have higher prevalence of mental health concerns, addictions, and infectious diseases like Human Immunode ciency Virus and Hepatitis C [2] [3]. IEH also use acute care services more frequently, with 71% using Emergency Department (ED) services over a 6 month period [4] [5]. They also have increased rates of hospitalizations, due not just to acute medical needs but also because of their complex social barriers [6] [7]. Despite increased rates of hospitalizations, the effectiveness of IEH's hospital stay is diminished by their return to homelessness [8]. Long-term outcomes remain poor for IEH, with mortality rates being up to 9 times higher than the average population [9] [10].
Not only are IEH hospitalized more often than housed individuals, they also have longer lengths of stay (LOS) in hospital, which presents an additional burden on both the individual and on society. In general, IEH spend between 2.3 and 4.1 more days in hospital [11] [12] [13] than housed individuals, however this data is aggregate data not identifying which disease types most contribute to this LOS. Hwang et al. noted that in Canada, even after adjustment for age, gender, and resource intensity weight, hospitalizations for IEH cost over $2500 more than for housed individuals [12]. Concomitant mental health concerns are common in IEH and impact both rates of hospital admissions, LOS, and hospitalization costs [14], with admissions to psychiatry costing over $1000 more for IEH (even while adjusting for length of stay) [12].
Aside from the presence of mental health conditions, it remains unclear whether speci c medical conditions are more likely to lead to an increased LOS in IEH compared to their housed counterparts, information which is crucial to determine targeted interventions to avoid increased LOS. It is often believed that IEH may require longer stays in hospital due to lack of community supports; for example, IEH may spend increased time in acute care setting to receive antibiotics for endocarditis and osteomyelitis due to lack of medication coverage or outpatient parenteral options [15]. In contrast to hospital LOS, there is variable evidence on the impact of homeless on LOS for emergency department visits. One study showed increased ED LOS for IEH [16], while another study speci cally examining patients with mental health concerns noted no differences in ED LOS in homeless versus housed individuals [17]. Most studies of ED utilization do not speci cally compare LOS, or the differences in ED diagnoses, between IEH and housed individuals [18] [19]. A better understanding of how medical diagnoses and needs impact the increased use of both inpatient and ED care can help to identify gaps in our current inpatient and outpatient health service delivery, to ultimately inform the development of interventions to address these speci c illness-speci c needs [20].
In this cohort study, we aimed to determine the most responsible diagnoses that most affect ED and inpatient LOS among IEH. To our knowledge, this study is the rst to assess diagnosis-modifying LOS in this population. This study will deepen our understanding on health issues faced by IEH, and to prioritize and determine disease management strategies in the community.

Study design and population
We designed a retrospective cohort study examining IEH who accessed four tertiary care hospitals and one urgent care center in Calgary, Canada, from April 1, 2017 to March 31, 2018. The exposure was homelessness (operational de nition described below). Ethnicity is not recorded within the health region's dataset. If an individual was seen or admitted multiple times, only the rst interaction with acute care (ED or inpatient) was utilized to simplify our analysis. The primary outcomes were differences in LOS between IEH and housed individuals within both ED and inpatient settings, with the speci c aim of examining which diagnoses and demographics contributed to these differences.

Data Sources
Administrative data for individuals aged 18 years or greater was provided by the provincial health authority, and consisted of the Discharge Abstract Database (for hospitalizations) and National Ambulatory Care Reporting System database (for ED visits). A unique ICD-10 code (International Classi cation of Diseases, 10th Revision ) for homelessness was utilized to identify IEH. As this code was only available from 2017 onwards, and as the code for homelessness is optional, additional information regarding homelessness was obtained through individual addresses. If the address was labeled as "no xed address" or the postal code provided corresponded to one of Alberta's homeless shelters, the patient was included into the study cohort. Because Calgary homeless shelters tend to be located in the downtown business districts where there are few residential developments, very few regular residences share similar postal codes with the shelters. This method of identifying cohorts of IEH has been used in previous studies [7,8]. Due to limitations in identifying homelessness using administrative databases, individuals who are precariously housed, those with low-income housing, and those who were couch sur ng were classi ed as being housed.
IEH were excluded if their health care identi cation was missing. If a particular ICD-10 category had fewer than 5 individuals with that diagnosis, they were also omitted due to di culties with measuring statistical signi cance. Additionally, signi cant data outliers de ned as individuals who were admitted to hospital for 90 days or more were excluded. The rationale for this latter exclusion criterion is that the extended lengths of stay for these individuals likely represent alternate level of care days [21], where they are medically stable with their admission being no longer attributable to acute care management.

Statistical Analysis
Once IEH were identi ed in the database, ICD-10 codes were used for the classi cation of the primary diagnosis these clients were evaluated for in acute care. ICD-10 code subcategories were combined for most diagnoses. If an individual diagnoses had a small sample size less than 5, they were omitted from analysis. The mean LOS for IEH and housed individuals (HI) within the ED and inpatient settings were calculated separately, and mean differences between these two were calculated by subtracting the mean LOS for IEH by the mean LOS for HI. Mean differences in LOS were adjusted for age group, sex as well as diagnosis based on principal diagnosis using analysis of covariance to limit confounding effects [22] [23]. 95% con dence intervals (CI) for the difference in mean LOS were determined using the Student t statistic. Statistical Analysis Software (SAS) version 9.3 was used to perform all statistical analyses.
The study received ethics approval at the University of Alberta research ethics board as well as internal research review by Alberta Health Services.

