Article Text

Download PDFPDF

Homelessness: patterns of emergency department use and risk factors for re-presentation
  1. G Moore1,
  2. M F Gerdtz1,
  3. G Hepworth2,
  4. E Manias1
  1. 1Melbourne School of Health Sciences, Faculty of Medicine, Dentistry and Health Sciences, The University of Melbourne, Victoria, Australia
  2. 2Department of Mathematics and Statistics, Statistical Consulting Centre, The University of Melbourne, Victoria, Australia
  1. Correspondence to Dr Gaye Moore, Centre for Palliative Care Education and Research, PO Box 2900, 6 Gertrude St, Fitzroy, VIC 3065, Australia; dgmoore{at}


Objectives To describe patterns of service use and to predict risk factors for re-presentation to a metropolitan emergency department (ED) among people who are homeless.

Methods A retrospective cohort analysis was undertaken over a 24-month period from a principal referral hospital in Melbourne, Australia. All ED visits relating to people classified as homeless were included. A predictive model for risk of re-presentation was developed using logistic regression with random effects. Rates of re-presentation, defined as the total number of visits to the same ED within 28 days of discharge, were measured.

Results The study period was 1 January 2003 to 31 December 2004. The re-presentation rate for homeless people was 47.8% (3199/6689) of ED visits and 45.5% (725/1595) of the patients. The final predictive model included risk factors, which incorporated both hospital and community service use. Those characteristics that resulted in significantly increased odds of re-presentation were leaving hospital at own risk (OR 1.31; 95% CI 1.10 to 1.56), treatment in another hospital (OR 1.45, 95% CI 1.23 to 1.72) and being in receipt of community-based case management (OR 1.31, 95% CI 1.11 to 1.54) or pension (OR 1.34, 95% CI 1.12 to 1.62).

Conclusions The predictive model identified nine risk factors of re-presentation to the ED for people who are homeless. Early identification of these factors among homeless people may alert clinicians to the complexity of issues influencing an individual ED visit. This information can be used at admission and discharge by ensuring that homeless people have access to services commensurate with their health needs. Improved linkage between community and hospital services must be underscored by the capacity to provide safe and secure housing.

  • Emergency department
  • homeless
  • hospital re-presentation
  • risk screening
  • behavioural model for vulnerable populations
  • clinical assessment, effectiveness
  • education, assessment
  • mental health, mental illness
  • mental health, drug abuse
  • nursing, emergency departments

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Homeless people are socially isolated and often lack the skills required to navigate the healthcare system.1 Research shows that this group of people encounter multiple barriers to accessing healthcare.2 3 For homeless people, the emergency department (ED) may be the only place they can obtain treatment for their health problems,4 and many homeless people are frequent ED attenders.5 6 In Australia, being homeless is known to be a strong sociodemographic predictor for ED re-presentation.7

The Australian federal government has made a commitment to halve homelessness by 2020 and offers supported accommodation to those who need it.8 The government has a policy that no homeless person is to be discharged from a hospital facility into homelessness. It is important that greater attention is given to the hospital services provided to homeless people so that the environmental, health, behavioural and psychosocial needs of homeless people are addressed in an effective and efficient manner. Community services also need to be examined so that barriers to effective access can be rectified.

Community- and hospital-based case management (CM) programmes have been designed to meet the needs of frequent ED attenders, including people who are homeless.5 9 10 In Australia, CM is conceptualised as a social model of health that involves assessment, planning and advocacy to meet an individual's health needs.11 Research indicates that although multidisciplinary CM can improve access to housing support and other community-based health services, these programmes may cause an initial increase in the frequency of ED use.5 Furthermore, the success of these programmes relies on the availability of safe and secure housing.12

Although use of CM is linked to increased ED visits, it also indicates that a high level of health need and is not necessarily the reason for an ED re-presentation. Further investigation is required to understand the complexity of hospital and community service use, which impacts on ED re-presentation rates. Accurate identification of risk factors for re-presentation to the ED among people who are homeless is required to initiate early referral to hospital- and community-based services. Describing the patterns of health service use among homeless people provides an essential context for understanding their current ED use. This information may be used to inform the development of interventions that take into account the role of community and hospital resources in providing healthcare delivery to this vulnerable group of ED service users.


A retrospective cohort analysis was conducted to identify the patient and visit characteristics of all presentations to one principal referral hospital ED located in metropolitan Melbourne, Australia, excluding those who died, for the period 1 January 2003 to 31 December 2004. This ED has an annual census of approximately 32 000 presentations per annum.

