Interrelationships of long-term changes in recovery domains among patients with schizophrenia spectrum disorders: a six-year follow-up study. Authors: Lars de Winter, Auke Jelsma, Jentien M. Vermeulen, Astrid Vellinga, Marieke van der Pluijm, Jaap van Weeghel, Ilanit Hasson-Ohayon, Cornelis L. Mulder, Nynke Boonstra, Wim Veling, Lieuwe de Haan, GROUP investigators Published: 2025 ABSTRACT BackgroundSocial functioning, symptoms, cognition and well-being are all considered important recovery domains for patients with schizophrenia spectrum disorders (SSD). In the current study, we investigated to what extent longitudinal changes in one recovery domain moderated changes in another recovery domain, and which personal, clinical or social patient characteristics influence changes in recovery domains for patients with SSD.MethodsWe investigated data from GROUP, a prospective, longitudinal six-year study of patients with SSD. We analyzed longitudinal changes in 15 outcomes (i.e., subdomains of symptoms, social functioning, cognition, quality of life or well-being), through a linear mixed-effects model analysis. Changes in other outcomes, patients’ duration of illness (DOI), and personal, clinical, or social patient characteristics at baseline were treated as covariates in the analyses.ResultsWe observed long-term improvement in all outcomes, except for depressive symptoms and subjective well-being. Larger improvements in positive symptoms, social functioning and executive functioning were indicated for patients with a short DOI. We found multiple, moderating effects, mainly for patients with a short DOI, between changes in cognition, social functioning, symptoms and quality of life. Limited moderating effects were indicated between these recovery domains and well-being. Baseline levels of outcomes, schizophrenia diagnosis, education level, ethnicity, living situation, employment status and marital status were all moderators for changes in most outcomes.ConclusionsSymptoms, social functioning, cognition and quality of life often concurrently improved over the course of SSD. These longitudinal changes were mostly driven by social patient characteristics, such as education level, ethnicity, living situation, employment status and marital status.Clinical trial numberNot applicable.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12888-025-07700-8. Background Schizophrenia spectrum disorders (SSD) are characterized by disturbances in perception, organization of thoughts, and negative symptoms [1]. People with SSD often struggle in their daily life due to problems with societal functioning, cognitive functions, symptoms and well-being [2–6]. The prognosis of improvement over time in SSD differs. Previous studies indicated large improvements in positive symptoms, small to moderate improvement in negative and depressive symptoms, moderate improvement in social functioning, and only marginal or no improvements in cognition and well-being over time in patients with SSD [6–11]. Longitudinal changes are also influenced by a variety of clinical, social and personal patient characteristics, especially by the patient’s duration of illness (DOI), defined as the duration since the first classification of SSD. Previous studies indicated that patients with a short DOI showed larger improvement in symptoms, social functioning and cognition than patients with longer DOI [7–9, 12]. The larger improvement was specifically indicated in subgroups of patients with a DOI lower than 5 years. In addition to DOI, previous research indicated that male sex, older age, schizophrenia diagnosis, higher levels of negative symptoms, ethnic minority status, and low social economic status were associated with less improvement in various outcomes, such as symptoms, social functioning, cognition, or well-being [3, 7–11, 13–18]. Finally, previous studies showed cross-sectional associations between levels of social functioning, symptoms, cognition and well-being [3, 5, 18–21]. However, these associations were not found in a longitudinal design. In the current study, we provided new insights into the longitudinal interrelationships between the following four clusters of outcomes, from now on indicated as recovery domains: symptoms, social functioning, cognition and quality of life or well-being. We investigated if longitudinal changes in one recovery domain moderate longitudinal changes in other recovery domains, and which personal, clinical or social patient characteristics influence longitudinal changes in the recovery domains. Furthermore, we specifically analyzed moderating effects between changes in recovery domains within specific subgroups of patients based on their DOI. We aimed to answer the following questions: 1) To what extent do changes in one recovery domain moderate changes in other recovery domains? 2) Do moderating effects of changes in recovery domains differ between subgroups of patients with a different DOI? 3) Which personal, clinical or social patient characteristics at baseline further explain changes in different recovery domains? Methods Study design For this study, we used data from the Genetic Risk and Outcome of Psychosis (GROUP) study [22]. GROUP is a multisite longitudinal cohort study included data from patients and their parents and siblings, from 36 mental health care institutes in the Netherlands and Dutch-speaking Belgium. In this study, we exclusively focused on the patient subgroup. Participants were assessed at baseline, three-year and six-year follow-up. The study protocol was approved by the Ethical Review Board of the University Medical Centre Utrecht. Written informed consent was obtained from all participants before the baseline assessment. Patient population We included patients that met the following inclusion criteria: 1) age of 18 years or older; 2) a main diagnosis within the classification of schizophrenia spectrum disorders (SSD [23]). A distribution of the main diagnoses is included in Table 1; and 3) good command of Dutch language. In the GROUP study, a total number of 1119 patients were recruited [22], of whom 1008 met our inclusion criteria.Table 1Descriptive statistics of study sample at baselineDemographic variables*M (SD)RangeN Study sampleAge of the participant27.7 (7.5)18–611008Age of the participant at illness onset23.0 (6.5)2–51997Duration of illness (in years)4.7 (4.5)1–41997Duration of untreated psychosis (in months)10.8 (29.5)0–395867Number of psychotic episodes1.8 (1.2)1–81007N participants%N Study sampleMain diagnosisBrief psychotic disorderDelusional disorderPsychotic disorder due to medical conditionPsychotic disorder NOSSchizoaffective disorderSchizophreniaSchizophreniform disorder24211100117690552.4%2.1%0.1%9.9%11.6%68.5%5.5%1008ComorbiditySubstance use disorderBipolar disorderMajor depressive disorderAnxiety disorderPersonality disorderOther diagnoses29192952112.9%1.9%2.9%0.5%0.2%1.1%1008DOI subgroup< 2 years2–5 years > 5–10 years > 10 years2594322367026.0%43.3%23.7%7.0%997LocationAmsterdamGroningenMaastrichtUtrecht26924827621526.7%24.6%27.4%21.3%1008Education level (ISCED)ISCED 0: early childhood educationISCED 1: primary educationISCED 2: lower secondaryISCED 3: upper secondaryISCED 4: post-secondary non tertiaryISCED 5: Short-cycle tertiaryISCED 6–8: Bachelor’s, Master’s, Doctoral or equivalent level712711536625701310.7%12.7%11.5%36.5%25.6%0.0%13.1%1003EthnicityWhiteOther ethnic background77820978.8%21.2%987SexMaleFemale78022877.4%22.6%1008Living