Identification of psychopathological course trajectoriesin schizophrenia

Markus Jäger a,n, Prisca Weiser a, Thomas Becker a, Karel Frasch a, Gerhard Längle b,c,Daniela Croissant d, Tilman Steinert e,f, Susanne Jaeger e,f, Reinhold Kilian a

a Ulm University, Department of Psychiatry and Psychotherapy II, BKH Günzburg, Ludwig-Heilmeyer-Str. 2, 89312 Günzburg, Germanyb Zentrum für Psychiatrie Südwürttemberg, Bad Schussenried, Germanyc Department of Psychiatry and Psychotherapy, University of Tübingen, Tübingen, Germanyd PP.rt Hospital for Psychiatry, Psychotherapy and Psychosomatics, Reutlingen, GermanyeUlm University, Department of Psychiatry and Psychotherapy I, Ravensburg, GermanyfZentrum für Psychiatrie Südwürttemberg, Weissenau, Ravensburg, Germany

a r t i c l e i n f o

Article history:Received 16 November 2012Received in revised form27 November 2013Accepted 30 November 2013Available online 7 December 2013

Keywords:SchizophreniaCourseTrajectoriesPANSS

a b s t r a c t

Course trajectory analyses have been performed primarily for treatment response in acute episodes ofschizophrenic disorders. As yet, corresponding data for the long-term course are lacking. Within amulticenter prospective observational study, 268 patients with schizophrenia were assessed at dischargefrom hospital and followed up after 6, 12, 18, and 24 months. A latent class growth analysis wasperformed on the scores from the Positive and Negative Syndrome Scale (PANSS). A two-class conditionallatent class model showed the best data fit (Entropy: 0.924). The model divided the sample into a groupwith amelioration in all PANSS subscales (60%) and a group with stable positive/negative anddeteriorating general psychopathology symptoms (40%). Global functioning (GAF score), gender, age,living situation and involuntary admission predicted course trajectory class membership. The model waspredictive of significant differences between the two groups in health care service costs and quality oflife. The results underline the heterogeneous course of the illness, which ranged from amelioration todeterioration over a 2-year period. Statistical models such as trajectory analysis could help to identifymore homogenous subtypes in schizophrenia.

& 2013 Elsevier Ireland Ltd. All rights reserved.

1. Introduction

Schizophrenia is a heterogeneous disorder (Picardi et al., 2012;Lang et al., 2013). The identification of subgroups with a differ-ential course and outcome is an important challenge for psychia-tric research (Andreasen, 2006). Such a subdivision could improvethe individual treatment of schizophrenia (Correll et al., 2011).As yet, several attempts have been made to establish a coursetypology of schizophrenic disorders (Bleuler, 1978; Leonhard,1979; Crow, 1980; Huber et al., 1980; Watt et al., 1983). However,these traditional course typologies are based merely upon clinicaljudgment.

More recently, empirical research has been performed withstandardized instruments and statistical models such as clusteranalysis or latent regression modeling. For example, Burti et al.(2009) identified five course patterns in schizophrenic disordersthat correlated with the DSM-IV course specifiers (APA, 1994)), and

Levine et al. (2011) described four groups. However, the informationon hospitalizations or psychotic episodes used in these analyseswas obtained from retrospective assessments or case register data.As yet, most trajectory analyses of prospective psychopathologicaldata from instruments such as the Positive and Negative SyndromeScale (PANSS) or the Scale for the Assessment of Negative Symp-toms (SANS) are limited to the acute treatment episode of schizo-phrenic disorders (Levine and Leucht, 2010; Levine and Rabinowitz,2010; Case et al., 2011, Stauffer et al., 2011; Schennach et al., 2012;Levine and Leucht, 2013). Only a few studies have examined thecourse of illness over a longer period. Levine et al. (2010) identifiedfive different trajectories of treatment response over a 2-yearperiod; all five groups were characterized by amelioration overthe longer course of illness. However, these data are limited topatients suffering from recent-onset psychoses and stem from arandomized controlled trial with rigid inclusion criteria. Further-more, Levine et al. (2012) reanalyzed data of the CATIE study andidentified three course trajectories in the entire sample. Theseresults need further validation.