Data sources

At the study site, a computerised patient administration system (PAS) facilitated reporting to the Victorian Emergency Minimum Data Set. This was a patient tracking system that contains real-time data for all ED presentations including their demographic profile, clinical information, admission and discharge times and disposition. For each patient in the PAS, there was a unit record number and an ED visit number. The PAS facilitated analysis of administrative and clinical data at both visit and patient levels.

The PAS was used to identify all visits to the ED by people who were homeless. For each ED presentation, information about housing status was captured by the variable “usual accommodation”. To ensure the accuracy of this variable, we examined the medical records of all patients identified as homeless to confirm and identify the level of homelessness. A file audit was conducted to further characterise the clinical and sociodemographic profiles of homeless people and describe their ED and community-based service use. The file audit was conducted by three abstractors. Training was conducted over 2 days, and independent review of abstractor accuracy was performed daily. One abstractor conducted 80% of the data collection. All the abstracters were blinded to the etiologic relationships being investigated in the study.7 Abstractors used defined options to identify categories within key variables to reduce coding error. Inferential integrity was maintained within the database to ensure accuracy of data entry.

Cohort selection

All the patients who attended the ED during the 2-year study period were included, except those where death was recorded as the discharge outcome. The term re-presentation was defined as attendance at the same ED within 28 days after the discharge date either from the ED or as an inpatient. For presentations occurring in December 2004 that had resulted in a subsequent re-presentation in January 2005, these were identified and included in the analysis.


The behavioural model for vulnerable populations (BMVP) was used to provide a theoretical framework for understanding the complex range of factors associated with a person's use of healthcare services.13 The prototype for the BMVP was first applied in 196812 to explore the factors that are associated with consumer use of healthcare services and has been adapted13 to focus on the socially disadvantaged populations. The model was used in this study to provide understanding of the variables contributing to re-presentation to the ED and to generate new knowledge on the complexity of issues affecting homeless people with respect to their healthcare needs. The model has three operational perspectives: health status, service use and socioeconomic status. These perspectives incorporate the reasons for access to the ED, the use of the ED and other health services and the social and financial background that underpins a person's health status. We have adapted the model slightly to take into consideration the environmental influences relevant to homeless people. Figure 1 is a flow chart showing how the predictive model includes all aspects of the BMVP. The nine variables identified as predictive of re-presentation to the ED are representative of the four domains derived from the BMVP: environmental influences, functional health status, health status and health behaviour.

Figure 1

Behavioural model for vulnerable populations with logistic regression model.

We used Chamberlain and Mackenzie's classification of housing status to define homelessness and further organised it according to four subgroups.14 Primary homelessness includes people living on the streets or in squats. Secondary homelessness includes people living in crisis accommodation. Tertiary homelessness includes people living in boarding houses. The fourth group includes people who reside in other types of unstable housing.

The study was approved by the human research ethics committee of the health service in which the research was conducted. The protocol was subsequently registered as a project with external approval by the collaborating university's human research ethics committee.

Data analysis

ED visits by people identified as homeless (6689/64 177) were aggregated into a patient dataset (n=1595) that contained the percentage of attendances resulting in re-presentation for each patient. Rigorous data cleaning and verification of information was performed before data analysis. Missing information was either located or listed as unknown. A process of cross-checking was conducted to ensure accuracy of information and to provide a means of validating the data.

There were unequal numbers of presentations per patient (range 1–152). To account for the correlation between visits by the same patient, a generalised linear mixed model was used, with representation (yes/no) as the binary response variable and the patient modelled as a random effect. This model is also known as hierarchical logistic regression or a logistic regression with random effects. Such a model uses all the data to estimate the correlation between visits within a patient and gives an appropriate effective weighting to visits and patients. If all patients had only one visit, the model would simplify to standard logistic regression. The fitting of the model was performed using GenStat (8th edition; VSN International, Hemel Hempstead, UK). A generalised linear mixed model was fitted for each potential explanatory variable on its own. Variables for which the p value for testing the hypothesis of no effect was >0.05 were excluded from further model fitting. All the remaining explanatory variables were entered into a backwards stepwise procedure until all variables still in the model were significant at the 0.05 level.

The estimated parameters and the standard errors from the final model were used to calculate adjusted odds ratio (OR) and associated 95% confidence intervals (CIs) for each explanatory variable. There were nine such variables, eight of which related to patient characteristics and one to ED visit characteristics. For explanatory variables not in the final model, unadjusted ORs and 95% CIs were calculated. Significant two-factor interactions among these nine variables were then added to the model, but these interactions did not improve prediction, so this more complicated model was not considered further. An analysis was also conducted on all presentations to the ED (n=64 177), so that a comparison between the non-homeless population and the homeless population could be made.