situationLiving aloneLiving with parentsLiving with partnerSheltered livingOther living situation32535796966334.7%38.1%10.2%10.2%6.7%937Marital statusNot married or living togetherMarried or living togetherDivorced842883087.7%9.2%3.1%960Antipsychotic use (%)87599.4%880SymptomsM (SD)RangeN Study samplePANSS positive symptoms score13.9 (6.6)2–41946PANSS negative symptoms score15.0 (6.6)3–39935PANSS overall symptoms score54.7 (16.8)30–120967PANSS disorganization16.8 (6.3)10–44938PANSS Hostility and excitement12.0 (4.0)8–38957PANSS general psychopathology score28.0 (8.5)16–65967Depressive symptoms: CDS total score**2.0 (2.9)0–16699Social FunctioningGAF total score54.4 (16.0)1–100931SFS total score**126 (22.8)50–189681SFS Independence competence**29.8 (6.1)11–39681SFS Independence performance**36.4 (3.5)11–39683N participants%N Study sampleEmployment statusBoth competitive and voluntary employmentCompetitive employment onlyVoluntary employment onlyNo competitive or voluntary employment10331710942110.8%33.4%11.5%44.3%950CognitionM (SD)RangeN Study sampleAttention: CPT mean RT correct responses432.1 (84.4)211.3–872.5899Executive functioning: response shifting task1.6 (5.6)−0.5–39871Intelligence: WAIS IQ score94.9 (16.0)57–146957Motor skills: WAIS block design9.6 (3.3)1–18961Overall cognition: WAIS-III total scaled score37.1 (9.8)12–64957Processing speed: WAIS digit symbol substitution7.3 (3.0)1–19961Verbal memory: 15 WLT total items correct immediate recall22.8 (6.1)3–41958Well-beingSWN-20 total score64.7 (7.2)25–98509WHO-QOL score3.4 (1.0)1–5907*CDS = Calgary Depression Scale for Schizophrenia; CPT = Continuous Performance Test; DOI = Duration of illness; GAF = Global Assessment of Functioning; ISCED = International Standard Classification of Education (European Commission, 2011); NOS = Not otherwise specified; PANSS = Positive and Negative Syndrome Scale; RT = Reaction time; SFS = Social Functioning Scale; SWN = 20 = The 20-item Subjective Well-being under Neuroleptics Scale; WAIS = Wechsler Adult Intelligence Scale; WLT = Word Learning Test; WHO-QoL = World Health Organization Quality of Life scale**First assessment of CDS and SFS took place at T2 (3-year follow-up) and is considered as baseline assessment in this overview Table 1Descriptive statistics of study sample at baselineDemographic variables*M (SD)RangeN Study sampleAge of the participant27.7 (7.5)18–611008Age of the participant at illness onset23.0 (6.5)2–51997Duration of illness (in years)4.7 (4.5)1–41997Duration of untreated psychosis (in months)10.8 (29.5)0–395867Number of psychotic episodes1.8 (1.2)1–81007N participants%N Study sampleMain diagnosisBrief psychotic disorderDelusional disorderPsychotic disorder due to medical conditionPsychotic disorder NOSSchizoaffective disorderSchizophreniaSchizophreniform disorder24211100117690552.4%2.1%0.1%9.9%11.6%68.5%5.5%1008ComorbiditySubstance use disorderBipolar disorderMajor depressive disorderAnxiety disorderPersonality disorderOther diagnoses29192952112.9%1.9%2.9%0.5%0.2%1.1%1008DOI subgroup< 2 years2–5 years > 5–10 years > 10 years2594322367026.0%43.3%23.7%7.0%997LocationAmsterdamGroningenMaastrichtUtrecht26924827621526.7%24.6%27.4%21.3%1008Education level (ISCED)ISCED 0: early childhood educationISCED 1: primary educationISCED 2: lower secondaryISCED 3: upper secondaryISCED 4: post-secondary non tertiaryISCED 5: Short-cycle tertiaryISCED 6–8: Bachelor’s, Master’s, Doctoral or equivalent level712711536625701310.7%12.7%11.5%36.5%25.6%0.0%13.1%1003EthnicityWhiteOther ethnic background77820978.8%21.2%987SexMaleFemale78022877.4%22.6%1008Living situationLiving aloneLiving with parentsLiving with partnerSheltered livingOther living situation32535796966334.7%38.1%10.2%10.2%6.7%937Marital statusNot married or living togetherMarried or living togetherDivorced842883087.7%9.2%3.1%960Antipsychotic use (%)87599.4%880SymptomsM (SD)RangeN Study samplePANSS positive symptoms score13.9 (6.6)2–41946PANSS negative symptoms score15.0 (6.6)3–39935PANSS overall symptoms score54.7 (16.8)30–120967PANSS disorganization16.8 (6.3)10–44938PANSS Hostility and excitement12.0 (4.0)8–38957PANSS general psychopathology score28.0 (8.5)16–65967Depressive symptoms: CDS total score**2.0 (2.9)0–16699Social FunctioningGAF total score54.4 (16.0)1–100931SFS total score**126 (22.8)50–189681SFS Independence competence**29.8 (6.1)11–39681SFS Independence performance**36.4 (3.5)11–39683N participants%N Study sampleEmployment statusBoth competitive and voluntary employmentCompetitive employment onlyVoluntary employment onlyNo competitive or voluntary employment10331710942110.8%33.4%11.5%44.3%950CognitionM (SD)RangeN Study sampleAttention: CPT mean RT correct responses432.1 (84.4)211.3–872.5899Executive functioning: response shifting task1.6 (5.6)−0.5–39871Intelligence: WAIS IQ score94.9 (16.0)57–146957Motor skills: WAIS block design9.6 (3.3)1–18961Overall cognition: WAIS-III total scaled score37.1 (9.8)12–64957Processing speed: WAIS digit symbol substitution7.3 (3.0)1–19961Verbal memory: 15 WLT total items correct immediate recall22.8 (6.1)3–41958Well-beingSWN-20 total score64.7 (7.2)25–98509WHO-QOL score3.4 (1.0)1–5907*CDS = Calgary Depression Scale for Schizophrenia; CPT = Continuous Performance Test; DOI = Duration of illness; GAF = Global Assessment of Functioning; ISCED = International Standard Classification of Education (European Commission, 2011); NOS = Not otherwise specified; PANSS = Positive and Negative Syndrome Scale; RT = Reaction time; SFS = Social Functioning Scale; SWN = 20 = The 20-item Subjective Well-being under Neuroleptics Scale; WAIS = Wechsler Adult Intelligence Scale; WLT = Word Learning Test; WHO-QoL = World Health Organization Quality of Life scale**First assessment of CDS and SFS took place at T2 (3-year follow-up) and is considered as baseline assessment in this overview Descriptive statistics of study sample at baseline Brief psychotic disorder Delusional disorder Psychotic disorder due to medical condition Psychotic disorder NOS Schizoaffective disorder Schizophrenia Schizophreniform disorder 24 21 1 100 117 690 55 2.4% 2.1% 0.1% 9.9% 11.6% 68.5% 5.5% Substance use disorder Bipolar disorder Major depressive disorder Anxiety disorder Personality disorder Other diagnoses 29 19 29 5 2 11 2.9% 1.9% 2.9% 0.5% 0.2% 1.1% < 2 years 2–5 years > 5–10 years > 10 years 259 432 236 70 26.0% 43.3% 23.7% 7.0% Amsterdam Groningen Maastricht Utrecht 269 248 276 215 26.7% 24.6% 27.4% 21.3% ISCED 0: early childhood education ISCED 1: primary education ISCED 2: lower secondary ISCED 3: upper secondary ISCED 4: post-secondary non tertiary ISCED 5: Short-cycle tertiary ISCED 6–8: Bachelor’s, Master’s, Doctoral or equivalent level 7 127 115 366 257 0 131 0.7% 12.7% 11.5% 36.5% 25.6% 0.0% 13.1% White Other ethnic background 778 209 78.8% 21.2% Male Female 780 228 77.4% 22.6% Living alone Living with parents Living with partner Sheltered living Other living situation 325 357 96 96 63 34.7% 38.1% 10.2% 10.2% 6.7% Not married or living together Married or living together Divorced 842 88 30 87.7% 9.2% 3.1% Both competitive and voluntary employment Competitive employment only Voluntary employment only No competitive or voluntary employment 103 317 109 421 10.8% 33.4% 11.5% 44.3% *CDS = Calgary Depression Scale for Schizophrenia; CPT = Continuous Performance Test; DOI = Duration of illness; GAF = Global Assessment of Functioning; ISCED = International