Against this background, the present study examined thecourse of psychopathological symptoms over a 2-year period in

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/psychres

Psychiatry Research

0165-1781/$ - see front matter & 2013 Elsevier Ireland Ltd. All rights reserved.http://dx.doi.org/10.1016/j.psychres.2013.11.031

n Corresponding author. Tel.: þ49 82 21 96 2204; fax: þ49 82 21 96 2737.E-mail address: Markus.Jaeger@bkh-guenzburg.de (M. Jäger).

Psychiatry Research 215 (2014) 274–279

patients from a prospective observational study on antipsychoticdrug treatment of schizophrenic disorders in routine psychiatriccare. The aims were (i) to identify psychopathological coursetrajectories, (ii) to search for variables predicting course trajectoryclass membership, and (iii) to examine the association of classmembership with patient quality of life and costs of psychiatricservice use.

2. Methods

2.1. Subjects

The sample stems from a multicenter prospective observational study (ELAN:“Effects of Long-term Atypical Neuroleptic Treatment under Routine Conditions”study) that followed patients with schizophrenia or schizoaffective disorders for2 years after discharge from psychiatric inpatient treatment (Frasch et al., 2012;Kilian et al., 2012; Längle et al., 2012). The primary aim of the ELAN study was toexamine whether antipsychotic maintenance therapy with quetiapine is moreeffective than treatment with olanzapine or risperidone. Patients were recruited in2005 and 2006 and data collection was continued until December 2008. Baselineassessments were conducted in the week before hospital discharge and follow-upassessments were performed after 6, 12, 18, and 24 months.

Inclusion criteria were a diagnosis of schizophrenia or schizoaffective disorder,an age of 18 years or older, and a recommendation for at least 12 months'antipsychotic maintenance treatment. Patients qualified for study inclusion if theywere administered quetiapine, olanzapine, or risperidone in routine clinicaltreatment; each of these drugs could be administered either as monotherapy orin combination with other psychotropic drugs. Medication changes and disconti-nuation of treatment were permitted at any time. Exclusion criteria comprisedsubstance dependence, an organic psychiatric disorder, and developmental dis-ability. Eligible study participants were selected on the basis of inpatient admissionfiles at nine psychiatric hospitals in South Germany. In the recruitment process,study workers established compatibility with inclusion criteria by using a standar-dized screening procedure. Patients were only included in the study after providinginformed consent. The study protocol was approved by the local ethics committees.

Five hundred and thirty patients were eligible for inclusion and asked toparticipate in the study; 156 patients (29%) refused to participate. This group wassignificantly younger than those who participated and had fewer inpatientepisodes and a lower rate of schizoaffective disorder diagnoses. A total of 374patients agreed to participate and gave informed consent. Patients with a diagnosisof schizoaffective disorder (n¼106, 28.3%) were excluded from the analysesbecause the nosological status of schizoaffective disorder is quite ambiguous(Jäger et al., 2011). Therefore, 268 patients were included in the final analyses.The sample characteristics (baseline and follow-up assessments) are presented inTable 1. Comparisons of the baseline sample characteristics with each follow-upsample found no systematic differences and confirmed the missing at randomassumption.

2.2. Assessments

Diagnoses were derived from the International Diagnosis Checklists (IDCL) forICD-10 (Hiller et al., 1993). This instrument consists of 30 symptom lists in pocketform for the most common psychiatric diagnoses and allows an operationalizeddiagnostic decision with a high reliability.

Psychopathological characteristics were assessed with the PANSS (Kay, 1991).The PANSS, a widely used 30-item scale for assessing schizophrenic symptoms, iscomposed of three subscales: positive symptoms, negative symptoms, and generalpsychopathology (Kim et al., 2012).

Functional capacity was measured with the Global Assessment of Functioning(GAF) scale from DSM-IV (APA, 1994)); quality of life, with the Lancashire Quality ofLife Profile, LQoLP-EU (Gaite et al., 2000); and Cognitive Processing Speed (CPS),with the Digit Symbol Coding subtest from the Wechsler Adult Intelligence ScaleWAIS (Kaufman and Lichtenberger, 2002). Information about psychiatric serviceuse, including consumption of psychotropic medication, was obtained by means of

the Client Sociodemographic and Services Receipt Inventory (CSSRI) (Chisholmet al., 2000). This instrument was developed to measure the mental health serviceutilization and costs, and it covers the domains “sociodemographics”, “usual livingsituation”, “employment and income”, “service receipt” and “medication profile”(Chisholm et al., 2000). The CSSRI allows estimating the costs of psychiatric care(Roick et al. 2001).

