We identified 31 relevant screening tools from the 22 studies included in this narrative literature review. All of the included studies were from the United States (US) [3, 4, 6–25]. Healthcare settings included Veterans Health Administration (VHA) outpatient clinics [4, 6, 8, 10–12, 19–21], primary care clinics [3, 9, 13–15, 22, 23, 25], emergency departments [3, 9, 13, 25], specialty outpatient clinics [15, 24, 25], inpatient hospital care [15], home-visiting programs [25], community health centers [18, 25], and veteran homeless center programs [17]. The organizations involved were serving financially disadvantaged patients [3, 13, 18, 23], pediatric patients [22, 25] and rural patients [16]. A significant proportion of studies (n = 10) targeted the US veteran population [4, 6, 8, 10–12, 17, 19–21]. This reflects the US Department of Veterans Affairs’ (VA) recognition of this population’s vulnerability to homelessness, and subsequent screening and intervention initiatives [4].

All screening tools assessed the patient’s current housing status and/or concerns about future housing instability (Table 1). Seven screening tools utilised direct questioning about housing stability, with questions such as “Over the past year, were you homeless or living in a shelter at any time?” [3], or “Are you worried that in the next 2 months, you may not have stable housing?” [15]. These types of questions relied on the patient’s evaluation of their current living situation. Most screening tools (n = 23) included open-ended questions such as, “What is your housing situation today?” [3, 9, 13, 15, 18, 23], “In the last 12 months, how many times have you or your family moved from one home to another?” [15, 25], or “Do you have any concerns about being evicted or not being able to pay the rent?” [15].

Screening tools used timeframes ranging from the past 12 months to the upcoming 2 months. Questions were binary (yes/no) [3, 4, 6, 9–16, 18–23, 25], single-answer multiple choice [3, 4, 6, 10–12, 14, 15, 18–22, 25], multiple-answer multiple choice questions [3, 14, 15, 18, 22, 25] or numerical questions [3, 9, 13–15, 18, 20, 23, 25]. Mode of screening administration, in the descending order of number of tools that utilised the particular mode, included self-administration on paper [14–16, 22–25], assisted completion via face-to-face interview between provider and patient (based on a paper form/template in the electronic health record) [4, 6, 10–12, 14, 15, 17, 19–21], self-administration electronically (computer/tablet) [3, 9, 13–15, 18, 22–25] or assisted completion via telephone interview [18, 22, 25]. Some tools had more than one mode of administration. Screening tools often assessed other social risk domains (e.g., food insecurity, financial instability) along with housing-related risks [3, 9, 13–15, 18, 22–25].

Sensitivity and specificity data were only available for the Veterans Health Administration (VHA) Homelessness Screening Clinical Reminder (HSCR) [12, 20]. The HSCR is a screening tool completed via a provider-led interview prompted by the Computerized Patient Record System (CPRS). For the 2017–2018 veteran sample, specificity (86.5%) was significantly greater than sensitivity (32.3%) when screening for homelessness [12]. Similarly, specificity (90.0%) was much higher than sensitivity (16.0%) when screening for risk of homelessness [12]. These statistical values suggest that a considerable proportion of patients experiencing homelessness are not being identified by this screening tool. Several other studies provided data on the living situation of positively screened patients without formal statistical analysis, in order to offer a degree of insight on the tool’s validity [4, 10, 19, 21]. Using the HSCR, over half the patients who screened positive for current homelessness reporting living in a homeless situation for most of the 2 months prior [4, 19].

One of the studies around veterans explored the potential for artificial intelligence to be employed in screening for unstable housing [8]. In this study, an open-source natural language processing (NLP) tool was trained to recognise electronic medical record documents as having either “evidence of homelessness” or “no evidence” [8]. Human review determined a precision of 70% of positive flags for homelessness, suggesting that with further development, NLP tools and other forms of artificial intelligence could be used to create effective and efficient screening processes [8].

A wide variety of staff were involved in screening (Table 1). 12 studies did not specify what type of healthcare provider administered the screening tool [3, 4, 6, 9, 10, 13, 15, 18, 19, 22, 23, 25]. In one systematic review, six out of eight tools screening for housing instability had associated training for screening professionals [25]. While there were no studies comparing tool administration between different providers, Montgomery et al. found that veterans screened by a behavioral health or social work provider were almost twice as likely to access services within 30 days after a positive screening than those screened by physicians [21].