Standard Classification of Education (European Commission, 2011); NOS = Not otherwise specified; PANSS = Positive and Negative Syndrome Scale; RT = Reaction time; SFS = Social Functioning Scale; SWN = 20 = The 20-item Subjective Well-being under Neuroleptics Scale; WAIS = Wechsler Adult Intelligence Scale; WLT = Word Learning Test; WHO-QoL = World Health Organization Quality of Life scale **First assessment of CDS and SFS took place at T2 (3-year follow-up) and is considered as baseline assessment in this overview Procedure Potentially eligible patients were given detailed explanations of the study procedure and were asked to sign informed consent. Assessments were administered on location or during in-home visits by trained assessors. Full assessments with the patients lasted on average four hours, divided into two separate meetings to account for fatigue effects. To prevent fatigue effects during cognitive tasks, assessors started with the neuropsychological testing. Materials We selected the following patient characteristics and assessment instruments: Patient characteristics Age and age at onset of first psychotic episode, duration of illness (DOI) and duration of untreated psychosis to first contact with mental healthcare (DUP), number of psychotic episodes, main and comorbid diagnoses (DSM-IV-TR [23]), assessment site, education level, employment status, ethnicity, sex, living situation, marital status and antipsychotic use. Symptoms Symptoms were assessed using the Positive and Negative Syndrome Scale (PANSS [24]). The PANSS assessed psychiatric symptoms, including positive, negative and general psychopathology symptoms in psychosis. Depressive symptoms were assessed through the Calgary Depression Scale (CDS [25]). Social functioning. Social functioning was assessed through the Global Assessment of Functioning scale (GAF [23]) and the Social Functioning Scale (SFS [26]). Cognition Cognition was assessed through multiple assessment scales. Attention was measured by the mean reaction time of correct responses in the Continuous Performance Task (CPT [27]). Executive functioning was assessed using the proportional cost index score of the response shifting task [28]. Motor skills were assessed through the block design task of the Wechsler Adult Intelligence Scale-III (WAIS-III [29]). Overall cognition was measured using the overall scaled score of WAIS-III [29]. Processing speed was assessed with the digit symbol substitution test of the WAIS-III [29]. Finally, verbal memory was measured by the number of correct responses of the immediate recall of the word learning task (WLT [30]). Quality of life or well-being Well-being was assessed with the sum score of all 20 items of the Subjective Well-being Under Neuroleptic Treatment questionnaire (SWN-20 [31]). We recoded all items of the SWN-20 in a way that a higher score indicated better subjective well-being. Furthermore, quality of life was assessed with the mean item score of the World Health Organization Quality of Life scale (WHO-QOL BREF [32]). Outcome assessments were conducted at three points of assessment over a six-year follow-up period: at baseline, 3-year follow-up and 6-year follow-up. An exception was made for CDS and SFS. These were only assessed at 3-year and 6-year follow-up only. Statistical analysis Statistical analyses were executed through SPSS version 29 [33]. First, we executed a descriptive analysis of the baseline patient characteristics and outcomes. Longitudinal analyses between baseline (T1), 3-year (T2) and 6-year follow-up (T3) were conducted using a linear mixed-effects model analysis. This analysis was most suitable for longitudinal analyses with more than 2 assessments, as it allowed inclusion of participants regardless of missing follow-up observations, and because it controls for the hierarchical nature of the data [34]. Assessment period was included as time variable (level 1), and participants were included as a subject variable (level 2). We also calculated effect sizes of changes in recovery domains (i.e., η2) and assessed the magnitude of effect [35]. Patients’ DOI, personal, clinical or social patient characteristics at baseline and changes in other outcomes between baseline and follow-up, were added as covariates in separate univariate mixed-model regression analyses (i.e., covariate ×time ×outcome interactions in the linear mixed-model analysis). We controlled for assessment site and demographic variables in each analysis based on the best model fit according to Akaike’s information criterion (AIC). Furthermore, we divided patients within specific DOI subgroups (DOI < 2 years; DOI 2–5 years; DOI 5–10 years; DOI > 10 years) and performed separate analyses of moderating effects within these subgroups. We exclusively analyzed moderating effects within DOI subgroups when these were shown to be significant in the general mixed-model regression analysis. Finally, due to the large number of variables included in the analyses of moderating effects there is a higher chance of finding false-positive findings due to multiple testing. Therefore, we controlled for these multiple testing effects by executing Benjamini-Hochberg corrections, with the false discovery rate set on 0.05 [36]. Results Patient characteristics Patient characteristics are presented in Table 1. We included 1008 patients with SSD. Their mean age at baseline was 27.7 years (SD = 7.5), 780 patients (77.4%) were male, and 778 (78.9%) had a white ethnicity (i.e., Dutch or Belgian origin). Furthermore, 421 patients (44.9%) lived independently, 88 (9.2%) were married or living together, 388 participants (38.7%) finished post-secondary or tertiary education, 420 patients (44.2%) were in competitive employment, and 875 participants (99.4%) used antipsychotics. Mean DOI was 4.7 years (SD = 4.5) and distributed as follows: 259 patients (26.0%) had a DOI < 2 years, 432 (43.3%) a DOI 2–5 years, 236 (23.7%) a DOI 5–10 years, and 70 patients (7.0%) had a DOI > 10 years. Follow-up data was not available for 269 patients (26.7%) after three-year follow-up and for 402 participants (39.9%) after six-year follow-up. Analysis of subgroup differences (see Supplementary Materials 1) showed that the drop-out sample had a significantly lower level of education (χ2 = 37.6; p < 0.01), a non-white ethnic background (χ2 = 32.5; p < 0.01) and were less frequently in competitive (χ2 = 5.1; p < 0.05) or voluntary employment (χ2 = 6.7; p < 0.05). Overall longitudinal changes in recovery domains As shown in Table 2, results showed a medium decrease in positive (η2 = 0.09), negative (η2 = 0.11), and disorganization symptoms (η2 = 0.12), but no significant changes in depressive symptoms over time. Furthermore, we found a medium improvement in GAF total scores (η2 = 0.10) and a small improvement in SFS total score (η2 = 0.02) and independence sub-score (η2 = 0.00). We also found a small improvement in attention (η2 = 0.02), medium improvement in processing speed (η2 = 0.10) and a large improvement in executive functioning (η2 = 0.25), motor skills (η2 = 0.17), overall cognition (η2 = 0.15) and verbal memory (η2 = 0.18). Finally, we found a medium decrease in subjective well-being (η2 = 0.06), but a medium improvement in subjective quality of life (η2 = 0.09).Table 2Longitudinal changes outcome domainsBaseline3-year follow-up6-year follow-upMixed model analysisSymptoms*M (SD)RangeNM (SD)RangeNM (SD)RangeNFdfpη2PANSS Positive