The participants in this study were recruited between 2021 and 2023 from the Community Mental Health Teams (CMHT) and basic health areas of Tenerife (Spain), one of the eight Canary Islands, geographically located in southern Europe. In 2021, Tenerife had a resident population of 930,570, and the Canary Islands reported a GDP per capita of €19,021, with the service sector as the predominant economic activity [23]. The unemployment rate in the Canary Islands has traditionally been among the highest in Spain, averaging around 20% during the study period. At the onset of the study (2021), unemployment exceeded 25%, gradually decreasing to about 14% by 2023. These figures remain substantially higher than the national average in Spain (approximately 14% in this period) and, than those reported in the European EQOLISE trial (around 10%–12%), thus supporting the characterization of this setting as a high-unemployment context.

The sample size calculation was based on the total score on the Positive and Negative Syndrome Scale (PANSS) as the primary outcome variable. A clinically relevant difference of 8.3 points between groups was assumed, with a standard deviation of 9.9 points, based on Leddy-Stacy and Rosenheck [24]. Using a 95% confidence level and 80% statistical power, the required sample size was 24 patients per group. To account for potential attrition, this number was increased by 30%, resulting in an initial target of 32 patients per group (31 in the IPS group due to one last-minute withdrawal).

A total of 557 participants were recruited, of whom 494 were excluded for different reasons (see flowchart, Figure 1). The high number of participants who declined participation (n = 250), was primarily related to the lingering effects of the COVID-19 context, which limited willingness to engage in face-to-face activities, along with other factors such as concerns about stigma, self-perceived inability to work, low motivation to engage in employment programs, and practical barriers (e.g., limited time availability, transportation difficulties, and caregiving responsibilities). The remaining 63 eligible participants were randomly and blindly assigned to either IPS group (n = 31) or the VR group (n = 32). During the follow-up, 7 participants in the IPS group and 13 in the VR group discontinued participation, leaving 24 and 19 participants, respectively, for the final analysis. All the participants gave their verbal and written informed consent to participate in the study.

The inclusion criteria used were the following: age between 18 and 65 years, diagnosis of schizophrenia (ICD-10 diagnoses), capacity to consent to voluntary participation in the study, being clinically stable without acute psychotic symptoms or having been hospitalized during the last months, being followed by the public community mental health teams during the study and being motivated to obtain a regular job.

The exclusion criteria used were having a common mental and/or other organic disorder, having a history of head injury with loss of consciousness, and being pregnant or breastfeeding.

The criteria foreseen for the withdrawal of subjects from the study were the explicit request by the subject to stop participating in the study, difficulty traveling to the place of the interviews, scheduling conflicts, change of place of residence outside the autonomous community where the study was carried out or death.

IPS was first implemented in Tenerife in 2004, following the scientific literature of the Dartmouth IPS Center. Since then, the program has been coordinated by the Insular Council of Psychosocial Rehabilitation and Community Action of the Canary Health Service together with Simpromi S.L., in collaboration with the University of La Laguna. The program has stable public funding, integration with mental health services, and has been described in previous publications [25, 26]. Its consolidation was further strengthened by hosting the first IPS Spain meeting, which paved the way for its subsequent inclusion in the IPS International Learning Community (IPS Europe, 2023) [27].

Fidelity to the IPS model was ensured by consistently adhering to the core IPS principles (zero exclusion, focus on competitive employment, integration with clinical services, rapid job searches, individualized support). As reported in [26], a fidelity assessment conducted in 2015 using the Supported Employment Fidelity Scale (IPS-15) yielded a score of 66 out of 75, corresponding to good implementation. In addition, a more recent fidelity assessment was conducted in 2020, prior to the current study period, using the Supported Employment Fidelity Scale by Becker et al. (IPS-25) [28]. This evaluation, performed by an independent psychologist not affiliated with the IPS team, yielded a score of 106 out of 125, consistent with good fidelity to the IPS model (unpublished data).

