Considerations for Defining Homelessness in Public Health Data Collection

Public health problems cannot be addressed without timely and accurate data. However, data that provide insight into populations that may be at disproportionate risk for disease, including people experiencing homelessness, are insufficiently captured. Although the associations between homelessness and disease have been well documented,^1-6 data on housing status are not universally or consistently collected in routine public health data. ⁷ Improving collection of data on housing status in public health data collection efforts is necessary to address health disparities among people experiencing homelessness and advance health equity research and practice.

Collecting data related to homelessness and disease is complicated for several reasons, but one of the most salient reasons is that defining homelessness is challenging. Several federal agencies use their own definitions to identify people who might be eligible for assistance programs, creating confusion about which definition should be used for public health purposes. In addition, definitions of homelessness at federal agencies have changed over time, further complicating the collection of homelessness data. The US Department of Housing and Urban Development (HUD) defines homelessness as lacking a fixed, regular, and adequate nighttime residence ⁸ ; this definition includes both people with a primary nighttime residence of a public or private place not meant for human habitation (eg, cars, parks, public spaces, abandoned buildings) and people residing in temporary shelters (eg, emergency homeless shelter, transitional housing). The US Department of Education (DOE) uses a broader lens, defining homelessness to include school-aged children whose housing situation meets the HUD criteria for homelessness while also including those who share housing with other people by doubling up or couch surfing; those who live in motels, hotels, or trailer parks; and those who are abandoned at hospitals. ⁹

As a result of these challenges, data systems that collect housing data rarely define homelessness in the same way as other interoperable systems. Within the US Department of Health and Human Services,^10,11 various criteria are used to define homelessness depending on the program, with most programs using the core elements of HUD’s definition of literal homelessness. One example is the Centers for Disease Control and Prevention’s (CDC’s) National Tuberculosis Surveillance System. The Division of Tuberculosis Elimination has developed a definition of homelessness based on the HUD definition of homelessness to be used when assessing tuberculosis cases. ¹² In public health surveillance, housing status is often captured indirectly from data collected for different purposes, such as electronic medical records created primarily for clinical care or homeless service utilization records. Because data on housing status are collected in a nonstandardized way, public health surveillance is unable to apply precise or consistent criteria for homelessness. Moving toward a more standardized way of collecting data on housing status could improve the usefulness and comparability of systems.

Definitions of homelessness and housing instability for public health data collection efforts are not constrained by the need to determine assistance eligibility and should be developed with additional considerations for assessing disease risk, incidence, and prevalence, which we outline in this commentary.

The COVID-19 pandemic further highlighted public health challenges resulting from various definitions of homelessness used by public health data collection systems. ¹³ Approximately half of US jurisdictions reported to CDC that they were collecting data on homelessness and COVID-19; however, these jurisdictions used various criteria to determine whether someone with COVID-19 was experiencing homelessness, often triangulating multiple data sources to determine or validate homelessness status. ¹⁴ As a result, it has not been possible to determine a definitive national incidence rate of COVID-19 among people experiencing homelessness in the United States using available data. Routinely and accurately identifying people experiencing homelessness in public health data sources would make it possible to provide specific, timely, and strategic resources to mitigate public health risks and close the gap on health disparities faced by people experiencing homelessness. For example, public health officials could use high-quality data on homelessness to rapidly identify an outbreak in a congregate setting, resulting in a timely response to mitigate spread. On a population level, improved data could identify common risk factors associated with housing status during infectious disease transmission and support appropriate resource allocation, such as opening isolation spaces or prioritizing vaccination among people experiencing homelessness.^15-17

Not all experiences of homelessness are equally applicable or relevant when determining disease rates or risk factors. For example, examining infectious diseases with short incubation periods may require collecting information about recent time frames or recent exposures. Duration of homelessness, such as chronic homelessness that lasts continuously for 12 months or 12 months cumulatively during a 3-year period, may be more relevant than assessing homelessness in the past week to the assessment of chronic health conditions or to the analysis of causes of disease, including social determinants of health.

Rather than propose a single definition of homelessness for public health data collection, the purpose of this commentary is to outline key considerations for public health practitioners and researchers as they seek the definition of homelessness that is best suited for the goals of a data collection. This process involves consideration of numerous factors, including (1) distinguishing between homelessness and housing instability, (2) determining the setting of homelessness and purpose of public health data collection, (3) identifying the period and duration of homelessness, and (4) linking current and future public health data to existing population estimates.