All assessments were conducted by study staff trained in the use of the studyinstruments. Rater trainings for the PANSS were performed by clinical experts(Kilian et al., 2012).

2.3. Statistical analyses

To identify latent heterogeneity in the growth processes of the PANSS subscalesfor positive symptoms, negative symptoms, and general psychopathology, weconducted a Latent Class Growth Analysis (LCGA) as a special form of the generalmixture model (GMM) (Matschinger et al., 2006). The number of latent classes wasdetermined by comparing the Bayesian Information Criterion (BIC), the measure ofEntropy (E), and the Lo-Mendel-Rubin adjusted Likelihood Ratio Test (LMR-LRT) (Loet al., 2001; Matschinger et al., 2006; Jung and Wickrama, 2008). A lower BIC valueindicates a better model fit (Muthén, 2004; Muthén and Muthén, 2010). E variesbetween 0 and 1, and a value of E40.8 indicates a sufficient probability (Muthén,2004). The LMR-LRT provides a statistic for improving the fit of a model with kclasses in comparison to a model with k�1 classes (Lo et al., 2001).

The model was extended to a Conditional Latent Class Growth Analysis (CLCGA)to examine the effects of patients' baseline characteristics on class membership.The following baseline characteristics were included in the CLCGA: gender; age;having a partner; living in one's own apartment; lower education level; involuntaryadmission to psychiatric hospital; number of previous inpatient episodes; numberof suicide attempts; the General Assessment of Functioning (GAF) score; CognitiveProcessing Speed (CPS) test score; and treatment group (quetiapine, olanzapine,risperidone). To examine the association of class membership with subjectivequality of life, we computed linear random effects regression models with randomintercepts and fixed effects of class membership on the overall quality of life scoreand the health-related quality of life subscale from the LQoLP-EU as dependentvariables. LQoLP-EU scales were transformed to a range between 1 and 7, where1 indicates the lowest quality of life; and 7 the highest.

To examine the relationship of class membership to health care service costs,we computed linear random effects regression models with random intercepts andfixed effects of class membership on the direct and indirect 6-month costs ofmental illness and on the total costs of medical treatment. Nonparametric boot-strapping was applied to estimate the standard errors, to correct for deviation fromnormal distribution in the cost data.

CLCGA was conducted with MPLUS 6 (Muthén and Muthén, 2010), and randomeffects regression models were computed by using the xtlogit and xtreg proceduresof STATA 12.

3. Results

3.1. Identification of latent classes in psychopathological coursetrajectories

As indicated by the model fit parameters for the CLCGA(Table 2), the BIC improved with the number of classes from31421.7 for the one-class model to 19453.4 for the three-classmodel. The LMR-LRT for the two-class model revealed that thismodel has a better fit than the one-class model (LMR-LRT¼1186.7;p¼0.000). However, the three-class model failed to show a betterfit than the two-class model (LMR-LRT¼252.5; p¼0.353). Thecomparison of the entropy values suggested that class member-ship can be better predicted for the two-class model (E¼0.924)than for the three-class model (E¼0.897).

Table 1Sample characteristics.

Baseline 6 months 12 months 18 months 24 months

N (%) 268 (100) 229 (85) 228 (85) 216 (81) 216 (81)Age (years, S.D.) 38.8, 712.4 38.9, 712.4 38.7, 712.4 39.1, 712.2 38.8, 712.1Female gender (%) 41.8 41.5 40.4 43.1 42.1Global functioning (GAF-value, S.D.) 54.6, 712.7 58.2, 714.5 59.0, 715.0 60.0, 716.2 60.1, 716.4

GAF: Global Assessment of Functioning.