symptoms13.9 (6.6)2–4194611.8 (5.8)3–3970012.4 (6.6)2–3856233.52 < 0.010.09PANSS Negative symptoms15.0 (6.6)3–3993512.7 (5.5)4–4169912.5 (5.9)4–4156243.32 < 0.010.11PANSS Disorganization symptoms16.8 (6.3)10–4493814.4 (5.3)11–4669914.7 (5.9)11–4555944.02 < 0.010.12CDS Depressive symptomsXXX2.0 (2.9)0–166992.2 (3.3)0–193140.2710.600.00Social functioning*GAF total score54.4 (16.0)1–10093160.4 (16.1)25–10062958.9 (16.6)21–10054533.72 < 0.010.10SFS total scoreXXX126 (22.8)50–189681128.4 (22.2)58–1875464.910.030.02SFS Independence performanceXXX36.4 (3.5)11–3968336.5 (3.7)0–3954626.01 < 0.010.00Cognition*Attention: CPT mean RT correct responses432.1 (84.4)211.3–872.5899443.1 (83.8)272.6–847.8692443.5 (82.5)256–722.35478.02 < 0.010.02Executive functioning: response shifting task1.6 (5.6)0–398711.7 (6.5)0–396390.2 (0.5)0–1.91934.52 < 0.010.25Motor skills: WAIS-III block design9.6 (3.3)1–1896110.5 (3.3)3–1865911.0 (3.8)1–1956662.12 < 0.010.17Overall cognition: WAIS-III total scaled score37.1 (9.8)12–6495739.5 (9.6)15–6365740.8 (10.2)12–6756355.12 < 0.010.15Processing speed: WAIS-III digit symbol substitution7.3 (3.0)1–199618.0 (3.1)1–186638.2 (3.1)1–1956433.82 < 0.010.10Verbal memory: 15 WLT total items correct immediate recall22.8 (6.1)3–4195825.4 (6.4)8–4366324.0 (6.8)5–4356371.12 < 0.010.18Well-being*SWN-20 total score64.7 (7.2)25–9850963.8 (6.1)44–9143162.5 (6.4)37–9428511.72 < 0.010.06WHO-QOL score3.4 (1.0)1–59073.7 (0.9)1–56893.8 (0.9)1–554032.22 < 0.010.09*CDS = Calgary Depression Scale for Schizophrenia; CPT = Continuous Performance Test; GAF = Global Assessment of Functioning; PANSS = Positive and Negative Syndrome Scale; RT = Reaction time; SFS = Social Functioning Scale; SWN = 20 = The 20-item Subjective Well-being under Neuroleptics Scale; WAIS = Wechsler Adult Intelligence Scale; WLT = Word Learning Test; WHO-QoL = World Health Organization Quality of Life scale Table 2Longitudinal changes outcome domainsBaseline3-year follow-up6-year follow-upMixed model analysisSymptoms*M (SD)RangeNM (SD)RangeNM (SD)RangeNFdfpη2PANSS Positive symptoms13.9 (6.6)2–4194611.8 (5.8)3–3970012.4 (6.6)2–3856233.52 < 0.010.09PANSS Negative symptoms15.0 (6.6)3–3993512.7 (5.5)4–4169912.5 (5.9)4–4156243.32 < 0.010.11PANSS Disorganization symptoms16.8 (6.3)10–4493814.4 (5.3)11–4669914.7 (5.9)11–4555944.02 < 0.010.12CDS Depressive symptomsXXX2.0 (2.9)0–166992.2 (3.3)0–193140.2710.600.00Social functioning*GAF total score54.4 (16.0)1–10093160.4 (16.1)25–10062958.9 (16.6)21–10054533.72 < 0.010.10SFS total scoreXXX126 (22.8)50–189681128.4 (22.2)58–1875464.910.030.02SFS Independence performanceXXX36.4 (3.5)11–3968336.5 (3.7)0–3954626.01 < 0.010.00Cognition*Attention: CPT mean RT correct responses432.1 (84.4)211.3–872.5899443.1 (83.8)272.6–847.8692443.5 (82.5)256–722.35478.02 < 0.010.02Executive functioning: response shifting task1.6 (5.6)0–398711.7 (6.5)0–396390.2 (0.5)0–1.91934.52 < 0.010.25Motor skills: WAIS-III block design9.6 (3.3)1–1896110.5 (3.3)3–1865911.0 (3.8)1–1956662.12 < 0.010.17Overall cognition: WAIS-III total scaled score37.1 (9.8)12–6495739.5 (9.6)15–6365740.8 (10.2)12–6756355.12 < 0.010.15Processing speed: WAIS-III digit symbol substitution7.3 (3.0)1–199618.0 (3.1)1–186638.2 (3.1)1–1956433.82 < 0.010.10Verbal memory: 15 WLT total items correct immediate recall22.8 (6.1)3–4195825.4 (6.4)8–4366324.0 (6.8)5–4356371.12 < 0.010.18Well-being*SWN-20 total score64.7 (7.2)25–9850963.8 (6.1)44–9143162.5 (6.4)37–9428511.72 < 0.010.06WHO-QOL score3.4 (1.0)1–59073.7 (0.9)1–56893.8 (0.9)1–554032.22 < 0.010.09*CDS = Calgary Depression Scale for Schizophrenia; CPT = Continuous Performance Test; GAF = Global Assessment of Functioning; PANSS = Positive and Negative Syndrome Scale; RT = Reaction time; SFS = Social Functioning Scale; SWN = 20 = The 20-item Subjective Well-being under Neuroleptics Scale; WAIS = Wechsler Adult Intelligence Scale; WLT = Word Learning Test; WHO-QoL = World Health Organization Quality of Life scale Longitudinal changes outcome domains *CDS = Calgary Depression Scale for Schizophrenia; CPT = Continuous Performance Test; GAF = Global Assessment of Functioning; PANSS = Positive and Negative Syndrome Scale; RT = Reaction time; SFS = Social Functioning Scale; SWN = 20 = The 20-item Subjective Well-being under Neuroleptics Scale; WAIS = Wechsler Adult Intelligence Scale; WLT = Word Learning Test; WHO-QoL = World Health Organization Quality of Life scale Moderators effects of changes in recovery domains Figure 1 provides an overview of moderating effects of changes between recovery domains. Specific statistics are discussed below and presented in Table 3 and both Supplementary Materials S2 and S3.Fig. 1Moderators of changes in outcome domains. Lines indicate that changes in one outcome domain moderates changes in another outcome domainTable 3Significant moderating effects change scores per recovery domain and duration of illness (DOI) subgroupOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yPositive symptomsCovariateFpFpFpFpFpChange GAF total score105.1 < 0.0136.3 < 0.0154.2 < 0.0111.9 < 0.015.20.03Change motor skills7.80.018.0 < 0.010.00.982.70.106.10.02Change overall cognition13.4 < 0.0110.3 < 0.013.90.0482.40.130.40.53Change processing speed12.5 < 0.0119.2 < 0.014.00.0470.00.961.60.21Change verbal memory5.10.0215.3 < 0.010.00.961.10.290.00.97Change quality of life13.3 < 0.015.20.021.50.227.0 < 0.011.40.24Negative symptomsOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange GAF total score44.9 < 0.0120.7 < 0.0120.0 < 0.016.20.013.50.07Change executive functioning30.4 < 0.0118.00.040.40.553.40.160.00.99Change motor skills14.8 < 0.016.60.013.70.060.50.487.80.01Change overall cognition19.0 < 0.0111.7 < 0.017.8 < 0.010.70.404.00.049Change processing speed20.4 < 0.0117.5 < 0.019.8 < 0.010.20.691.30.25Change verbal memory17.5 < 0.0112.6 < 0.010.60.426.00.026.50.01Change quality of life10.1 < 0.013.20.080.80.367.7 < 0.013.40.07Depressive symptomsOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange motor skills6.50.017.50.011.20.280.00.871.50.23Change overall cognition5.40.024.00.0461.00.330.80.370.40.55Change processing speed13.6 < 0.018.3 < 0.013.90.052.40.120.10.80Change verbal memory6.9 < 0.015.00.031.10.290.30.570.90.36Change quality of life37.0 < 0.0113.4 < 0.013.70.0619.5 < 0.014.00.049Disorganization symptomsOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange GAF total score66.4 < 0.0122.0 < 0.0123.5 < 0.0117.10.013.80.06Change executive functioning4.80.04157.2 < 0.010.70.441.20.360.00.99Change motor skills14.40.016.00.021.50.224.80.036.40.01Change overall cognition16.6 < 0.015.50.028.6 < 0.015.90.020.20.69Change processing speed19.8 < 0.0119.1 < 0.0111.3 < 0.010.10.720.60.43Change verbal memory13.0 < 0.0111.3 < 0.010.00.9410.2 < 0.011.10.29Change quality of life5.00.030.00.980.10.718.1 < 0.010.20.70GAF total score*OverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange positive symptoms4.40.0413.6 < 0.0128.8 < 0.016.90.012.30.13Change negative symptoms37.8 < 0.0117.2 < 0.0110.4 < 0.015.00.031.50.23Change depressive symptoms4.80.030.00.851.30.261.80.182.10.16Change disorganization symptoms42.2 < 0.0118.2 < 0.018.4 < 0.0112.4 < 0.012.60.11Change motor skills17.7 < 0.013.60.064.50.046.10.027.40.01Change overall cognition19.4 < 0.019.3 < 0.016.00.026.70.010.70.42Change processing speed29.9 < 0.0138.5 < 0.016.4 < 0.012.30.130.10.71Change verbal memory6.00.019.3 < 0.010.00.942.00.160.70.42Change quality of life103.4 < 