Sampling process

The sampling process targeted people with disabilities receiving services under the national scheme for socio-pedagogical support (Convention ASP). Eligible participants were individuals living in community-based residential settings providing 24-hour support or attending day-care services operated by one of the 14 participating service providers governed under the ASP Convention.

Due to data protection regulations and the absence of a comprehensive registry of people with disabilities in Luxembourg, including those covered by the ASP Convention, it was not possible to obtain a complete list of the target population. Instead, the sampling frame was established using administrative data provided by the participating disability service providers, which included all users receiving services under the ASP Convention. In September 2021, a total of 1,115 individuals met the inclusion criteria and constituted the sampling frame.

To ensure that the study sample was representative of the target population, a stratified random sampling strategy was implemented based on four characteristics: age, sex, residential status, and employment status. The final analytical sample consisted of 152 individuals. A weighting procedure was subsequently applied to preserve the distribution of these four characteristics in the target population, thereby ensuring that the analytical sample accurately reflected the demographic and service-related composition of the 1,115 eligible individuals.

Procedure

For each of the 152 participants, service provider representatives completed a pre-interview. This pre-interview sought to get to know the respondent better, their living environment, their personal and family situation and disability characteristics. It was also intended to find out more about their preferred mode of communication and their capacity to participate in the interview, either on their own or with support, as well as the language spoken, to creating an environment conducive to interaction.

The questionnaire (S2 Appendix) was administered via face-to-face semi-structured interviews at the service providers’ premises. Interviews took place between September, 1 2021 and March, 30 2022. Safety measures concerning the COVID19 crisis were applied [COVID-19 safety measures included wearing masks, limiting the number of people in a room, maintaining physical distance, ensuring proper ventilation by opening windows, and postponing interviews if there was any risk of illness or exposure]. To limit the potential influence of pandemic-related restrictions on responses, participants were explicitly asked to report their social interactions and neighbourhood contacts based on their situation prior to the COVID-19 restrictions. The questionnaire was developed jointly by the multidisciplinary research team, building on previous benchmarks and examples in the literature [9,29,30]. The questionnaire also underwent review by experienced representatives of service providers, operating in the field of disability, and revision after pilot interviews. Two pilot interviews were conducted with people with disabilities who were not included in the study sample but were supported by Convention ASP and received services from the same providers as study participants. All interviewers participated in at least one pilot interview, either as an interviewer or observer, under the supervision of a lead researcher. The pilot interviews were used to train interviewers, test the questionnaire, and assess the feasibility of the Computer-Assisted Personal Interviewing (CAPI) method. Based on the pilot findings, the CAPI approach was replaced by a paper-and-pencil format, and several questionnaire items were reworded to improve clarity and relevance.

In total, five specifically trained interviewers conducted the interviews, including the pilot interviews. Participants could choose to be accompanied by a trustworthy person of their choice. In case of communication difficulties of the participant, supportive communication was used in the form of pictograms, or a trusted person could respond on behalf of the participant. When the participant was not present, a trusted person responded entirely on their behalf (proxy interview). The proxy interviewer was expected to have known the disabled person for at least one year in typical daily situations. All interviews were conducted using paper and pencil and were subsequently entered into a CAPI using the CONVERSO software.

Ethical considerations

Prior to data collection, the Ethics Committee of the research institute approved the protocol of this study, including the content of the consent form, the questionnaire and the data collection method. All invited participants were then informed about the survey being conducted and their opportunity to participate if they wish. A consent form was sent to them. This document was filled in by the participant himself or, where applicable, by a legal guardian. Moreover, an information letter, also in easy language, was sent to those involved, describing the purpose of the study, the rights of participants and legal guardians, and the use of the data.

Quality of life

The QoL of individuals with disabilities was assessed using an adapted version of the widely recognised instrument developed by [29]. This model conceptualises QoL across eight key domains: emotional well-being, interpersonal relations, material well-being, personal development, physical well-being, self-determination, social inclusion, and rights. Guided by this framework, we constructed a 16-item scale designed to capture a broad yet concise representation of these domains (see Table 1). In adapting Schalock’s instrument, we did not include the full original questionnaire. Given its length, the shorter version was selected to reduce respondent burden, maintain participant engagement throughout the interview, and minimise the risk of fatigue, incomplete responses, and dropout. The questionnaire also encompassed other important constructs, such as practicability, acceptability, and temporal burden, which required limiting the number of items to prevent respondent fatigue. The 16 items included in the QoL measure were carefully selected to ensure balanced coverage of the eight domains, while maintaining alignment with Schalock’s theoretical model (Table 1). The respondents answered using a 5-point Likert scale, from 1 (completely true) to 5 (not true at all). When needed, the items were recoded, so that the higher the score, the greater the QoL.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Scale of quality of life.