M. Jäger et al. / Psychiatry Research 215 (2014) 274–279 275

The slope of the growth processes in the PANSS subscales(Table 3) reveals change in symptoms over the five assessments.The intercept represents the estimated mean value at baseline. Forthe total population (one class model), all growth parametersindicated a decrease in symptoms on the PANSS positive subscale(slope¼�0.257; p¼0.000), the PANSS negative subscale (slope¼�0.314; p¼0.000), and the PANSS general psychopathologysubscale (slope¼�0.583; p¼0.000). Growth parameters for thetwo-class model revealed (i) a significant decrease in symptomson all PANSS subscales in members of class 1 (amelioration) (60%,n¼154), and (ii) a significant increase in General Psychopathology(GP) symptoms, while positive and negative symptoms remainedunchanged, in members of class 2 (deterioration GP) (40%, n¼103).

3.2. Prediction course trajectory class membership

The relationships of patients' baseline characteristics to theirpredicted class membership are presented in Table 4. As indicatedby the Odds Ratios (OR), the probability of membership in class 1(amelioration)was higher for female patients (OR¼3.056; p¼0.020)and for patients who lived in their own apartment (OR¼4.154;p¼0.002) and decreased with age (OR¼0.952; p¼0.026) and aninvoluntary admission (OR¼0.360; p¼0.023). With increasing GAFscore at baseline, patients had a higher probability of being assignedto class 1 (amelioration) (OR¼1.115; p¼0.000). The treatment group(quetiapine, olanzapine, risperidone) had no predictive value fortrajectory class membership.

3.3. Association of class membership with the patients' subjectivequality of life and psychiatric service use costs

Relations of latent class membership to the patients' health-related and overall subjective quality of life over the 2-year studyperiod are presented in Table 5. Regression coefficients revealedthat the health-related quality of life is 3.3 mean scale points lowerand the overall quality of life 9.6 mean scale points lower inpatients belonging to class 2 (deterioration GP) in comparison tothose belonging to class 1 (amelioration). As indicated by the R2,class membership explained 15% of the variance in health-relatedquality of life and 25% of the variance in overall quality of life.

Relations of latent class membership to the 6-month costsof psychiatric inpatient care and to total costs of psychiatric careare presented in Table 6. Regression coefficients indicated that6-month inpatient costs for members of class 2 (deterioration GP)are on average € 3475.5 higher than for members of class 1(amelioration), and 6-month total psychiatric treatment costs areon average € 8813.4 higher. As revealed by the R2, about 2% ofthe variance in inpatient costs and 11% of the variance in total

Table 2Model fit parameter for conditional latent class growth analyses (CLCGA).

Model BIC LMR-LRT, p Entropy n (%)

Class 1 Class 2 Class 3 Class 4

CLCGA1 Class 31421.7 – – 268 (100) – – –

2 Classes 19602.9 1186.7, 0.000 0.924 154 (60) 103 (40) – –

3 Classes 19453.4 252.5, 0.353 0.897 63 (25) 48 (19) 146 (57) –

BIC: Bayesian Information Criterion.LMR-LRT: Lo-Mendel-Rubin adjusted Likelihood Ratio Test.

Table 3Growth curve parameter for the conditional latent class models.

Model PANSS-POS PANSS-NEG PANSS-GP

Intercept, p Slope, p Intercept, p Slope, p Intercept, p Slope, p

1 Class (total sample) 11.742, 0.000 �0.257, 0.000 14.322, 0.000 �0.314, 0.001 28.592, 0.000 �0.583, 0.0002 Classes

Class 1 (60%) (amelioration) 9.839, 0.000 �0.354, 0.000 12.199, 0.000 �0.551, 0.000 25.528, 0.000 �1.106, 0.000Class 2 (40%) (deterioration GP) 14.239, 0.000 0.026, 0.872 16.729, 0.000 0.355, 0.058 32.428, 0.000 0.604, 0.043

The intercept represents the estimated mean value at baseline. The slope represents the estimated change of the scale value.PANSS-POS: positive symptom subscore, PANSS-NEG: negative symptom subscore, PANSS-GP: general psychopathology subscore.

Table 4Relationship of baseline characteristics to predicted class membership class 1(amelioration) vs. class 2 (deterioration GP).