The principal investigator (PI), who was blinded to group assignment, assessed mental health, clinical and functional outcomes at baseline and at six-month follow-up. The following instruments were used:

Global Assessment of Functioning Scale (GAF) [29]. GAF is a numerical and descriptive scale that provides a single score reflecting a patient’s overall psychological, social, and occupational functioning. Scores range from 0 to 100, with higher scores indicating better functioning.

Lancashire Quality of Life Test and Satisfaction with Life Scale (ESV) [30]. This is a structured interview to measure the health and wellbeing of individuals with mental disorders. Subjective ratings are made using a visual analog scale (Life Satisfaction Scale) (1 = worst, 7 = best). The instrument also assesses global wellbeing and the interviewer scores the patient’s quality of life and makes an estimate of the reliability of the responses.

Positive and Negative Syndrome Scale (PANSS) [31]. It assesses symptomatology. The PANSS is an interview and observation-based scale consisting of 30 items: 7 for the positive scale (PANSS-P), 7 for the negative scale (PANSS-N) and 16 for the general psychopathology scale (PANSS-PG). Each item is evaluated according to a Likert scale of 7 degrees of intensity, where 1 is always equivalent to the absence of symptoms and 7 represents extreme severity.

Anxiety-Depression Scale (HADS) [32]. It is a self-administered questionnaire of 14 items, composed of two subscales of 7: one for anxiety (odd items) and another for depression (even items). The intensity or frequency of the symptom over the previous week is evaluated on a 4-point Likert scale (range 0–3).

Groningen Social Disabilities Schedule Test (GSDS) [33]. Based on social role theory, the GSDS evaluates social functioning across multiple domains, including self-care, family relationships, marriage, parenthood, citizenship, social, and occupational roles.

AF5 self-concept test [34]. It consists of 30 items, equally divided into 5 dimensions of self-concept: academic/work, social, family, emotional and physical. Responses are recorded on a scale from 1 (strongly disagree) to 99 (strongly agree).

In addition to the variables obtained from the aforementioned instruments, relevant clinical history data were collected from participants, including hospitalizations (number of admissions to short- or medium-stay beds during the past 6 months, as determined by the treating psychiatrist) and relapses (defined as exacerbations of symptoms measured by the PANSS).

The study was approved by the ethics committee (CI 2024-669) and conducted in accordance with the Declaration of Helsinki and European/Spanish legislation. Written informed consent was obtained from all participants, ensuring confidentiality and data protection.

Participants were referred from public CMHTs to IPS teams, which consisted of employment specialists from SINPROMI (Sociedad Insular para la Promoción con Discapacidad). The principal investigator (PI) was independent from these services and blinded to group assignment during recruitment and baseline assessments. Randomization occurred only after informed consent and baseline evaluation, minimizing selection bias.

Initial interviews collected sociodemographic level, educational level, previous work history and comorbidities. Participants were informed that employment could be obtained at any time, either during IPS or following VR. In addition, previous employment information, income and hospitalizations prior to the study were collected. Validated scales for mental health and functional outcomes were administered at baseline and repeated at six-month follow-up, together with employment and hospitalization data.

Data from the surveys were entered into an Excel spreadsheet and statistical analysis was performed using the Statistical Package for the Social Sciences (SPSS) Program (version 25, 2017) and GraphPad Prism 9.

To describe the qualitative variables included, the count and percentage were calculated, while the mean and standard deviation were used for the continuous variables. During the baseline phase, the comparison of the continuous variables was performed with the t-student test for independent samples, after verifying that they followed a normal distribution (Kolmogorov-Smirnov test). The chi-square test was used to compare qualitative variables, and in the case of 2 × 2 tables in which the expected frequency was less than 5, Fisher’s exact test was used.

The comparison of the information obtained in the baseline and follow-up phases for the quantitative variables of the study was carried out with a repeated measures analysis of variance (RM-ANOVA). Group and the interaction between group and time (baseline and follow-up) were introduced as a factor. In addition to statistical significance, effect sizes (Cohen’s d) were calculated to quantify the magnitude of within-group and between-group changes over time.

