Vulnerability index

In the context of homelessness, the first vulnerability index was developed in 2007 by Common Ground, a homelessness service in New York City, for the purpose of assessing medical vulnerability and to prioritise housing for people experiencing homelessness [12]. Based on the findings of earlier research [10], this original vulnerability index provides a framework to prioritise housing need by identifying the most medically vulnerable people experiencing homelessness through a standardised assessment, which quantifies the individual’s risk for mortality. It comprises nine criteria, including length of homelessness (days, months, years), number of hospital admissions and emergency department admissions in a year, age, cirrhosis of the liver, end stage renal disease, history of frost-bite/hypothermia, HIV+/AIDS, mental health, substance use and chronic health conditions. Scores range from 0 to 8, and to receive a score higher than zero, an individual must have experienced at least six months of homelessness [13]. It takes approximately 15 minutes to complete and relies on self-reported data from clients [14]. It has been widely implemented, particularly in the United States (US), where its use has been documented in 62 US communities [15] and in Canada [16].

Defining Australian homelessness

According to the Australian Bureau of Statistics’ (ABS) definition, homelessness is when a person has no suitable accommodation and lives in an inadequate dwelling, has no or limited/non-extendable tenure, and has no control of/no access to or space for social relations [17]. Homelessness is an umbrella term used to describe four broad population groups 1) Rough/Street sleeping; 2) Supported accommodations (e.g. refuges & crisis accommodation; 3) Short–term accommodation without tenure (e.g. boarding houses, hostel, caravan, couch surfing); and 4) Accommodation in institutional settings (e.g. hospitals, drug and alcohol rehabilitation centres, jail) [18]. Aboriginal and Torres Strait Islander people are over-represented in homelessness in Australia, making up 3% of the total population but 20% of all persons who were experiencing homelessness on Census night in 2016 [17]. The causes of homelessness are extremely complex and include structural factors such as criminal justice system association and poverty along with individual factors such as trauma, substance use, and mental illness [9]. The likelihood of accessing and receiving health care decreases with the complexity and length of homelessness [19].

In Australia, more than 116,000 people are experiencing homelessness [17]. The City of Sydney Bi-Annual Street Count reports 334 people sleeping on the streets and 505 people sleeping in crisis and temporary accommodation [20]. The lifetime prevalence of health conditions among the inner-Sydney homeless population were recorded in the Registry Week survey, 30 November– 2 December 2015 [20]. The most prevalent health conditions reported were hepatitis C, followed by asthma, heart problems and liver disease [20]. Two Australian studies examined emergency service use over two separate time periods (24 months over 2003–4, and one month in 2009), and both reported that people experiencing homelessness had lower rates of access to general practitioners (GPs) than general population [11, 21].

Rationale for the review

The authors of this review practice at St Vincent’s Hospital Homeless Health Service, Sydney, Australia. Established in 2010, the Homeless Health Service consists of a multidisciplinary team of nurses, doctors, allied health professionals, drug and alcohol clinicians, oral health educators, peer support workers, and Aboriginal health workers. The service provides holistic, multidisciplinary healthcare and support to people experiencing homelessness in the City of Sydney Local Government Area (LGA). To date, the Homeless Health Service has not employed any form of assessment tool to identify clients experiencing homelessness at particular risk of mortality and morbidity. This review is the first stage of a project to develop and implement a tool to prioritise people experiencing homelessness for access to healthcare services based on their health need and their capacity to access healthcare.

Protocol and registration

A scoping review of the literature was conducted between January and April 2020, reported here in accordance with the Preferred Reporting Items for Systematic Review and Meta-analyses Scoping Review extension (PRISMA-ScR) guidelines [22]. A study protocol was developed and registered initially as a systematic review on PROSPERO (doi: 10.15124/CRD42020166983) in January 2020, and later amended to a scoping review on account of the breadth of the research questions.

The search strategy was identified by the authorship team, a multidisciplinary team of clinicians and academics with considerable knowledge of health service delivery to people experiencing homelessness. Prior to agreeing to the selected search terms (Fig 1) a scoping review of terms was conducted, which included the terms measure* OR risk factor* and resulted in a diverse range of papers that were non-applicable to the research questions and were therefore excluded. The final search terms are shown in Fig 1 and were used to search in each database, MEDLINE, Embase, CINAHL, Scopus, PubMed, Web of Science. A copy of the search strategy is provided as a S1 File.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Search terms.

https://doi.org/10.1371/journal.pone.0254100.g001

A search of prominent authors in the field of homelessness was also undertaken (Conroy, E., O’Connell, J., Parsell, C., Teasdale, M., Baldry, E.). In addition, the search terms Vulnerability Index AND Homelessness were searched through Google Scholar. Saturation was reached after the first four pages of results.

