Results

Study characteristics

A total of 1021 articles were initially identified through the literature search. Following removal of duplicates and screening of titles and abstracts, 14 studies were retained for full-text review (see Supplementary Table S5 and Fig. 1). These studies, encompassing approximately 2555 participants, assessed the measurement properties of the S-QoL18, S-QoL41, and S-QoL-MCAT across diverse populations with schizophrenia and across multiple geographic regions.

Fig. 1: Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram of study selection.

A summary of the general characteristics of the included studies is presented in Table 2. The 14 included studies encompassed a range of populations with schizophrenia, including homeless individuals and patients with cognitive impairment, across multiple countries (France, Taiwan, Morocco, Bolivia, Chile, and Peru). Sample sizes ranged from 41 to 517 participants, ages varied from 29.8 to 38.6 years on average, and the instruments were primarily self-administered, with some studies using mixed administration modes. Languages of administration included French, Spanish, Chinese, and Moroccan Arabic. Across studies, samples were predominantly male (70%), with reported proportions of male participants ranging from 58% to 100%.

Author | Country | Population | Sample (%male/%female) | Age (years), mean (SD) | Language | PROM
Auquier, 20031 | France | Schizophrenia | 207 (70/30) | 37.3 (10.9) | French | S-QoL41
Reine, 200516 | France | Schizophrenia | 207 (68/32) | 37.4 (10.9) | French | S-QoL41
Boyer, 20102 | France | Schizophrenia | 517 (70/30) | 36.5 (10.8) | French | S-QoL18
Chou, 201112 | Taïwan | Schizophrenia | 41 (70/30) | 37.9 (7.36) | Chinese | S-QoL41
Auquier, 20137 | France | Homeless individuals with Schizophrenia | 236 (58/42) | 37.9 (10.8) | French | S-QoL18
Baumstark, 20138 | France | Individuals with Schizophrenia and cognitive impairment | 113 (70/30) | 38.6 (10.8) | French | S-QoL18
Baumstark, 20149 | France | Individuals with Schizophrenia and cognitive impairment | 113 (70/30) | 38.6 (10.8) | French | S-QoL18
Caqueo-Urizar, 201411 | Bolivia, Chile, Peru | Schizophrenia | 253 (65/35) | Bolivia: 33.7 (11.3)Chile: 37.9 (15.3)Peru: 35.2 (10.3) | Spanish | S-QoL18
Chou, 201513 | Taïwan | Schizophrenia | 41 (60/40) | 37.9 (7.36) | Chinese | S-QoL18
Faget-Agius, 201523 | France | Schizophrenia | 81 (100/0) | 30 (7.5) | French | S-QoL18
Faget-Agius, 201624 | France | Schizophrenia | 131 (73/27) | 35.8 (11) | French | S-QoL18
Girard, 201625 | France | Homeless individuals with Schizophrenia | 55 (72/28) | <30 | French | S-QoL41
Michel, 20186 | France | Schizophrenia | 517 (70/30) | 36.5 (10.9) | French | MCAT
Kachouchi, 202626 | Morocco | Schizophrenia | 43 (74/26) | 29.8 (8.07) | Moroccan Arabic | S-QoL18

PROM development and content validity of Schizophrenia Quality of Life questionnaire (S-QoL)

The Schizophrenia Quality of Life questionnaire (S-QoL41) was developed using a patient-centred approach based on Calman's HRQoL model, which defines QoL as the gap between expectations and lived experience.1 Initial items were generated through interviews with 20 patients of varying clinical profiles, leading to 97 items after achieving conceptual saturation. Pretesting in 40 patients reduced the pool to 85 items. Further refinement using expert review and psychometric analyses, including Rasch modelling, ensured unidimensionality. The final instrument includes 41 items across eight dimensions: psychological well-being, self-esteem, relationships, resilience, physical well-being, autonomy, and sentimental life. Cognitive interviews further optimised clarity, relevance, and response formats prior to large-scale validation.1

The shortened version (S-QoL18) was developed using pooled data from four multicentre studies (n = 517). Item reduction combined classical test theory, item response theory, and expert review to maintain conceptual coverage while reducing burden. Differential item functioning (DIF) analyses confirmed stability across demographic and clinical subgroups.2 Additional studies in patients with cognitive impairments supported preserved dimensionality, item relevance, and overall comprehensibility.8,9

The computerized adaptive version (S-QoL-MCAT) was derived entirely from the S-QoL41 item bank, with no new items or domains. Item wording was unchanged; only item selection and order were adapted, with a mean of 25 items administered (SD = 5; SEM < 0.55). The original eight-factor structure was confirmed using multidimensional item response theory, supporting preserved content validity consistent with the paper-based versions6 (Supplementary Table S6). Findings on each measurement properties are also summarised in Supplementary Table S7.