Distinguishing Between Homelessness and Housing Instability

In public health data collection, data on homelessness and housing instability need to be collected as separate data elements to truly capture and understand the influences of these structural determinants of health. People experiencing housing instability are likely to be excluded from existing formal definitions of homelessness; their experience is not captured by a binary measure (eg, homeless vs not homeless). Housing instability is formally defined to include an unsustainable burden of rent cost, the risk of eviction, or frequent moves.^17,18 The aforementioned differences in definitions also come into play. People who are doubled-up (ie, staying with others due to housing instability and financial insecurity, sometimes called “couch surfing”) may be classified as “not homeless” according to the HUD definition of homelessness but classified as “homeless” by the DOE when it applies to young people. People experiencing housing instability who are technically housed according to the HUD definition may experience disproportionate health impacts compared with their stably housed counterparts, including food insecurity,^18,19 higher rates of domestic or sexual violence, ²⁰ and increased disease transmission risk. ²¹ Consequently, not including people who may be unstably housed in data collection could generate an incomplete picture of disease risk.

Because experiences of housing instability can look different across contexts, unique challenges exist in identifying people experiencing housing instability. For example, both people experiencing homelessness and people experiencing housing instability might access support services and be entered into data systems or records used by homeless service agencies. It may then be helpful to identify housing instability as an independent data element to distinguish the heterogeneity of experiences for people accessing housing or meal support services, programs, and spaces. However, one limitation of such data sources is that not all people experiencing housing instability will access these services and may go unrecorded. The lack of a comprehensive and unbiased data source is a central challenge to the ability to determine the size of the population of people experiencing housing instability.

CDC’s Medical Monitoring Project, a nationally representative survey of adults diagnosed with HIV, expanded data elements in 2018 to assess the effect of housing instability separately from homelessness. ²² Findings showed that housing instability, even among people not experiencing homelessness as strictly defined by HUD, is associated with poor HIV outcomes. Data on housing instability and homelessness are important to independently collect to understand the impact of housing status on health. While unique challenges in assessing who is experiencing housing instability may exist, it is important for public health practitioners to determine whether they are interested in the experience of housing instability, homelessness, or both and to make an intentional effort to capture data on these distinct experiences and their influence on health outcomes.

Setting of Homelessness and Purpose of Data Collection

The fluidity of an individual’s housing experience can be challenging to capture in public health surveillance. Thus, describing the setting of homelessness can provide insight into the risks and exposures a person may have experienced. Some people experiencing homelessness sleep in emergency shelters, while others sleep outside or in places not meant for human habitation. Distinguishing individuals in a congregate setting from those who are experiencing unsheltered homelessness can be important, as infection rates may differ for sheltered and unsheltered people. For example, the prevalence of COVID-19 was found to be higher among people in congregate shelter settings than among people staying in encampments (ie, temporary structures/enclosed places not meant for long-term human habitation) ²³ in several studies, although both groups would be considered to be experiencing homelessness by federal definitions.^24,25 In addition, people may rotate between sheltered and unsheltered homelessness, accessing emergency shelter services some nights and sleeping outside other nights. When distinguishing the setting of homelessness, it may also be beneficial to assess the duration spent in each setting to sufficiently assess risks and exposures. Such findings highlight the importance of collecting specific information on the setting of homelessness when determining disease risk, incidence, and prevalence within this heterogenous group.

Understanding the purpose of data collection should further help to identify the most appropriate inclusion criteria for any definition of homelessness. If the goal of data collection is to understand disease risk and potential exposures in a congregate setting, questions should be tailored to assess experiences in shelters. On the contrary, if the goal is to understand homelessness broadly across a variety of settings and facilities, data collection could additionally include an assessment of where someone reported sleeping most often during a particular period. Asking people if they are homeless yields a different response than asking people where they slept the previous night or where they sleep most often. In some cases, such as being doubled-up or sleeping in a car, people may not self-identify as homeless but may be defined as experiencing homelessness using federal definitions. Accounting for nuance in the setting of homelessness experienced could provide helpful context to understand trends in disease risk and transmission. Public health practitioners should also consider how to capture or describe the experiences of people who frequently transition between congregate shelters and outside locations, especially when exploring risk factors that may differ between settings.