Search & eligibility criteria

All search results (n = 430) were exported to EndNote X8 for eligibility assessment. After removing duplicates (n = 213) the abstracts were screened against the inclusion and exclusion criteria, shown in Table 1, independently by two authors. There were no disagreements between the authors. The remaining 34 studies were read independently in full by at least two authors and screened against the eligibility criteria and then discussed with all authors for consensus.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Inclusion and exclusion criteria.

https://doi.org/10.1371/journal.pone.0254100.t001

Critical appraisal

Given that the results of this scoping review were to be used to underpin the development of a clinical tool, the research team undertook a critical appraisal of the included studies. The critical appraisal was undertaken using the Joanna Briggs Analytical Cross Section Design Appraisal criteria [24]. Two authors (JC, EG) independently appraised each study and compared their findings, any disagreements were resolved through discussion.

Data charting and synthesis

A data charting form was developed by two of the authors (EG & JC) and reviewed by all authors, there was unanimous agreement in the characteristics extracted from the included studies. The data charting was undertaken independently by two authors (EG & JC) and then findings were compared and discussed. There were no disagreements between the two authors and the findings extracted were extremely similar. Data extraction included author, year, country, participants, study design, results and conclusion. Also extracted from the studies were the characteristics included in each vulnerability index, including the demographic data, health history and health service usage.

Selection of sources of evidence

This scoping review investigated the characteristics assessed by vulnerability indices and the health outcomes achieved by applying vulnerability indices to people experiencing homelessness. There was unanimous agreement that six studies met the inclusion and exclusion criteria and were included in this systematic review. The results of the search process are outlined in Fig 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. PRISMA diagram.

https://doi.org/10.1371/journal.pone.0254100.g002

Critical appraisal of sources of evidence

Overall, the selected papers were of a high quality and therefore none of the studies were removed on the basis of rigour (Table 2). Each study gained ethics/board approval, recruitment processes were well described, and appropriate statistical analysis was conducted. Across all of the studies, three papers clearly met all of the appraisal criteria. In two of the studies the management of confounding variables was unclear [16, 25].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Critical appraisal.

https://doi.org/10.1371/journal.pone.0254100.t002

In one other study the description of the settings from which the subjects were drawn was unclear [26]. This study used primary data from a previous study and excluded data from communities where greater than 50% of data were missing and it was not explicitly stated, which communities were excluded [26].

Characteristics of sources of evidence

The final analysis included six studies (Table 3), all quantitative, four from the US, one each from Colombia and Canada, published between 2008–2018. The length of data collection for the studies ranged from six nights [16] to two years [27], and the total number of participants was 27,509. Data for the studies were collected directly from participants at drop-in centres, temporary shelters [25] or solely on the street [16, 28], from existing homelessness databases, [27] and data from previous campaigns including the 100K campaign [26] and Directions Home in Texas [13].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Data charting vulnerability indices.

https://doi.org/10.1371/journal.pone.0254100.t003

The aims of the studies varied. One study used a vulnerability index to establish a health profile of a homeless population, [16] one study compared characteristics of ‘unsheltered and sheltered’ people experiencing homelessness using a vulnerability index, [26] another used a vulnerability index to assess health service usage [25]. One study examined the validity and/ or reliability of a vulnerability index [27] and another conducted an empirical evaluation of the predictive capacity of demographic, health and clinical variables on vulnerability score [13] and another developed and examined the validity of an HIV vulnerability index [28].

Where reported, the mean age of participants was 48 years [13, 25] and 40 years [16, 28]. Most participants were male [13, 16, 25] and the majority of participants reported at least one severe medical condition [13, 16, 25]. Mean length of continuous homelessness was reported as 4.7 years [25] and just under six years [16]. The most common medical conditions reported by participants were heart disease and skin conditions, [25] hepatitis C [16, 25] and asthma [16].

Results of individual sources of evidence

Characteristics of vulnerability indices.

The included studies used different vulnerability indices and included different characteristics, as shown in Fig 3. Four studies used the original vulnerability index developed through the 100K campaign [12, 13, 16, 25]. One study developed an HIV vulnerability index, [28] to explore the associations between homelessness, social factors and HIV risk behaviours in participants living in Colombia. One study examined the reliability and validity of the Vulnerability Index–Service Prioritization Decision Assistance Tool (VI-SPDAT), a 50-item assessment composed of yes/no questions grouped into four themes: history of housing and homelessness, risks, socialisation and daily functions and wellness. In addition to these domains there is one demographic question to assess age. The responses to the domains are scored and subtotalled to an overall score of 0 to 20 [27].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Characteristics of vulnerability indices reported in the included studies.

https://doi.org/10.1371/journal.pone.0254100.g003

The health history reported through the four studies using the original vulnerability index was similar and included cold injury (frostbite) [13, 16, 25, 26]. The study of the VI-SPDAT elicited participants’ history of frostbite and/or heat stroke [27] as well as detailed health history including tri-morbidity and an assessment by the surveyor of the participant’s level of hygiene and any signs of serious health conditions. The term ‘tri-morbidity’ is understood as the combination of a serious physical health condition, serious mental health condition and drug and alcohol dependence [29]. The study of the HIV vulnerability index reported sexual history, participant’s knowledge of HIV transmission, drug use frequency and social networks as well as participants’ perceptions of rejection and discrimination [28]. Based on their findings the authors developed a model for HIV vulnerability that includes components of risky behaviours (e.g. drug use, no condom use), lack of support and presence of rejection by others in the form of discrimination or stigmatisation, and erroneous knowledge of HIV and AIDS [28].