Cohen’s Kappa coefficient of concordance was used to analyze the evolution of the categorical variables between baseline and 6 months for the GSDS scale. In those cases, in which only two categories (“excellent” and “satisfactory”) were given, McNemar’s test was additionally calculated.

A multivariate logistic regression model was used to analyze employability data, with explanatory variables selected using the Backward Wald method.

All statistical tests were two-tailed, with a significance threshold of p < 0.05. Given the exploratory nature of this trial and the modest sample size, no correction for multiple comparisons was applied. Findings should therefore be interpreted with caution and considered preliminary.

Mental health and functional outcomes are summarized in Supplementary File 1. Overall, IPS participants showed greater improvements in global functioning (GAF) compared with VR, while other non-vocational outcomes-quality of life, symptomatology (PANNS), anxiety and depression (HADS), social functioning (GSDS), and self-concept (AF)- did not differ significantly between groups. Improvements over time were noted in both groups, particularly in GAF and quality of life, suggesting that participation in either program may have beneficial effects. The only non-vocational outcome showing a group effect was GAF, with significantly better scores in the IPS group, consistent with its inclusion of functional domains related to employment.

As shown in Table 2 and Supplementary File 2, repeated-measures ANOVA indicated that, among participants who completed follow-up, PANSS scores decreased in both groups at 6 months, reflecting an overall reduction in psychotic symptoms, anxiety, and depression. A significant time × group interaction was observed for social self-concept (p = 0.014), participants in the IPS group improved over time, whereas those in the VR group showed a decline. For all other non-vocational variables assessed (quality of life, other self-concept dimensions, and social functioning), no significant group differences were observed, indicating that improvements over time were similar in both IPS and VR participants. Effect sizes (Cohen’s d) for within-group and between-group changes over time are presented in Supplementary File 3. Consistent with the inferential results, medium to large effect sizes were observed for improvements in overall functioning (GAF) and in the social self-concept domain, favoring the IPS group. Within-group analyses also showed large effects for reductions in psychotic symptoms across both groups.

The relationship between baseline and six-month scores on each scale showed higher Pearson correlation coefficients in the IPS group compared with VR (GAF: 0.855 vs. 0.772; PANSS positive: 0.897 vs. 0.518; PANSS negative: 0.658 vs. 0.488; PANSS global: 0.761 vs. 0.668; HADS global: 0.798 vs. 0.603, respectively). These findings suggest a stronger association between initial functioning and subsequent improvement among participants receiving IPS.

High concordance values (close to 1) were observed between baseline and six-month evaluations, indicating that employment interventions did not produce changes or worsen patients’ conditions, as reflected in GSDS domain scores. No significant differences were found between the groups.

When comparing completers with dropouts (Supplementary File 4), participants who discontinued the study had lower scores on the PANSS negative and general subscales, as well as on HADS depression. This pattern suggests that individuals with fewer symptoms were more likely to drop out, whereas those with more severe symptomatology tended to remain in the study.

Employment rates were markedly higher in the IPS group, where 18 of 24 participants (75.0%) obtained competitive employment, compared with 3 of 19 participants (15.8%) in the VR group (χ² = 14.880, p < 0.001; Supplementary File 5). Logistic regression indicated that IPS participants were 16 times (95%CI 3.43, 74.70; p < 0.001) more likely to obtain competitive employment than those in VR.

To further explore whether employment status influenced mental health and functional outcomes, repeated-measures ANOVA was conducted (see Table 3). Being employed was associated with greater improvements in global functioning (GAF) and quality of life over time. Specifically, quality of life increased across both employed and unemployed participants (p = 0.026), but those who obtained employment experienced a significantly greater improvement (p = 0.043). Similarly, GAF showed significant changes over time (p = 0.001) and between groups (p = 0.001), with the largest gains among those who achieved employment.

For symptomatology, reductions were observed in both employed and unemployed participants across the three PANSS subscales (positive, negative, and general) over time (all p < 0.001). However, no significant group differences were found between those who obtained employment and those who did not (p = 0.718, p = 0.601, p = 0.933, respectively), indicating that employment status did not lead to symptom exacerbation or hospitalizations.