OR p

Female 3.056 0.020Age 0.952 0.026Having a partner 0.489 0.136Living in one's own apartment 4.154 0.002Lower education level 1.029 0.945Involuntary admission 0.360 0.023Number of previous inpatient episodes 0.988 0.654Suicide attempts 1.038 0.953GAF value 1.115 0.000CPS 1.154 0.084Treatment with olanzapinea 1.271 0.638Treatment with risperidonea 0.742 0.518

GAF: Global Assessment of Functioning; CPS: Cognitive Processing Speed.a Reference class: quetiapine.

Table 5Relationship of latent class membership to the health-related and overall subjectivequality of life.

Model/variable Health-related qualityof life

Overall qualityof life

b p R2 b p R2

Two-class modelClass 1 (amelioration) Reference categoryClass 2 (deterioration GP) �3.3 0.000 0.15 �9.6 0.000 0.25

Constant 14.5 0.000 52.4 0.000

M. Jäger et al. / Psychiatry Research 215 (2014) 274–279276

psychiatric treatment costs could be explained by latent classmembership.

4. Discussion

4.1. Two class model of psychopathological course trajectories

The results of our study indicate that a two-class conditionallatent class model of trajectories based on PANSS subscales showsthe best data fit. Analyses of the relations of class membershipwith quality of life and costs of psychiatric treatment revealed thatthe model is predictive of significant differences between groupsin terms of quality of life and costs.

The results are in contrast to previous trajectory analyses,which identified four (Case et al., 2011) or five (Levine et al.,2010; Levine and Leucht, 2010; Levine and Rabinowitz, 2010;Stauffer et al., 2011; Schennach et al., 2012) latent classes. Whilethese studies concurred in finding an amelioration of psycho-pathological symptoms in all course trajectories, we also identifieda latent class with stable positive/negative symptoms and dete-riorating general psychopathology symptoms. However, our studyexamined patients in a more stable period of their illness, afterdischarge from hospital, whereas the studies mentioned abovetried to identify course trajectories for treatment response in acuteepisodes of schizophrenic illness. In the acute stages, one cantypically observe a strong decrease in psychopathological symp-toms (Jäger et al., 2010). Furthermore, in contrast to many previousstudies (Levine et al., 2010; Levine and Rabinowitz, 2010; Caseet al., 2011; Schennach et al., 2012) patients suffering fromschizoaffective disorder were excluded from the trajectory ana-lyses in the present study. Lastly, we conducted growth analysesfor the three PANSS subscores (positive symptoms, negativesymptoms, general psychopathology) whereas trajectory analysesin other studies were limited to the PANSS total score. This mayexplain the lower number of latent classes identified by ouranalyses. Apart from that, diverging results could be explainedby different selection criteria in clinical studies. On the other hand,our results are quite comparable to those of Levine et al. (2012)who reanalyzed the data of the CATIE study and identified two orthree course trajectories with a range from amelioration todeterioration.

The two class model does not correspond directly to one of thetraditional clinical course typologies (Bleuler, 1978; Leonhard,1979; Crow, 1980; Huber et al., 1980; Watt et al., 1983). Never-theless, the course trajectory with amelioration of positive, nega-tive, and general psychopathology symptoms could correspondapproximately to concepts like “cycloid psychoses” (Leonhard,1979), “type I schizophrenia” (Crow, 1980), or episodes “with noor minimal impairment” (Watt et al., 1983). In contrast, thetrajectory with increasing general psychopathology symptomscould fit in roughly with concepts such as “systematic schizo-phrenia” (Leonhard, 1979) or “type II schizophrenia” (Crow, 1980)or courses “with increasing impairment” (Watt et al., 1983).

4.2. Prediction of course in schizophrenia

The present results revealed that gender, age, involuntaryadmission and living situation are predictors for course trajectoryclass membership. However, the statistically most significantpredictor for membership in the group with decreasing symptoms,i.e. a relatively favorable outcome, was a high GAF score at baseline.This means that patients with good functioning at discharge fromhospital appear to show a further improvement of symptoms overthe longer term. The predictive value of GAF score has been shownfor treatment response in acute episodes of schizophrenic disorders(Jäger et al., 2009; Schennach et al., 2012). Furthermore, membersof the two trajectory classes seemed to differ with respect to allthree PANSS subscores. In fact, previous studies have identifiedPANSS subscores at discharge as predictor for the further course(Spellmann et al. 2012). However, for statistical reasons it was notpossible to include these values in the conditional latent classmodel as predictor variables (Muthén and Muthén, 2010).