0.017.9 < 0.011.00.3113.8 < 0.015.30.03SFS total score*OverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange positive symptoms4.20.049.8 < 0.015.60.022.40.138.3 < 0.01Change negative symptoms15.2 < 0.0112.0 < 0.012.50.112.00.160.10.78Change depressive symptoms**4.00.0470.10.820.90.351.20.274.50.04Change executive functioning7.40.012.70.235.30.051.10.370.00.99Change motor skills5.30.023.60.062.30.131.10.301.60.21Change overall cognition11.3 < 0.015.80.029.1 < 0.010.20.634.40.04Change processing speed27.2 < 0.0117.4 < 0.0123.6 < 0.010.10.782.50.12Change verbal memory10.1 < 0.012.00.162.40.126.00.022.80.10Change quality of life13.8 < 0.0110.0 < 0.011.30.254.20.041.80.18SFS Independence performance*OverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange negative symptoms7.5 < 0.014.50.040.50.471.00.313.10.08Change motor skills9.70.010.50.487.0 < 0.011.60.212.30.14Change overall cognition17.1 < 0.013.30.0713.5 < 0.010.30.604.10.047Change processing speed21.5 < 0.0110.0 < 0.0118.7 < 0.010.10.792.70.10Change quality of life16.3 < 0.0110.1 < 0.012.30.134.60.031.00.31AttentionOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange disorganization symptoms5.30.022.80.101.50.221.10.310.60.43Executive functioningOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange subjective well-being7.70.010.20.894.80.030.80.370.00.92Motor skillsThere was no significant influence of change scores for this outcomeOverall cognitionOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange negative symptoms9.0 < 0.010.20.697.0 < 0.017.10.010.00.92Change depressive symptoms5.00.031.40.241.10.292.40.130.40.53Change GAF total score6.20.010.40.516.40.010.70.421.50.22Change quality of life10.1 < 0.013.90.050.10.736.80.012.60.11Processing speedOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange negative symptoms8.7 < 0.012.30.135.50.021.40.240.40.54Verbal memoryOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange depressive symptoms5.60.024.00.0480.40.554.40.040.10.76Change GAF total score6.70.010.870.350.30.568.50.012.20.14Subjective well-beingOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange processing speed4.10.040.50.490.30.589.3 < 0.010.30.62Quality of lifeOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange depressive symptoms6.00.020.80.364.90.0210.10.000.10.77Change GAF total score34.1 < 0.0117.0 < 0.0111.1 < 0.017.80.010.40.55Change motor skills4.50.040.50.470.00.904.30.044.00.048Change overall cognition6.20.011.80.183.10.080.40.552.50.12Change processing speed13.8 < 0.0112.2 < 0.014.10.041.00.320.40.52Change verbal memory4.20.044.60.030.00.872.20.140.30.59*GAF = Global Assessment of Functioning; SFS = Social Functioning Scale**Moderator was not significant after Benjamini-Hochberg correction of multiple testing Moderators of changes in outcome domains. Lines indicate that changes in one outcome domain moderates changes in another outcome domain Table 3Significant moderating effects change scores per recovery domain and duration of illness (DOI) subgroupOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yPositive symptomsCovariateFpFpFpFpFpChange GAF total score105.1 < 0.0136.3 < 0.0154.2 < 0.0111.9 < 0.015.20.03Change motor skills7.80.018.0 < 0.010.00.982.70.106.10.02Change overall cognition13.4 < 0.0110.3 < 0.013.90.0482.40.130.40.53Change processing speed12.5 < 0.0119.2 < 0.014.00.0470.00.961.60.21Change verbal memory5.10.0215.3 < 0.010.00.961.10.290.00.97Change quality of life13.3 < 0.015.20.021.50.227.0 < 0.011.40.24Negative symptomsOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange GAF total score44.9 < 0.0120.7 < 0.0120.0 < 0.016.20.013.50.07Change executive functioning30.4 < 0.0118.00.040.40.553.40.160.00.99Change motor skills14.8 < 0.016.60.013.70.060.50.487.80.01Change overall cognition19.0 < 0.0111.7 < 0.017.8 < 0.010.70.404.00.049Change processing speed20.4 < 0.0117.5 < 0.019.8 < 0.010.20.691.30.25Change verbal memory17.5 < 0.0112.6 < 0.010.60.426.00.026.50.01Change quality of life10.1 < 0.013.20.080.80.367.7 < 0.013.40.07Depressive symptomsOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange motor skills6.50.017.50.011.20.280.00.871.50.23Change overall cognition5.40.024.00.0461.00.330.80.370.40.55Change processing speed13.6 < 0.018.3 < 0.013.90.052.40.120.10.80Change verbal memory6.9 < 0.015.00.031.10.290.30.570.90.36Change quality of life37.0 < 0.0113.4 < 0.013.70.0619.5 < 0.014.00.049Disorganization symptomsOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange GAF total score66.4 < 0.0122.0 < 0.0123.5 < 0.0117.10.013.80.06Change executive functioning4.80.04157.2 < 0.010.70.441.20.360.00.99Change motor skills14.40.016.00.021.50.224.80.036.40.01Change overall cognition16.6 < 0.015.50.028.6 < 0.015.90.020.20.69Change processing speed19.8 < 0.0119.1 < 0.0111.3 < 0.010.10.720.60.43Change verbal memory13.0 < 0.0111.3 < 0.010.00.9410.2 < 0.011.10.29Change quality of life5.00.030.00.980.10.718.1 < 0.010.20.70GAF total score*OverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange positive symptoms4.40.0413.6 < 0.0128.8 < 0.016.90.012.30.13Change negative symptoms37.8 < 0.0117.2 < 0.0110.4 < 0.015.00.031.50.23Change depressive symptoms4.80.030.00.851.30.261.80.182.10.16Change disorganization symptoms42.2 < 0.0118.2 < 0.018.4 < 0.0112.4 < 0.012.60.11Change motor skills17.7 < 0.013.60.064.50.046.10.027.40.01Change overall cognition19.4 < 0.019.3 < 0.016.00.026.70.010.70.42Change processing speed29.9 < 0.0138.5 < 0.016.4 < 0.012.30.130.10.71Change verbal memory6.00.019.3 < 0.010.00.942.00.160.70.42Change quality of life103.4 < 0.017.9 < 0.011.00.3113.8 < 0.015.30.03SFS total score*OverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange positive symptoms4.20.049.8 < 0.015.60.022.40.138.3 < 0.01Change negative symptoms15.2 < 0.0112.0 < 0.012.50.112.00.160.10.78Change depressive symptoms**4.00.0470.10.820.90.351.20.274.50.04Change executive functioning7.40.012.70.235.30.051.10.370.00.99Change motor skills5.30.023.60.062.30.131.10.301.60.21Change overall cognition11.3 < 0.015.80.029.1 < 0.010.20.634.40.04Change processing speed27.2 < 0.0117.4 < 0.0123.6 < 0.010.10.782.50.12Change verbal memory10.1 < 0.012.00.162.40.126.00.022.80.10Change quality of life13.8 < 0.0110.0 < 0.011.30.254.20.041.80.18SFS Independence performance*OverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange negative symptoms7.5 < 0.014.50.040.50.471.00.313.10.08Change motor skills9.70.010.50.487.0 < 0.011.60.212.30.14Change overall cognition17.1 < 0.013.30.0713.5 < 0.010.30.604.10.047Change processing speed21.5 < 0.0110.0 < 0.0118.7 < 0.010.10.792.70.10Change quality of life16.3 < 0.0110.1 < 0.012.30.134.60.031.00.31AttentionOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange disorganization symptoms5.30.022.80.101.50.221.10.310.60.43Executive functioningOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange subjective well-being7.70.010.20.894.80.030.80.370.00.92Motor skillsThere was no significant influence of change scores for this outcomeOverall cognitionOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange negative symptoms9.0 < 0.010.20.697.0 < 0.017.10.010.00.92Change depressive symptoms5.00.031.40.241.10.292.40.130.40.53Change GAF total score6.20.010.40.516.40.010.70.421.50.22Change quality of life10.1 < 0.013.90.050.10.736.80.012.60.11Processing speedOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange negative symptoms8.7 < 0.012.30.135.50.021.40.240.40.54Verbal memoryOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange depressive symptoms5.60.024.00.0480.40.554.40.040.10.76Change GAF total score6.70.010.870.350.30.568.50.012.20.14Subjective well-beingOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange processing speed4.10.040.50.490.30.589.3 < 0.010.30.62Quality of lifeOverallDOI < 2 yDOI 2–5 yDOI 5–10 yDOI > 10 yCovariateFpFpFpFpFpChange depressive symptoms6.00.020.80.364.90.0210.10.000.10.77Change GAF total score34.1 < 0.0117.0 < 0.0111.1 < 0.017.80.010.40.55Change motor skills4.50.040.50.470.00.904.30.044.00.048Change overall cognition6.20.011.80.183.10.080.40.552.50.12Change processing speed13.8 < 0.0112.2 < 0.014.10.041.00.320.40.52Change verbal memory4.20.044.60.030.00.872.20.140.30.59*GAF = Global Assessment of Functioning; SFS = Social Functioning Scale**Moderator was not significant after Benjamini-Hochberg correction of multiple testing Significant moderating effects change scores per recovery domain and duration of illness (DOI) subgroup *GAF = Global Assessment of Functioning; SFS = Social Functioning Scale **Moderator was not significant after Benjamini-Hochberg correction of multiple testing Moderating effects of changes in other recovery domains on changes in symptoms Changes in symptoms were moderated by changes in GAF total score, multiple domains of cognition and quality of life. Changes in GAF were a moderator for changes in all symptom domains, except for depressive symptoms. These moderating effects applied for patients with a DOI < 2 years, 2–5 years, 5–10 years, and, exclusively for changes in positive symptoms, a DOI > 10 years. Regarding cognition, changes in motor skills and processing speed moderated changes in all symptom domains, changes in executive functioning moderated changes in both negative and depressive symptoms, changes in overall cognition moderated changes in positive, negative and depressive symptoms and changes in verbal memory moderated changes in positive, negative and disorganization symptoms. All moderating effects applied for patients with a DOI < 2 years. Finally, changes in quality of life moderated changes in all symptom domains. All moderating effects applied for patients with a DOI between 5 and 10 years. Moderating effects of changes in other recovery domains on changes in social functioning Changes in social functioning were moderated by changes in all symptom domains, multiple domains of cognition and quality of life. Changes in negative symptoms were a moderator for changes in all domains of social functioning; this consistently applied for patients with a DOI < 2 years. Changes in positive symptoms were a moderator for changes in GAF and SFS total scores, specifically for patients with a DOI < 2 years and 2–5 years. Finally, changes in disorganization and depressive symptoms moderated changes in GAF scores, specifically for patients with a DOI shorter than 10 years. Regarding cognition, changes in motor skills, overall cognition and processing speed were a moderator for changes in all domains of social functioning. For overall cognition, those moderating effects applied to patients with a DOI 2–5 years and for processing speed to patients with a DOI < 2 years. Changes in verbal memory were a moderator for changes in SFS and GAF scores, and changes in executive functioning was a moderator for changes in SFS total scores. Finally, changes in quality of life moderated changes in all domains of social functioning. These moderating effects specifically applied for patients with a DOI < 2 years and 5–10 years. Moderating effects of changes in other recovery domains on changes in cognition Changes in cognition were moderated by changes in negative and depressive symptoms, GAF total scores and well-being. Changes in negative symptoms were a moderator for changes in overall cognition and processing speed, specifically for patients with a DOI between 2 and 5 years. Changes in depressive symptoms were a moderator for changes in overall cognition and verbal memory, and changes in disorganization were a moderator for changes in attention. Changes in GAF scores moderated changes in overall cognition and verbal memory but were not specifically indicated in any DOI subgroup. Finally, changes in subjective well-being were a moderator for changes in executive functioning, and changes in quality of life were a moderator for changes in overall cognition. Both moderating effects specifically applied for patients with a DOI between 2 and 5 years. Moderating effects of changes in other recovery domains on changes in quality of life or well-being Quality of life and well-being were moderated by changes in depressive symptoms, GAF total score and multiple domains of cognition. Changes in depressive symptoms moderated changes in quality of life, specifically for patients with a DOI 2–5 and 5–10 years. Furthermore, changes in GAF total scores were a moderator for changes in quality of life, specifically for patients with a DOI less than 10 years. Finally, we found that changes in processing speed were a moderator for changes in both subjective well-being and quality of life, and changes in motor skills, overall cognition and verbal memory were a moderator for changes in quality of life. None of these moderating effects were consistently indicated in patients with a specific DOI. Moderating effects of DOI, personal, clinical and social characteristics of patients Moderating effects of personal, clinical and social patient characteristics at baseline on changes in recovery domains are discussed below, and shown in Fig. 2, and Supplementary Materials S4.Fig. 2Moderating effects of patient characteristics at baseline*. * GAF = Global assessment of Functioning; SFS = social functioning Scale Moderating effects of patient characteristics at baseline*. * GAF = Global assessment of Functioning; SFS = social functioning Scale Duration of illness Patients with a shorter baseline DOI were associated with larger improvement in positive symptoms, GAF total score and executive functioning. Age Patients with a younger age at baseline were associated with larger improvement in negative symptoms, disorganization symptoms, GAF total score, executive functioning, and verbal memory. Baseline level of outcome A lower baseline level or higher baseline severity of the investigated outcome were associated with larger improvement in positive, negative, and disorganization symptoms, GAF total score, motor skills, processing speed, verbal memory, subjective well-being and quality of life. Duration of untreated psychosis (DUP). Patients with a shorter DUP at baseline were associated with larger improvement in positive symptoms, motor skills and quality of life. Education level Patients with a higher education level at baseline were associated with larger improvement in positive, negative and disorganization symptoms, all three outcome domains of social functioning, executive functioning, motor skills, overall cognition, processing speed, verbal memory and quality of life. Employment status Patients with competitive employment at baseline were associated with larger improvement in all domains of symptoms and social functioning, motor skills, overall cognition, processing speed, verbal memory and quality of life. Ethnicity Patients with a white ethnicity were associated with larger improvement in all domains of symptoms and social functioning, executive functioning, motor skills, overall cognition, processing