Structural validity

The original eight-factor structure of S-QoL41 was consistently supported across studies. Auquier et al., (2003)1 identified an eight-factor solution explaining approximately 52% of the total variance, and subsequent confirmatory factor analysis (CFA) confirmed adequate model fit (root mean square error of approximation [RMSEA] = 0.035, comparative fit index [CFI] = 0.95, goodness-of-fit index [GFI] = 0.99, and standardized root mean square residual [SRMR] = 0.015). Similar results were reported for the computerized adaptive version, with retention of the original dimensions and acceptable fit indices (RMSEA = 0.05, CFI = 0.95).6

For S-QoL18, structural validity was supported by factor analyses, Rasch modelling, and item-fit analyses. Rasch INFIT statistics were within acceptable ranges across all eight dimensions, supporting unidimensionality and adequate item functioning.2,7, 8, 9,11 The eight-factor structure was also maintained in patients with executive dysfunction and broader cognitive impairment, with acceptable CFA and item-fit results.8,9 For S-QoL-MCAT, multidimensional item response theory and CFA likewise supported the original eight-factor structure (RMSEA = 0.05, CFI = 0.95).6

Internal consistency

Internal consistency was evaluated across multiple studies and populations. For S-QoL41, Cronbach's α ranged from 0.72 to 0.95 for total scores and from 0.54 to 0.93 for subscales, with the lowest values reported for Self-esteem (α = 0.54) and Autonomy (α = 0.64) in smaller samples.1,12,14,16 For S-QoL18, total score Cronbach's α ranged from 0.72 to 0.90, and most subscales showed α ≥ 0.70, although lower values were reported for Resilience (α < 0.50) and Self-esteem (α = 0.51) in specific samples.2,7, 8, 9,11,21,26 Comparable Cronbach's α coefficients were observed among patients with executive dysfunction and broader cognitive impairment.8,9 For S-QoL-MCAT, empirical marginal reliability coefficients ranged from 0.80 to 0.92 across dimensions.6

Cross-cultural validity/measurement invariance (COSMIN)

Evidence for cross-cultural validity was limited for S-QoL41, as most studies did not assess differential item functioning (DIF) or measurement invariance.1,12,14,16,22 For S-QoL18, DIF was identified across gender, age, education, or clinical form,2 whereas item functioning was comparable between homeless and non-homeless participants.7 One study identified DIF for a single item (“plan for future”) across countries,11 while another confirmed content validity following translation into Moroccan Arabic.26 The Chinese version retained the original eight-factor structure, although DIF was not assessed.13 For S-QoL-MCAT, negligible DIF was observed in only 6 of 123 tests, supporting invariance across gender, insight, and paranoia.6

Reliability

Test–retest reliability was assessed for S-QoL41 and S-QoL18. For S-QoL41, intraclass correlation coefficients (ICCs) ranged from 0.64 to 0.76 across subscales, with an ICC of 0.79 for the total S-QoL Index.1 Similar results were reported by Chou (2011), with ICCs ranging from 0.64 to 0.87,12 and significant test–retest correlations over 7–30 days were also observed.22

For S-QoL18, ICCs ranged from 0.50 to 0.78 across subscales, with an ICC of 0.73 for the total Index.2 Chou (2015) reported ICCs ranging from 0.58 to 0.84 across subscales and 0.87 for the total Index.13 For S-QoL-MCAT, test–retest reliability was not assessed.6 Inter-rater reliability was not evaluated in any study.

Measurement error

No study formally evaluated measurement error according to COSMIN criteria. Although Michel et al., (2018)6 reported RMSE and standard error of measurement (SEM) for S-QoL-MCAT, these estimates reflected score precision within the item response theory/computerized adaptive testing framework rather than measurement error as defined by COSMIN. No studies reported measurement error parameters such as SEM, smallest detectable change (SDC), minimal detectable change (MDC), or limits of agreement.

Construct validity (hypotheses testing)

For convergent validity, correlations between S-QoL instruments and QoL measures, recovery measures, self-esteem scales, and clinical indicators were generally consistent with the predefined hypotheses for S-QoL41, S-QoL18, and S-QoL-MCAT.1,2,6,8,9,11, 12, 13, 14,16,21,22

Known-groups validity was supported by the ability of S-QoL41 and S-QoL18 to discriminate between groups differing in symptom severity, remission status, depressive symptoms, functioning, cognitive impairment, hospitalisation history, and housing status.2,7,8,10,15 Associations with demographic variables were generally weak or absent, as hypothesised (Supplementary Table S8). Overall, 84.5% of the predefined hypotheses across construct validity analyses were confirmed.