Our results also showed that the medication group (quetiapine,olanzapine, risperdione) is not predictive of course trajectory classmembership. These findings are in contrast to those of Levine et al.(2012) who found that olanzapine treated patients were morelikely than other treatment groups (perphenazine, quetiapine,risperidone, ziprasidone) to belong to the trajectory of “respon-ders”. The diverging results could be explained by differences insampling process and study design (e.g. observational study vs. RCT).

4.3. Impact of statistical models for psychopathological research

Hitherto, trajectory analyses have not been able to confirmcomplex psychopathological course typologies, e.g. those of Bleuler(1978), Huber et al. (1980), or Leonhard (1979). Therefore, one couldargue for an abandonment of these traditional psychopathologicalconcepts, because they are not confirmed by empirical research.However, all statistical models based on psychopathological datashould be regarded with caution. Feinstein, for example, criticizedthe use of statistical approaches such as cluster analysis to identifyvalid types of psychiatric disorders, because “they all depend onmathematical principles that have no direct relationship to clinicalreality” (Feinstein, 1977). In fact, it is difficult to describe complexphenomenological concepts with statistical models. Authors likeParnas and Bovet (1995) have criticized a pure “psychometric”approach in psychopathological research, and Andreasen (2007)has even referred to a “death of phenomenology.” Future psycho-pathological research programs should link quantitative statisticalmodels with qualitative-phenomenological approaches.

4.4. Limitations

We have presented a post hoc analysis of an observationalstudy on antipsychotic drug treatment (ELAN Study). Therefore,the study was not designed to consider trajectory analyses.However, all previous trajectory analyses on the course of schizo-phrenic disorders also are limited to post hoc analyses (Levineet al., 2010; Levine and Leucht, 2010; Levine and Rabinowitz, 2010;Case et al., 2011, Stauffer et al., 2011; Schennach et al., 2012).Furthermore, one has to consider that only five assessments(baseline and follow-ups after 6, 12, 18, and 24 months) wereperformed over the study period. Beyond that, the restriction to aperiod of only 2 years could hamper the detection of morecomplex patterns in the longitudinal course of schizophreniawhich have been described in traditional psychopathologicalliterature (Bleuler, 1978; Leonhard, 1979; Huber et al., 1980).

Furthermore, the influence of treatment on psychopathologicalcourse trajectories was not adequately controlled, becausemedication changes and discontinuation of treatment were

Table 6Relationship of latent class membership to the 6-month costs of psychiatric care.

Model/variable Costs of psychiatricinpatient care

Total costs ofpsychiatric care

b p R2 b p R2

Two-class modelClass 1 (amelioration) Reference categoryClass 2 (deterioration GP) 3475.5 0.000 0.02 8813.4 0.000 0.11

Constant 5777.8 0.000 8976.6 0.000

M. Jäger et al. / Psychiatry Research 215 (2014) 274–279 277

permitted at any time of the ELAN study (Frasch et al., 2012; Kilianet al., 2012; Längle et al., 2012). Variables such as dosage andcombination of antipsychotic drugs, treatment adherence, orpsychotherapy and psychosocial interventions strongly influencethe clinical course of the illness. The present analysis, however,only included the antipsychotic treatment group at baseline (quetia-pine, olanzapine, risperdione). Lastly, the non-participation rate (29%of eligible patients) could affect the generalization of the presentresults.

4.5. Conclusions

Results of the current trajectory analysis underline the hetero-geneous clinical course of schizophrenic disorders over a 2-yearperiod. We identified both a course type with amelioration and acourse type with stability and deterioration of psychopathologicalsymptoms. Statistical models such as latent class growth analysiscould help to identify more homogenous subtypes in the course ofschizophrenia. Relationships of the course trajectories identifiedin the current study with traditional psychopathological conceptsrequire further evaluation. Future psychopathological researchshould combine statistical analyses with a phenomenologicalapproach.