Results
A total of 3,620 unique IEH accessed an acute care facility in 2017-18; 858 of these individuals were admitted to the hospital. Our comparator group consisted of 375,271 housed individuals.
Characteristics of IEH and differences in LOS in acute care settings The majority of individuals assessed in both the ED and inpatient settings were male across all age groups ( Table 1). The mean age of IEH using acute care resources was 38.6 years, and 48.5 years for housed individuals. Most IEH who used acute services were between the ages of 25-49, whereas age ranges were more evenly spread for housed individuals. Among IEH who accessed acute care, the average number of ED visits over one year was 3.19 compared to 1.28 visits for housed individuals. On average, homeless and housed individuals had a similar number of hospital admissions over the 1 year period, however, the LOS for IEH patients was over 1.5 fold that for housed individuals (10.27 vs. 6.70 days, p < .0001).  Table 2 shows the adjusted mean LOS, strati ed by age and sex. There were 2.84 times more males seen in the ED and 2.97 times more males admitted to hospital than females, which re ects the Calgary homeless population baseline demographic of 2.77 times more males than females [24]. IEH in almost all age groups spent statistically signi cantly more time in both the ED and in hospital than housed individuals.

Discussion
To our knowledge, this is the rst study that explores differences in LOS between IEH and housed individuals in both the ED and inpatient setting. It is also the rst study to explore how these differences vary by demographics and the principle diagnoses that prompted acute care use. In keeping with previous studies, we found that a disproportionately high prevalence of IEH in the ED and admitted to hospital were men between the ages of 25 and 34, where the majority of individuals experiencing homelessness in the community are between the ages of 45-64 [24]. Most were admitted to hospital for substance-related concerns. On average, IEH spent 1.62 more hours in the ED and 3.02 more days in hospital than housed individuals. This average increase in inpatient LOS for IEH is slightly lower than a previous study from New York (where LOS for IEH was 4.1 more days than for housed individuals) but is in keeping with a Canadian study which found a mean difference of 2.32 days in LOS between IEH and those who are housed [11][12].
The trend towards increased ED and hospital LOS was consistent across many diagnoses, many of which have little physiologic or clinical overlap. This suggests that the increased LOS in IEH may be more attributable to their underlying state of homelessness rather than factors related to the particular medical diagnoses. There were surprisingly two diagnoses that were associated with increased LOS for HI compared to IEH including depression, and blood alcohol and drug tests. These differences appear small and of questionable clinical signi cance.
By exploring the differences in ED LOS, our study highlights potential points of intervention to optimize ED work ow and bed occupancy. Within ED, IEH contribute to reduced work ow, leading to ED crowding which has been associated with decreased quality of care, delays in treatment commencement, and increased mortality [25]. The majority of the primary diagnoses that are most frequently seen in the ED in the homeless population might be managed in an outpatient setting, though we are limited by the lack of severity data captured in administrative databases. For example, both cellulitis and epilepsy were associated with an increased ED LOS for IEH. These are ambulatory care sensitive conditions (ACSC) [26] where acute care use might be avoided if they able to be optimally managed in the outpatient setting. A large number of diagnoses associated with increased ED LOS are related to mental health and addiction concerns. Further investment into community based mental health and addictions resources may be warranted.
When looking at diagnosis associated with increased LOS within inpatients; coronary atherosclerotic disease (CAD) is associated with almost 25 more days spent admitted to hospital compared to housed individuals. While CAD is not classi ed as an ACSC, risk factors for CAD such as hypertension and diabetes, as well as their consequences such as angina and heart failure are included in ACSC.
Furthermore, smoking is a known risk factor for CAD. There is a very high prevalence of smoking amongst IEH (57%) compared to housed individuals (27%) [27]. Our results highlight the need for interventions targeting CAD and their risk factors in IEH, such as focusing on resources for smoking cessation, hypertension, and diabetes. Maternal concerns associated with homelessness also demonstrated increased LOS with 6.45 more days spent in hospital, again demonstrating speci c needs for community prenatal and fetal-maternal care for IEH.
The strengths of our study include detailed hospitalization and ED data collected from multiple acute care facilities. Furthermore, our ndings are in keeping with prior evidence, suggesting that they are generalizable. The demographics of the Calgary population experiencing homelessness has also been demonstrated to be similar to IEH across the country [28][29].
There are limitations to our study. Due to our cohort identi cation methods, a small number of IEH were excluded from our study. For example, individuals without identi cation were excluded, though this comprised only 1/30 of our sample size. Furthermore, as we could not identify individuals who were precariously housed such as those who were couch sur ng, our cohort likely represents individuals more severe or chronic homelessness. Another limitation was our inability to account for illness severity, despite matching IEH and controls based on demographics and primary diagnoses, as this information is not captured within the administrative databases. That is, IEH may present to acute care facilities with similar diagnoses as housed individuals, but at a later stage and/or with increased severity, which could also explain their increased lengths of stay.

Conclusion
Homelessness is associated with increased ED and hospital inpatient LOS.

Declarations
Ethics approval was obtained for this study. As this was a large data study, consent to participate was not deemed necessary from ethics.