speed, verbal memory and quality of life. Living situation Patients who were living independently (alone or with a partner) at baseline were associated with larger improvement in positive, negative and disorganization symptoms, all domains of social functioning, attention, motor skills, overall cognition and processing speed. Marital status Patients who were married or living together at baseline were associated with larger improvement in positive, negative and disorganization symptoms, all domains of social functioning, attention, processing speed, subjective well-being and quality of life. Sex Sex No moderating effect of sex was found. Schizophrenia diagnosis Schizophrenia diagnosis Patients with a diagnosis of schizophrenia at baseline were associated with larger improvement in symptoms, negative symptoms, disorganization symptoms, GAF total score, independence, motor skills, overall cognition, processing speed, verbal memory, subjective well-being and quality of life. Discussion A reflection of the results of our study is reported below. Before reflecting on these findings, it is important to note that moderating effects do not imply causal relationships but instead indicate to what extent the moderating variable strengthens the direction of changes in investigated recovery domains between baseline and follow-up assessment [37]. Moderating effects on changes between recovery domains We found a vast variety of moderating effects between changes in recovery domains. The moderating effects that stood out (i.e., present in more than halve of all combinations of outcomes) were: changes in cognition moderate changes in symptoms, changes in cognition moderate changes in social functioning, and changes in symptoms moderate changes in social functioning. In contrast, only a few moderating effects (i.e., in less than 20% of all combinations of outcomes) were found between: changes in social functioning moderate changes in cognition, changes in social functioning moderate changes in well-being, changes in symptoms moderate changes in well-being, and changes in well-being moderate changes in cognition. The robust influence of changes in symptoms and cognition on changes in social functioning is in line with a previous meta-analysis [38]. Other studies indicated that lower symptom severity, especially negative symptoms, and higher level of cognition were associated with better social functioning [39–42]. Findings in current study suggest that improvement in social functioning is strengthened by improvement in cognition and symptoms, mainly for patients with a short DOI. These findings imply that a focus on both cognitive impairment and symptom severity in clinical practice may have the largest impact on improvement of social functioning, especially early in the course of illness. Moderating effects of changes in other recovery domains on changes in quality of life and well-being differed between the two subdomains. Changes in quality of life was moderated by changes in multiple outcome domains of cognition, symptoms and social functioning. In contrast, subjective well-being was hardly moderated by or a moderator for changes in any recovery domain. This implies that quality of life and well-being are different constructs that go along with different processes. Quality of life captures more explicitly subjective experiences of behavioral components, functional status and environmental conditions, as well as concepts of well-being and life satisfaction [43, 44]. In contrast, well-being, as measured by the SWN-20, captures patients’ perception of their health status, antipsychotic treatment, and nonmedical aspects of their lives [45, 46]. This suggests that the finding of more associations of quality of life with changes in other recovery domains, compared to changes in well-being, might be explained by the behavioral, social and environmental components that were captured in quality of life, in contrast with well-being. Finally, results indicated that cognition was a consistent moderator for changes in symptoms, social functioning and quality of life. However, changes in cognition were hardly moderated by changes in these recovery domains themselves. This might suggest that improvement in cognition is a driving force for improvement in other recovery domains [40, 47]. On the other hand, changes in cognition were not easily influenced by changes in other recovery domains. Previous research indicated that improvement in cognition is modest at best, showing a persistent pattern of cognitive impairment without profound improvement over time [9, 48, 49]. Current findings emphasize the need to develop effective ways to improve cognitive functions in patients with SSD as a potential gateway towards better improvement in other recovery domains. Moderating effects of duration of illness on changes in recovery domains Patients with a short duration of illness (DOI) showed larger improvement in positive symptoms, GAF total score and executive functioning, indicating that a short DOI was associated with larger improvements in these outcome domains. The influence of DOI not only applied to changes in individual recovery domains, but also to the moderating effects between changes in recovery domains. We observed that 43 (66.2%) of the significant moderating effects between recovery domains applied to a subgroup of patients with a DOI less than two years, whereas only 14 (21.5%) of the significant moderating effects between recovery domains applied to patients with a baseline DOI more than 10 years. This suggests that patients with a shorter DOI show larger mutual improvement in multiple recovery domains than patients with a longer DOI which suggests a large potential for responsiveness to different specialized treatment programs [50–52]. We also found that a short duration of untreated psychosis (DUP) is associated with larger improvement in positive symptoms. This substantiated our hypotheses that early interventions in psychosis elevate improvement mainly in positive symptoms. Moderating effects of personal, clinical and social patient characteristics at baseline Several personal, clinical and social patient characteristics at baseline moderated changes in recovery domains. We found that patients with a high education level, patients who lived independently, patients who were competitively employed and patients who were married or living together all moderated improvement in symptoms, social functioning and cognition. These findings underline that the social contexts of patients’ daily lives and an adequate standard of living are crucial for improvement in multiple recovery domains. Previous research already emphasized that quality and quantity of social relationships, employment and independent living are important predictors for remission and recovery of patients with SSD [19, 50, 53, 54]. However, current results indicated relationships between social factors of patients with SSD and longitudinal improvement in multiple recovery domains. These insights are relatively novel and important insights to unlock the potential for recovery for patients with SSD. Furthermore, a schizophrenia diagnosis was a moderator for reduced improvement in all recovery domains. These moderating effects were also reported in previous meta-analyses focusing on changes in specific recovery domains [7–10]. This is mainly explained by the diagnostic criteria of schizophrenia that