Measures

Demographic characteristics including gender, age, ethnicity and educational level were assessed. Ethnicity was categorized into ‘native Dutch’ when the participant and both parents were born in the Netherlands and as ‘non-native Dutch’ when participants were foreign born or when participants were born in the Netherlands but one or both of their parents were foreign born. Education was categorized as ‘lowest’ when the participant completed primary education at the most, as ‘low’ when the participant completed pre‐vocational education, lower technical education, assistant training or basic labor-oriented education, as ‘intermediate’ when the participant completed secondary vocational education, senior general secondary education or pre-university education, and categorized as ‘high’ when the participant completed higher professional education or university education.

The amount of debts (not including mortgages without overdue payments) was assessed: debts reported by participants showed a very skewed distribution with various outliers (range of the continuous data: 0–500,000 euros). Therefore, we dichotomized debts into ‘1000 euros or more’ (high; >first quartile) and ‘less than 1000 euros’ (low; <first quartile). This cut-off between high and low debts was data-driven because no normative data for the amount of debts were available.

The Dutch abbreviated version of the Lehman Quality of Life Interview (Wolf et al. 2002) was used to assess the adequacy of finances to cover basic expenditures. Participants were asked “During the past month, did you generally have enough money to cover (1) food, (2) clothing, (3) housing, (4) traveling around the city for things like shopping, medical appointments, or visiting friends and relatives, and (5) social activities like movies or eating in restaurants?” (yes or no). The mean number of covered expenditures (range 0–5) was calculated.

Stable housing was defined as at least 90 consecutive days independently housed or living in supportive housing (owned by care organizations) (yes or no) (Van Straaten et al. 2016).

Unmet care needs were assessed using a questionnaire developed by Impuls - Netherlands Center for Social Care Research (Lako et al. 2013). The response categories were based on the format of the Short-Form Quality of Life and Care questionnaire (QoLC) (Wennink and Wijngaarden 2004). Care needs were considered on eight life domains: finding housing, finances, basic skills, searching for work, physical health, mental health, dental care and safety. For each domain, two questions were asked: “Do you want help with …?” and “Do you get help with … ?”. An unmet care need variable was created for each life domain, which is scored affirmatively when participants indicated they wanted help, but did not receive help. All unmet care needs were summed to a total unmet needs variable, ranging from 0 to 8. The questionnaire has previously been used among homeless youth (Krabbenborg et al. 2013) and abused women (Jonker et al. 2012).

We asked participants: “Do you have health insurance?” (yes or no).

Social support was assessed by five items derived from scales developed for the Medical Outcome Study (MOS) Social Support (Sherbourne and Stewart 1991). Participants were asked to indicate how often different kinds of support were available to them through family and friends or other acquaintances, on a 5-point scale ranging from ‘none of the time’ to ‘all of the time’. Two social support measures (ranging from 0 to 5) were constructed by averaging across items: a family measure, and a friends and acquaintances measure. The MOS Social Support Survey has been used in several studies among homeless people (O’Toole et al. 1999; Nyamathi et al. 2000) and showed high convergent and discriminant validity and internal consistency (Sherbourne and Stewart 1991). The selection of items used in the present study has been successfully used in previous longitudinal research among homeless populations (Lako et al. 2013; Krabbenborg et al. 2013).

Experiences of relatedness were measured by one subscale of the Basic Psychological Needs questionnaire, based on the basic psychological need satisfaction-work version (Ilardi et al. 2006). Participants were asked to indicate their agreement with 7 items on a 7-point Likert scale, ranging from not true at all (1) to definitely true (7). An example of an item is: ‘People in my life care about me’. The scale has been used in previous studies (Gagné 2003), including a study among homeless youth (Krabbenborg et al. 2013). Adequate factor structure, internal consistency, reliability (Cronbach’s alpha = .92), discriminant validity and predictive validity have been demonstrated (Vlachopoulos and Michailidou 2006; Johnston and Finney 2010). The relatedness subscale score ranges from 0 to 7 and was constructed by averaging across the items of the subscale.

The Dutch abbreviated version of the Lehman QoL Interview (Wolf et al. 2002) was used to assess whether participants had a job by asking: “Do you have a job at this moment (paid job or volunteer work)?” (yes or no).

The Dutch abbreviated version of the Lehman QoL Interview (Wolf et al. 2002) was used to assess whether participants had been arrested by asking: “Have you been arrested or picked-up for any crimes in the past year?” (yes or no), and “Did you get any fines for any violations of the law in the past year?” (yes or no).

The Brief Symptom Inventory 18 (BSI-18) was used to measure psychological distress (Derogatis 2001). The BSI-18 is a short form consisting of 18 items taken from the Symptom Checklist-90-R (SCL-90-R) (Derogatis 1994), which correlates highly with the SCL-90-R. The BSI-18 assesses three symptom scales (i.e. depression, anxiety and somatization), which are included in a total score as an indication of general psychological distress. The BSI is a frequently used measure to evaluate psychological distress in studies among homeless populations (McCaskill et al. 1998; Kashner et al. 2002; Ball et al. 2005; Weinreb et al. 2006; Tsemberis et al. 2012). Respondents rated, from 0 (never experience symptom) to 4 (very often experience symptom), 18 items like “Nervousness or shakiness inside” and “Feelings of worthlessness”. The Dutch translation was used, with (provisional) norm scores for the Dutch population (De Beurs 2011). We compared the scores of the participants with the norm scores described in the manual for the Dutch community sample, with separate norm scores for men and women, and for different age categories (18–29 years and 30+ years) (De Beurs 2011). Because norms for t-scores are not available for the Dutch BSI-18 (De Beurs 2011), participants were categorized as having a high level of psychological distress if they scored in the upper 20^th percentile on a subscale compared with a Dutch community sample. Participants were categorized into two groups: participants with a high level and participants with less than a high level of psychological distress.

To create a change variable for psychological distress, participants were classified as having “reduced psychological distress” when they had a high level of distress at baseline and no high level of distress at follow-up, as “increased psychological distress” when they had no high level of distress at baseline, but a high level of distress at follow-up, and as “no change in psychological distress” when they had both at baseline and follow-up either a high or no high level of psychological distress. For all other variables, a change variable was created by subtracting the score at baseline from the score at follow-up.