Period and Duration of Homelessness

Different periods and durations of homelessness may also be uniquely relevant to specific types of public health investigations. After determining the inclusion criteria for housing status, it is important to define the period and duration of homelessness that the data collection effort should capture. Shorter time frames may be necessary to evaluate infectious disease spread, but longer time frames are important because homelessness is a structural determinant of health and acts as a predisposing factor to poor health outcomes.^26,27 In making this decision, public health practitioners and researchers should prioritize data that are most informative to the goal and type of data collection, whether it is outbreak investigation, routine surveillance, or other special studies.

Often, the disease incubation period is necessary to determine the relevant period if the data are being used to understand homelessness as an exposure or risk factor for a communicable illness. For example, some pathogens responsible for foodborne illness (such as Salmonella or norovirus) or respiratory illnesses (such as COVID-19) have short incubation periods; public health data collection related to these outbreaks might include a question about recent homelessness (eg, current, in the past 2 weeks). Conversely, for investigations of infections or conditions with longer incubation or latency periods, assessing homelessness for a longer duration or period may be necessary. From 2008 to 2015, a tuberculosis outbreak among people experiencing homelessness in Atlanta, Georgia, demonstrated that assessing homelessness during longer durations (such as one’s lifetime) was useful in addition to assessing homelessness in the 12 months before tuberculosis diagnosis. ²⁸ In 2020, the National Tuberculosis Surveillance System added a variable to assess any homelessness during one’s life. Currently, all verified cases of tuberculosis have information on lifetime homelessness as well as homelessness in the 12 months preceding tuberculosis evaluation in the National Tuberculosis Surveillance System. ²⁹ In addition, characterizing the duration of homelessness can be relevant to public health investigations, including whether the individual experienced homelessness that was chronic (often 1 year or more), episodic (multiple times but not consistently), or transient (a shorter period without recurrence). ³⁰ Data collection on these factors is also crucial to add to the evidence base in the literature, as little is known about the differences in disease risk and health outcomes related to durations and periods of homelessness throughout a person’s lifetime.

Linking to Existing Population Estimates

Several data sources can be used to estimate the total number of people experiencing homelessness in the United States, including the HUD Point-in-Time (PIT) count, ³¹ data available in local Homeless Management Information Systems (HMIS), and data available from the DOE. ³² HUD’s PIT count represents an estimate of people experiencing homelessness on any given night, while HMIS may provide information on the annual number of people experiencing homelessness. Population estimates, and the sources used to derive them, are critically important to the accuracy of outcome measures, including rates of infection, hospitalizations, and deaths among people experiencing homelessness. An ongoing need exists for improved collection of denominator data that enumerate people experiencing homelessness for many reasons, including linkages with public health data that could better monitor health needs in this population. However, if public health data collection will be linked to existing population or denominator sources, the current definition constraints of those sources must be considered. For example, when using the HUD PIT count as a population or denominator data source, the HUD definition of homelessness should be used in the public health data collection tools for maximum alignment. In our assessment of COVID-19 incidence among people experiencing homelessness, definitions of homelessness in public health data collection that varied from the HUD PIT count substantially limited our ability to generate incidence rates and compare incidence rates of COVID-19 among people experiencing homelessness with those of the general population.^14,33 More information on denominator data sources and their data use considerations is described elsewhere, including extensive discussion on the limitations of the HUD PIT count.^32-36

Public Health Implications

Ultimately, a single approach to collecting data on homelessness might not be appropriate for every public health data collection effort, but it is important to carefully consider the elements described here when planning data collection efforts. Local and state public health agencies can support the inclusion of housing status by reviewing their surveillance systems and identifying what definitions are currently in use in data collection. National-level organizations can support the inclusion of housing status by developing and validating question sets related to housing for widespread use in public health data collection. Further collaborative efforts across multiple public health organizations and agencies such as CDC, the Council of State and Territorial Epidemiologists, the Association of State and Territorial Health Officials, and the National Association of County and City Health Officials would allow for maximum investment and applicability of data collection tools for homelessness at all levels of public health governance. In addition, CDC recently posted considerations for defining homelessness in public health data collection, along with examples from data collection tools around the country, for public health practitioners to reference. ³⁷ More complete data on homelessness and health are imperative to improve data-driven public health action. Health equity cannot be accomplished without data equity, and people experiencing homelessness should be represented in public health data collection.