ED presentation characteristics

From 1 January 2003 to 31 December 2004, there were 64 177 presentations by 40 942 individuals to the ED. For all ED presentations, 4881/64 147 (7.6%) were classified as Australian Triage Scale (ATS) category 5 (non-urgent or clinic administrative), whereas for the homeless group 987/6689 (14.8%) were classified as ATS category 5 (table 1).

Table 1

ED visit characteristics by housing status (n=40 942)

The re-presentation rate was 18.0% (11 559/64 177) for all the visits and 14.4% (5894/40 942) for all the patients. The homeless population accounted for 10.4% (6689/64 177) of all the visits and 3.9% (1595/40 942) of all the patients. The re-presentation rate for the homeless was 47.8% (3199/6689) of ED visits and 45.5% (725/1595) of the patients.

Patient demographics

The number of presentations where people were homeless ranged from 1 to 152 per patient with a mean (SD) of 4.2 (7.0) visits per patient over the 2-year period. Males constituted 74.4% (n=1186) of all homeless people who presented. Of whom, 71.8% (n=1145) were single, 63.8% (n=1017) were receiving a pension and 6.6% (n=106) were indigenous. The incidence of mental illness in the study cohort was 54% (n=861), and for substance misuse, the incidence was 70.5% (n=1124). Presentation to the ED by homeless people was not influenced by the seasons, which showed an even distribution of visits throughout the year. Table 2 shows more demographic and clinical characteristics of the homeless people.

Table 2

Demographic and clinical characteristics of all homeless patients (grouped according to the behavioural model) who presented to the ED from 1 January 2003 to 31 December 2004 (n=1595)

Community and hospital services

Table 3 shows the frequency of use of hospital-based services by the homeless people according to levels of homelessness. The homeless people who resided in secondary, tertiary and unstable housing tended to be more frequent users of hospital-based services than the homeless people living on the streets. Hospital services were identified from the chart audit and included hospital in the home service that entailed hospital nurses giving care to inpatients in the home as an alternative to hospital care, care coordination by a multidisciplinary team, outpatient services, a hospital in the hospital specialised service that supplies a home environment to enable nurses to deliver care safely, interpreting service and the ED. Table 4 indicates the level of support from community-based services by level of homelessness. Similarly, there was an increase in community-based service use for people residing in secondary, tertiary and unstable housing when compared with the homeless people living on the streets. Community services included a broad range of support from general practitioners to family/friend carer. Hospital and community services were accessed by 67.5% (n=1076) of the homeless people, which indicates high service use in conjunction with a high re-presentation rate.

Table 3

Frequency of hospital-based services use by those people classified as homeless according to levels of homelessness (n=1595)

Table 4

Level of support from community-based services by level of homelessness (n=1595)

Risk factors influencing re-presentation

Patient variables that significantly influenced re-presentation of patients to the ED are shown in table 5. Although the discharge outcome, a residential care facility, was associated with substantially increased odds of re-presentation (adjusted OR 3.08; 95% CI 1.33 to 7.12), the CI is wide and the numbers were small (n=32). Presenting to another hospital within 12 months was associated with significantly increased odds of re-presentation (adjusted OR 1.45; 95% CI 1.23 to 1.72). Pensioners, individuals aged <65 years, individuals having a documented next of kin, identified drug misuse, individuals receiving community CM and individuals leaving at their own risk were all significantly associated with increased odds of re-presentation to the ED.

Table 5

Percentage of homeless visits resulting in re-presentation to the ED and adjusted ORs for the final model by significant patient and ED visit characteristics

Although medication non-compliance and dual diagnosis of mental illness and substance or alcohol misuse (unadjusted OR 1.54, 95% CI 1.30 to 1.83 and unadjusted OR 1.54, 95% CI 1.30 to 1.84, respectively) were found to be significant variables influencing re-presentation on their own, when fitted into the full model, they were no longer significant. Length of stay (LOS) for those admitted as inpatients and for the ED were not significant factors influencing the outcome (table 1). The level of homelessness was not significantly associated with repeated self-referrals to the ED.


This study showed that homeless people frequently used many hospital and community services. The nature and extent of this use have not been described for the homeless population. Despite this use, homeless people still featured predominantly in ED re-presentation rates. The results also demonstrate that homeless people experience enormous complexity of healthcare needs and socioeconomic vulnerability, which require extensive resources at hospital and community levels of service delivery.