https://doi.org/10.1371/journal.pone.0356610.t001

To ensure that participants were able to understand and respond to the items, several measures were implemented: interviewers received specific training on conducting interviews with people with disabilities, questions were reformulated when necessary, using simpler language, and pictograms were made available as a communication support. Where participants were unable to respond themselves, proxy responses were used, an approach also described in QoL research involving people with severe intellectual disabilities when self-report is difficult or not possible [31]. Finally, to enhance response validity, specifically trained interviewers monitored the consistency of responses throughout the interview and checked for potential inconsistencies between items addressing similar domains.

We used an adapted quality of life scale derived from Schalock’s framework. To assess its internal consistency, we calculated Cronbach’s alpha coefficient, a widely used measure of reliability that evaluates the extent to which items within a scale are correlated and consistently measure the same underlying construct. The analysis indicated that two items exhibited low item-total correlations and were not sufficiently consistent with the underlying construct measured by the scale. Consequently, they were removed from the final analysis to improve the scale’s internal consistency, bringing the number of items to 14 (Table 1). Cronbach’s Alpha value of the adapted scale is 0.57.

Explanatory variables

The factors considered in this study were grouped into four broad categories: socio-demographic characteristics, living environment, disability-related functional limitations, and social interactions. The socio-demographic characteristics included participants’ age, gender, and civil status. The living environment encompassed variables such as place of residence (e.g., city center, outside the city), type of residence (e.g., family home, institutional setting), type of room (e.g., private or shared), and whether the individual attended day care services. Disability-related functional limitations were captured using the ASP disability classification system. The ASP classification covers mild intellectual disability, moderate intellectual disability, severe and profound intellectual disability, poly-disability, motor disability, and autism spectrum disorder. Visual and hearing disabilities fall outside the scope of ASP coverage and are therefore not included in the sample.(See S3 Appendix, Table S3A for the corresponding ICD-11 categories of the ASP classification). Additional indicators included individual capacities in receptive and expressive communication, whether the person had limited mobility, required significant medical care, exhibited behavioural problems, or had objectively identified needs for support. The objective needs for support among individuals with disabilities were assessed using a 14-item scale specifically designed to capture a broad range of support requirements in performing activities such as getting dressed, eating, or going to the supermarket (See Table 2). The respondents answered using a 4-point Likert scale, from 1 (‘I can do it alone’) to 4 (‘someone has to do it for me’). To evaluate the internal consistency of the scale, Cronbach’s alpha was calculated. The resulting coefficient was 0.88, indicating a high level of internal reliability. Finally, the social interactions category included measures such as current employment (e.g., whether the person was working) and the frequency of contact in the neighbourhood.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Scale of objective needs for support.

https://doi.org/10.1371/journal.pone.0356610.t002

Sample

The selected sample was composed of people with disabilities benefitting from the ASP Convention and included in the initial sample of 152. Minors were excluded (n = 13) from the main analysis because they are still under legal protection of their parents and hence have lower opportunities of self-determination than adults. In addition, children with disabilities have a different living environment compared to adults with disabilities, as well as another perception of their own life as they typically would live in the parents’ home. Out of the 139 adults, we kept those with a minimum response rate of 60% to the QoL items to ensure data quality and to accurately capture the constructs measured by the scale. Robustness tests around this threshold are included in the Results section. These restrictions led to a final working sample of 124 adults.

Empirical strategy

We seek to identify the main factors associated with the QoL of people with disabilities using an Ordinary Least Squares (OLS) estimator such that:

where QoL_i represents the quality of life of individual i. The vector X_i includes a comprehensive set of factors such as disability characteristics, living environment settings, social participation, and socio-economic factors. The vector Z_i contains variables specifically related to interview conditions, such as the language of the interview and whether the respondent attended. is the error term.

To account for potential variation introduced by different interviewers, interviewer fixed effects are included. Additionally, since Luxembourg has 14 service providers operating under the ASP scheme—some of which specialise in supporting individuals with specific disabilities (e.g., autism)—service provider fixed-effects are incorporated to control for selection bias related to disability type and provider characteristics.