The health service usage reported by each of the studies varied, reflecting the different aims of the studies. For example, in their examination of the reliability and validity of the VI-SPDAT, Brown and colleagues (2018) elicited detailed data relating to the frequency of participants’ attendance to ED, crisis services, inpatient services, ambulance attendances, treatment for drug or alcohol dependence, and mental health review in the past six months. Two studies did not report health service use [16, 28].

Reported limitations of vulnerability indices

The included studies identified limitations of using vulnerability indices. Studies using the original vulnerability index highlighted the self-report nature of data collection as a limitation due to the potential subjectivity of responses [13, 25, 26]. The face to face administration of vulnerability indices was also identified as having the potential to impact on the transparency of responses [25]. In the Bowie & Lawson (2018) study none of the participants reported a diagnosis of HIV, and the authors believed this was unlikely to be valid and more likely a symptom of fear of disclosure. Furthermore, the potential to misunderstand the terminology used in a vulnerability index was also identified. For example, the original vulnerability index uses the term ‘heart disease’ and whilst some participants in the Bowie & Lawson (2018) study had hypertension they did not self-report it because they were not aware that hypertension would be included under the term heart disease.

Another limitation highlighted was the predictive capacity of the original vulnerability index. In their postdictive validity study of the vulnerability index, Cronley and colleagues (2013) suggest that their findings offer ‘…tentative support…’ for the validity of the vulnerability index and its capacity to predict hospital admissions, in so far as individuals who scored higher on the vulnerability index had more frequent hospital admissions [13]. Although, in relation to the self-reporting of chronic physical health conditions, mental health and substance use, hospital records were not a significant predictor of an individual’s vulnerability index score [13]. The authors conclude that further research is required to assess the accuracy of relying on self-reported health status [13].

The authors using the VI-SPDAT reported a tendency for participants to over-report and recommended triangulation of sources of data other than self-report [27]. The authors called for further psychometric testing and did not recommend the VI-SPDAT as the sole instrument for housing prioritisation [27].

Synthesis of results

In relation to the research questions posed by this scoping review, the characteristics included in the vulnerability indices used in the studies, the most common demographic characteristics were age and housing status/history, most commonly included health history characteristics were trimorbidity, substance use, mental health, liver, kidney and heart disease, emphysema and asthma, cold weather injury, diabetes and cancer. There were no specific health outcomes reported from applying any of the vulnerability index tools, as this was not an aim of any of the included studies.

Summary of evidence

This paper reports a scoping review of the application of vulnerability indices to people experiencing homelessness. In developing and implementing a tool for the Australian context, we identified three translational findings from this review.

Our first finding is that the breadth of characteristics included in existing vulnerability indices suggest that an index can be tailored to the context and service in which it will be used [28]. The most common health conditions identified in participants in the included studies were heart disease, skin conditions, hepatitis C and asthma. These conditions are also prevalent among people experiencing homelessness in Sydney. In addition, we would consider including traumatic brain injury (TBI) and heat stroke, which are already included in the VI-SPDAT [27]. Often secondary to unintentional injury [9], the lifetime prevalence of TBI is high among those experiencing homelessness and associated with poorer health outcomes and higher suicidality [30].

Secondly, our project requires that we develop a vulnerability index designed specifically to prioritise clients for access to health care. Both the VI-SPDAT and the original vulnerability index have been successfully applied to homelessness services across Australia to prioritise clients for housing [25, 27]. Now in its third iteration, the VI-SPDAT appears to be the most popular assessment tool in Australia, and is used by Micah Projects, Brisbane [31] Australian Alliance to End Homelessness [32], Homelessness NSW [20] and the Western Australia Initiative [33].

Thirdly, the tool needs appropriate psychometric testing. Whilst proven valuable as a data collection tool, the capacity of the original vulnerability index to predict outcomes is less certain [13, 25, 26]. Following their empirical evaluation, Cronley and colleagues [13] suggest the vulnerability index would be more effective if used in conjunction with other tools that collect data on social supports and health behaviours regarding vulnerability, rather than used on its own. Analysis of the reliability and validity of the VI-SPDAT, the original vulnerability index by the US Department of Housing and Urban Development and the National Alliance to End Homeless concluded that the vulnerability index and VI-SPDAT are limited in their ability to select the best intervention and predict the most successful intervention [34].

Recommendations for future research

The overall findings of this scoping review demonstrate the use of vulnerability indices as tools to identify the health status of homeless populations. In our St Vincent’s Hospital project, we have gained permission to modify the VI-SPDAT to develop a specific health version of this existing tool.

Limitations

This scoping review has some limitations. Using articles and databases in English language meant relevant non-English studies were not identified. The search was limited to peer-reviewed publications. In the search there were several government reports identified, and these were excluded. The sample of identified studies was relatively small. Only six studies met the inclusion criteria for this review, indicating the need for more peer-reviewed publications that apply a vulnerability index to homeless populations.