Six studies discussed the acceptability of the screening tool from the patient and/or provider’s perspective [9, 11, 13, 14, 21, 24]. De Marchis et al. reported that 79% of participants found the Accountable Health Communities’ (AHC) social risk screening tool appropriate and 65% felt comfortable with the inclusion of social risks in electronic health records (EHR) [13]. Pinkerton et al. reported that 92% of patients with ≥1 health-related socioeconomic risk (HRSR) found it appropriate to assess for HRSRs in clinical settings and 72% felt comfortable with EHR documentation [24]. Byhoff et al. conducted interviews with participants who completed a survey that utilised the AHC Screening Tool and found that they viewed screening for social risks as acceptable and crucial [9]. Participants conveyed they wanted their healthcare providers to be informed of their social circumstances, but did not expect them to resolve social issues [9].

Chhabra et al. carried out in-depth interviews with VHA clinical providers about the HSCR which reflected overall positive engagement with the tool [11]. Providers reported that the HSCR encouraged them to integrate patient housing status into routine assessment and subsequent clinical decision-making [11]. A later study in 2021 by Montgomery et al. provided further insight into provider perspectives on the HSCR, reporting that they found the tool easy to administer, but used it as a reminder to enquire about housing rather than following the language of the HSCR verbatim [21]. They attributed this to the overly formal or vague language of the HSCR and concerns about patients’ subjective definitions of what constitutes “stable housing” [21].

From a broader perspective of social risk screening and interventions, Eder et al. reported a majority of the articles (n = 31) in their literature search conveying patient satisfaction with and acceptability of screening and interventions, with frequent mentions of improvements in the patient-clinician relationship and high comfort levels [14]. However, 11 articles did raise issues of discomfort and confidentiality which will be further discussed in the “Barriers and Facilitators of Screening” section below [14]. 17 of 18 articles on clinician satisfaction with screening and interventions reported positive findings, including the time-efficiency of screening and subsequent improvements in the patient-clinician relationship, patient care and clinician knowledge and competence [14]. Overall, patient and clinician acceptability of the various screening tools appeared to be high with mutually perceived benefits in the therapeutic relationship and clinical care.

Six studies [9, 11, 13, 14, 16, 23] discussed potential barriers and/or facilitators to screening for housing instability. De Marchis et al. described that patients were more likely to perceive screening as appropriate if they were in a primary care setting, had previous exposure to healthcare-based social risk screening, or trusted their clinician [13]. Patients were more likely to perceive screening as less appropriate if they had previous experience of healthcare discrimination [13]. Patients were also more comfortable with EHR documentation of social risks if they had previously received assistance with social risks in a healthcare setting [13]. Patients in the qualitative study by Byhoff et al. expressed the importance of a patient-centered approach to social risk screening that shows empathy, compassion and respect [9].

In a quality improvement project, staff reported that they forgot to offer the self-administered screening tool [16]. After the project leader placed the screening tools in a brightly colored folder by their workstation, the screening rate improved from 68% to 45% in Week one and two respectively to 77% in Week three [16]. From interviews with VHA clinical providers about the HSCR, barriers to screening for housing (prior to the HSCR) included providers’ assumptions about who might be at risk, lack of training in medical school and residency, and lack of previous work experience in settings where housing instability is sought to be addressed on a system-wide level [11]. Other potential barriers were raised, such as providers feeling better positioned and more inclined to focus on the medical rather than social factors of the patient’s presenting complaint [11]. This was in consideration of their primarily medical expertise, the time constraints of each brief consultation and lack of familiarity with available resources [11].

Regarding facilitators of screening, incorporation of the HSCR into the Computerized Patient Record System has been effective in prompting screening as clinical providers have to click on it as they proceed through a consultation [11]. Palakshappa et al. found that a major barrier to systematic screening was a lack of personnel, especially in the context of free and charitable clinics as they are often dependent on volunteers [23]. Potential facilitators identified include having a dedicated patient advocate at clinics to assist with social needs [23].

Eder et al. identified patients’ concerns about stigma and privacy, clinician’s concerns about lack of referral resources, and concerns on a health system level about how social risk data is collected and managed across systems [14]. In summary, there is a need for: 1) training for clinicians on screening, 2) time efficiency of the screening tool, 3) incorporation of the screening tool into existing workflow, 4) a robust patient-doctor relationship, and 5) proper integration between healthcare and social care organizations.