Descriptive statistics

The characteristics of the selected sample – composed of 56 women and 68 men – are presented in Table 3. More than two-thirds were single and aged between 19 and 59. More than 75 percent had intellectual disabilities of varying degree, including autism, almost 6 percent have poly-disabilities and around 15 percent have motor disabilities. This classification of disability is based on categories as defined in the ASP Convention, which we have expanded by adding the autism category as well as the category “others” to capture any other disability situation. In addition, 33 percent had mobility-related limitations and 21 percent reported significant needs due to behavioural problems such as preventing self-harm, tantrums, emotional outbursts and aggression. Almost 16 percent could not express themselves at all during the interview, and a third of the sample could not express themselves sufficiently. The average score of needs for support was 2.08, over a scale from 1 (low support needs) to 4 (high support needs). In terms of living environment, around half lived in a central urban area, in an accommodation managed by a service provider and the vast majority occupied a single room. In terms of social interactions, the distribution was almost equally balanced between daily, weekly or every two- or three-weeks contact, between 15 and 20 percent of the sample, however, almost 37 percent reported having no contact at all with their neighbourhood. Finally, only 27 percent of the sample worked in workshops with a salary under the “disabled worker scheme”. Those non-working were entitled to a state guaranteed minimum monthly living income.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Descriptive statistics of the sample.

https://doi.org/10.1371/journal.pone.0356610.t003

Table 4 shows QoL average scores by disability ASP profile. The average QoL score of the sample was 3.77 (over a scale from 1 to 5). QoL decreased with an increasing level of intellectual disability. While people with mild intellectual disability, poly-disability or motor disability graded their QoL above the average level, those with severe intellectual disabilities or those who were autistic reported one of the lowest levels of QoL.

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Quality of life of people with disabilities.

https://doi.org/10.1371/journal.pone.0356610.t004

Statistical analyses

Table 5 presents the results of the OLS estimations. Column 1 displays the estimates without accounting for interview conditions and without including interviewer and service provider fixed effects. The final column reports the results from the fully specified model, incorporating all interview controls and fixed effects. There were no significant gender or age differences in terms of QoL. As for civil status, divorced people reported higher levels of QoL than single people (p = 0.039). To a lesser extent, being married was also associated with greater QoL scores, compared with being single, but the associated coefficient was only statistically significant at the 10-percent level (p = 0.074).

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. Factors associated with the quality of life of people with disabilities – OLS estimates.

https://doi.org/10.1371/journal.pone.0356610.t005

We aim at capturing a global picture of the disability situation by observing several indicators, including the type of disability, the expression and communication skills as well as mobility-related limitations, significant medical treatment, behavioural troubles and needs for support. While most of these indicators were not statistically significantly correlated with QoL, the needs-for-support variable was negatively associated with QoL (p = 0.003): people with higher needs for support reported lower QoL scores.

Several levels of the living environment were considered: the neighbourhood, the accommodation and the bedroom. We also included the accommodation settings, indicating whether the person is in day-care, i.e., spends the day in the premises of the service provider and returns home in the evening, or is in full-time care. There was no significant difference in QoL between living in the outskirt of the city or out of the city, compared to living in a more central area. People living in a private accommodation with parents rated their QoL higher than those living in an accommodation run by the service provider. The estimated coefficient was highly significant (p = 0.006) and its magnitude showed a large effect of more than 1 point. People sharing their room with two or more persons (if no separate beds) reported lower levels of QoL. The magnitude of the coefficient increased as more people share the room, showing a detrimental effect on QoL of more than 1.5 points (on a scale from 1 to 5) when the room was shared by three people or more. In addition, people in day-care reported lower QoL (p = 0.041) than those in full-time care.

People were asked how often they talk to those in their neighbourhood. There was no difference in QoL when accounting for fixed effects. However, when not including service providers and surveyor dummies, there was a decrease in QoL when interacting with others only every two or three weeks (p = 0.035), only every two or three months (p = 0.045). When interacting even less, the effect on QoL was still negative but only statistically significant at the 10-percent level (p = 0.072). Labour market status was not correlated with QoL.

Robustness tests

Additional analyses were conducted to assess the robustness of our findings using alternative subsamples and sensitivity checks. First, given the high proportion of individuals with intellectual disabilities in our sample, including autistic individuals, we performed a focused analysis on this subgroup. Second, in some cases, the respondent did not personally participate in the interview and was replaced by a proxy, a trusted person who responded on their behalf. This may lead to over- or under-reporting bias; thus, we repeated the analysis using only the subsample of self-respondents. Finally, we examined the sensitivity of our results to the response rate threshold imposed on the QoL scale. Specifically, we re-ran the analysis on two restricted subsamples: those who answered at least 50% and at least 70% of the QoL-related items. The results of these analyses are presented in S4 Appendix, Table S4A and confirm the reliability of the main findings.