is given to patients with severe symptoms and impaired functioning, indicative of an unfavorable prognosis of recovery [23]. These results justify that more intensive support and treatment targeting various recovery domains is needed especially for patients diagnosed with schizophrenia. In line with this reflection, we also found that lower baseline level or higher baseline severity of the investigated outcome were associated with larger improvement in most outcomes. This suggests that patients with lower levels of functioning and higher severity of symptoms have more potential to improve in targeted recovery domains. We also found that patients with a white ethnic background showed larger improvement in all recovery domains compared with patients with non-white backgrounds. This is in line with previous studies that indicated unfavorable changes in multiple recovery domains for people with a non-white background [55–57]. Explanations given for this unfavorable prognosis of changes in recovery domains were a higher level of social disadvantage, lower rates of referrals to suitable treatment programs and lower treatment adherence, potentially due to social and cultural distances between clinicians and patients with an ethnic minority status [56–59]. This suggests that mental healthcare is not culturally sensitive enough. It is an important topic for further research on how to reduce these inequalities between patients with different ethnic backgrounds. Study limitations Considering study findings, several limitations should be addressed. The first limitation is that outcomes were assessed at only 3 assessment points, which is a relatively low number given the large follow-up length in this study. Multiple assessments, especially in the first 3 years of follow-up, might have captured the dynamic process of longitudinal changes in recovery domains better than done in current study design. Furthermore, the outcomes of depressive symptoms, SFS total score and independence were only addressed at 3-years and 6-years follow-up. Therefore, we considered the 3-years follow-up the baseline assessment for these outcomes. Consequently, at the starting point of observation patients were already involved in the study for three years. Generally, the largest improvement in most outcomes took place between baseline and 3-year follow-up. Furthermore, 26.7% of the study sample dropped out after 3-years follow-up, resulting in a smaller and likely higher functioning population being included in the analysis model for these outcomes compared to GROUP participants at baseline. This suggests that longitudinal changes in these outcomes may be underestimated compared to the recovery domains that were assessed at all assessment periods. Therefore, we should take more caution in the interpretation of both longitudinal changes and moderating effects of these outcomes. Furthermore, especially for executive functioning and well-being, sample sizes were considerably smaller at both baseline and follow-up than for the other outcomes. Consequently, these outcomes might show different patterns of change than the other recovery domains. Changes could be underestimated due to smaller sample sizes and, therefore, lower statistical power, or overestimated due to a selective patient sample who were assessed by these outcomes. Another limitation we should take into consideration is that outcomes that mutually moderate one another might raise issues of simultaneity (i.e., the influence of the independent variable is jointly determined by the dependent variable). As a consequence, moderating effects between variables with potential simultaneity might be overestimated and should be handled with caution. It is also important to note that we used GAF total score as an outcome for overall social functioning. However, GAF also embedded symptom severity and clinical features related to psychosis. These aspects could not be isolated from the total scores, and therefore, changes in GAF might be attributed to a large extent by clinical or symptomatic remission. Another limitation is that some important outcomes, such as specific domains of social functioning (e.g., vocational functioning, activities or social skills), perceived or internalized stigma, social cognition and personal recovery were not available in this cohort study. Previous research indicated that these outcomes do improve modestly over time in SSD [7, 10] and are associated with changes in different recovery domains [3, 18, 38, 60]. Furthermore, though we opted for a relatively homogeneous sample by only including patients with a classification of schizophrenia spectrum disorder, both diagnoses and comorbidity vary between patients and might cause heterogeneity in the outcomes. We considered it not recommendable to control for the influence of specific diagnoses or comorbidity due to the low number of patients with a diagnosis other than schizophrenia or a specific comorbidity. Multiple analyses with subgroups containing small sample sizes limit the representativeness of the outcomes and increases the chance of finding false-positive results [61]. Therefore, we decided not to control for the effects of diagnosis or comorbidity in the analysis of current study. Finally, there might be risk for attrition bias in our study sample. Our analysis indicated that patients that dropped out of the study had a significantly lower level of education, more often a non-white ethnic background and were less frequently employed than those who remained at follow-up. All these factors were moderators of changes in most recovery domains in our study. Previous research also indicated that a lower level of education is associated with a less favorable prognosis of change in most outcomes [62]. This indicated that the drop-out sample most probably had less favorable outcomes than those who remained in the study. Consequently, longitudinal changes that were indicated in this study might be overestimated. We tried to minimize these effects by executing a mixed-model analysis which takes all observations into consideration. Conclusions We found strong indications of moderating effects between changes in symptoms, social functioning, cognition and quality of life over six-year follow-up, especially in patients with a shorter DOI. In particular, cognition moderated a large variety of other recovery domains and might therefore be a driving force for improvement in other recovery domains. Changes in subjective well-being showed weaker associations with changes in other recovery domains, suggesting that well-being has a relatively separate prognosis of change compared to changes in other recovery domains. Furthermore, favorable social contexts, in the form of paid employment, higher education level, independent living and being in a stable relationship elevate improvement in most recovery domains. Reduced improvement in subjective well-being and depressive symptoms were mostly explained by that these domains were exclusively assessed at 3- and 6-year follow-up. It is therefore important to target these outcomes in follow-up research. We can conclude that cognition, symptoms, social functioning and quality of life often concurrently improved. Special attention to changes in cognition, focusing on social context and an adequate standard of living are key elements for improvement in most recovery domains. Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1