Responsiveness/sensitivity to change

For S-QoL41, longitudinal studies showed changes consistent with clinical improvement, with significant improvements and small-to-moderate effect sizes (effect size [ES] ≥ 0.20) observed for the total score and key domains.1,16 Some domains showed smaller or non-significant changes, in line with predefined hypotheses (Supplementary Table S9). For S-QoL18, significant improvements were reported across multiple studies, particularly in Psychological Well-being, Self-esteem, Social Relationships, Sentimental Life, and total scores, with effect sizes generally meeting or exceeding the predefined threshold (ES ≥ 0.20).2,7,13 Group comparisons based on clinical improvement, symptom remission, and hospitalisation status were consistent with expected hypotheses. No responsiveness data were available for S-QoL-MCAT; therefore, this measurement property could not be evaluated for this version. Overall, 89% of the predefined hypotheses across responsiveness analyses were confirmed.

Item score distribution, floor and ceiling effects

Item score distributions and floor/ceiling effects were evaluated across S-QoL instruments. For S-QoL41, floor and ceiling effects were generally low (<10–10%) across most dimensions, although higher values were reported for the Sentimental Life dimension in some studies (up to 16.9%) and with wider ranges observed across subscales in individual samples.1,12,14,22 For S-QoL18, floor effects were generally low (<10–20%) across most dimensions, with higher values observed in domains such as Relationships with Friends, Relationships with Family, and Sentimental Life. Ceiling effects were typically low (<10–12%), with slightly higher values in Psychological Well-being and more impaired populations.2,7, 8, 9,11,13 Data on S-QoL-MCAT were not available.

Interpretability and feasibility

According to COSMIN criteria (>15–20% indicating problematic floor or ceiling effects), S-QoL41 generally showed acceptable score distributions across studies, although higher floor and ceiling effects were observed in some domains, particularly Sentimental Life and Psychological Well-being (up to 30.9% and 41.6%, respectively).1,12,14,22 For S-QoL18, most studies reported floor and ceiling effects below 15–20%, although higher values were observed in specific domains and populations (up to 38% and 35.1%, respectively).2,7, 8, 9,11,13 Floor and ceiling effects were not reported for S-QoL-MCAT. Regarding feasibility, S-QoL18 reduced respondent burden compared with S-QoL41, with shorter completion time (<5 min vs 13.6 min) and low missing data rates (2–10%).1,2 S-QoL-MCAT provides adaptive electronic administration with high measurement precision, although data on score distributions were not available.6 Information on implementation requirements (e.g., training, cost, and administrative burden) was limited across studies.

Overall quality of evidence

Overall, S-QoL instruments demonstrated sufficient measurement properties, with strongest evidence for structural validity, internal consistency, construct validity, reliability, and responsiveness across S-QoL41, S-QoL18, and S-QoL-MCAT. Certainty of evidence was generally moderate to high for S-QoL41 and S-QoL18, and moderate for S-QoL-MCAT (Supplementary Table S10).

Grading the quality of evidence

Most measurement properties were supported by studies of adequate methodological quality, with no downgrading for risk of bias. Evidence for cross-cultural validity and measurement error remained limited due to the absence of formal measurement invariance and measurement error analyses, and was therefore considered indeterminate. Minor inconsistencies across studies in factor structure and responsiveness led to a one-level downgrade for inconsistency. No downgrading was applied for imprecision or indirectness. Collectively, the evidence supports the use of S-QoL instruments as valid, reliable, and responsive measures of health-related quality of life in schizophrenia, while highlighting the need for further research on cross-cultural measurement properties and measurement error.

COSMIN-based recommendation for use

According to COSMIN 2024 criteria, S-QoL41, S-QoL18, and S-QoL-MCAT can be recommended for use in people with schizophrenia. S-QoL18 provides the most favourable balance between measurement performance and feasibility, while retaining the original multidimensional structure with reduced respondent burden. S-QoL41 remains appropriate when a comprehensive assessment of quality of life domains is required, particularly in research settings. S-QoL-MCAT offers an efficient computerized adaptive alternative with high measurement precision, although some measurement properties are less extensively evaluated compared with the paper-based versions.

Exploratory biological context

Neuroimaging evidence was additionally synthesised as a complementary and contextual layer to support the theoretical plausibility of S-QoL constructs. In a functional Single Photon Emission Computed Tomography (SPECT) study of 130 patients with schizophrenia, scores on S-QoL18 dimensions were associated with regional cerebral blood flow in areas implicated in emotion processing, social cognition, and decision-making, including Brodmann areas 6, 8, 9, 10, the striatum, and the praecuneus.24 Similarly, a magnetisation transfer imaging study demonstrated associations between S-QoL scores and microstructural integrity in the temporal lobes, insula, cerebellum, and occipital cortex.23 Given differences in imaging modalities (SPECT vs magnetisation transfer MRI), analytic approaches, and sample characteristics, these findings should be interpreted as exploratory and hypothesis-generating rather than confirmatory evidence of measurement validity. Overall, neuroimaging evidence provides complementary neurobiological support for the multidimensional QoL construct in schizophrenia.