We observed an increase in the use of hospital- and community-based services for people residing in a secondary, tertiary or unstable level of housing when compared with people living on the streets. In fact, almost double of the number of homeless people classified as residing in unstable types of housing used these services. We also noted that although there was a considerable difference in service use between people classified into the different levels of homelessness, there was no significant difference in their risk of ED re-presentation between the levels.7 These findings suggest that people who are living on the streets may experience more structural barriers to accessing these services compared with homeless people in other forms of accommodation. Homeless people in other forms of accommodation may have access to resources that provide opportunities for them to use hospital- and community-based services. Homeless people who live on the streets have been shown to struggle to manage basic needs, such as obtaining food, clothing and shelter.15

In the current study, over half of the homeless patient visits (59.0%, n=3949) occurred after business hours, therefore, increasing the difficulty of obtaining referral for accommodation and CM because community services were unavailable overnight. Other barriers may include appointment scheduling and criteria that restrict access based on presenting documentation such as health benefit cards that may have been lost or be out of date.16–18 In addition, lack of transportation, disjointed involvement of multiple services and lack of access to affordable medications may also contribute to homeless people preferring ED services. These findings highlight the need to develop clear pathways from the ED to health services that have the capacity to respond to this complexity and socioeconomic vulnerability. Additional research is required to identify the nature of structural barriers experienced by homeless people living on the streets and to develop pathways that will assist them to access appropriate health services to meet their needs.

This study highlights the complex nature of ED visits involving homeless people, and this complexity contributes to the increased risk of re-presentation to the ED among homeless people. The percentage of those in ATS category 5 who were classified as homeless was almost double that of non-homeless people (15% vs 7.6%). Further analysis of the data identified nine significant variables that describe the physical and social complexities of ED visits in this cohort. The characteristics of the ED environment, which is driven by the need to rapidly assess and manage urgent health problems is possibly not well suited to addressing the complicated nature of health concerns of homeless people. Mapping service use by homeless people can provide helpful information about the barriers to obtaining care outside the ED and points to the need to develop pathways to optimise health service access.

The ED remains a critical place for engaging with homeless people and offers an opportunity to support this group. This study used the ED as the initial contact and identification of a homeless person and then their risk of re-presentation. The outcome measure, re-presentation, identified the psychosocial vulnerability at point of entry into the hospital, with the aim to inform referral and discharge planning. Although hospital demand strategies have tried to address the demand of services by people who are homeless, little is understood in respect to the complex nature of their service use. Past research has focused on the ways homeless people use community and hospital services in isolation to each other.5 9 19–21 Future research needs to investigate the way homeless people access healthcare across a range of services to understand their specific healthcare issues.

The BMVPs through its four domains helped to shape the variables examined for their potential impact on re-representation in the ED and homeless people's access of hospital and community services (figure 1). The model was able to delineate the complexities associated with the characteristics of homeless people who re-presented to the ED. In terms of functional health status, homeless people tended to be cognitively impaired, to experience medication non-adherence, to have drug and alcohol problems and to have a dual diagnosis of a mental disorder and a problem associated with substance or alcohol abuse. Their health status demonstrated that they had many chronic illnesses. In terms of their health behaviour, they regularly accessed the ED compared with people who were not homeless and were heavy users of hospital and community services. Despite this large demand in service use, this group comprises a large proportion of people who re-present to the ED, which demonstrates that they need further increases in resources to reverse re-presentation trends.

Strengths and limitations

The study involved analysis of clinical audit data for presentations to the targeted hospital, and it was not possible to identify those people who may have presented to other hospitals after attending the study hospital. People who died as a discharge outcome were excluded from the study, but the data on subsequent deaths occurring after discharge were unreliable because of incomplete data. These individuals were not known and were, therefore, not excluded.

Although this study was conducted in one principal referral hospital, the work contributes broadly to our understanding of how homeless people use EDs and community services. There is a need to further extend this study to review other EDs in Australia and internationally to identify factors influencing the use of EDs and hospital and community services by homeless people.


The ED is a window of opportunity where early intervention strategies to improve the health status of homeless people may be implemented. Identification of risk factors for re-presentation to the ED will facilitate screening and early referral to hospital- and community-based services. Re-presentation to the ED highlights the complex and unresolved health needs for people who are homeless. The predictive model will enable a screening for these complex health issues to focus resources on this very vulnerable population.


The authors acknowledge the financial assistance by the Australian Research Council and the contribution of Dr Andrew Dent, previous Director of Emergency Department, St Vincent's Hospital.



  • Funding Financial support was provided through the Australian Research Council Linkage Project Scheme, project number LP0453587. Financial support was also provided by St Vincent's Hospital.

  • Competing interests None.

  • Ethics approval This study was conducted with the approval of St Vincent's Hospital, Melbourne, Australia.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Linked Articles

  • Primary survey
    Darren Walter