Acknowledgment

The ELAN study was financed by an educational Grant to theUniversities of Tübingen and Ulm from AstraZeneca, Wedel,Germany (Project no. 229/2004V). AstraZeneca had no role inthe development of the research questions; the design of thestudy; the collection, analyses and interpretation of data; or thewriting of the manuscript. AstraZeneca had the right to commenton the final draft of the paper before submission to the journal.

We wish to thank all patients who participated in the study. Inaddition, we thank Jochen Bügler, Gerhard W. Eschweiler, ThomasMesser, Max Schmauß, Carmen Pfiffner, Wiltrud Schepp, and FrankSchwärzler for supporting patient recruitment. We thank HeikeWiesner, Filiz Özfirat, Tanja Gieselmann, and Simone Triem fordata collection and data entry, and Ildiko Baumgartner for herwork in processing the data. Also, we thank Jacquie Klesing, ELS,for editing assistance with the manuscript.

References

American Psychiatric Association, 1994. Diagnostic and Statistical Manual of MentalDisorders (DSM-IV), 4th ed. American Psychiatric Association, Washington DC.

Andreasen, N.C., 2006. Standardized remission criteria in schizophrenia. ActaPsychiatrica Scandinavica 113, 81.

Andreasen, N.C., 2007. DSM and the death of phenomenology in America: anexample of unintended consequences. Schizophrenia Bulletin 33, 108–112.

Bleuler, M., 1978. The Schizophrenic Disorders: Long-term Patients and FamilyStudies. Yale University Press, New Haven. (S.M. Clemens, Trans.).

Burti, L., Mazzi, M., Koeter, M., Schene, A., Helm, H., Puschner, B., Bindman, J., Leese,M., Thornicroft, G., Tansella, M., 2009. An empirical method to identify patternsin the course of psychotic episodes of people with schizophrenia. InternationalJournal of Methods in Psychiatric Research 18, 265–278.

Case, M., Stauffer, V.L., Ascher-Svanum, H., Conley, R., Kapur, S., Kane, J.M., Kollack-Walker, S., Jacob, J., Kinon, B.J., 2011. The heterogeneity of antipsychoticresponse in the treatment of schizophrenia. Psychological Medicine 41,1291–1300.

Chisholm, D., Knapp, M., Knudsen, H.C., Amaddeo, F., Gaite, L., van Wijngaarden, B.,2000. Client sociodemographic and service receipt inventory—European ver-sion. British Journal of Psychiatry 177 (Suppl. 39), S28–S33.

Correll, C.U., Canas, F., Lamo, I., Levy, P., Montes, J.M., Papageorgiou, G., Rossi, A.,Surlason, R., Zink, M., 2011. Individualizing antipsychotic treatment selection inschizophrenia: characteristics of empirically derived patients subgroups. Eur-opean Psychiatry Suppl 1, 3–16.

Crow, T., 1980. Molecular pathology of schizophrenia: more than one diseaseprocess? British Medical Journal 280, 66–68.

Feinstein, A.R., 1977. A critical overview of diagnosis in psychiatry. In: Rakoff, V.M.,Gorman, H., Kedward, H.B., Preston, A.J., Stancer, H.C. (Eds.), PsychiatricDiagnosis. Brunner, New York, pp. 189–206.

Frasch, K., Weiser, P., Becker, T., Längle, G., Steinert, T., Niederreiner, C., Pfiffner, C.,Jaeger, S., Bayer, W., Eschweiler, G.W., Kilian, R., 2012. Psychotropic drugtreatment, clinical characteristics and cognitive processing speed in patientswith schizophrenia: results from the ELAN study. Pharmacopsychiatry 45,138–145.

Gaite, L., Vázquez-Barquero, J.L., Arraiga Arrizabalaga, A., Schene, A.H., Welcher, B.,Thornicroft, G., Ruggeri, M., Vázquez-Bourgon, E., Pérez Retuerto, M., Leese, M.,2000. Quality of life in schizophrenia: development, reliability and internalconsistency of the Lancashire Quality of Life Profile—European Version. EPSI-LON Study 8. European psychiatric services: inputs linked to outcome domainsand needs. British Journal of Psychiatry 39 (Suppl.), S49–S54.

Hiller, W., Zaudig, M., Mombour, W., Bronisch, T., 1993. Routine psychiatricexaminations guided by ICD-10 diagnostic checklists (International DiagnosticChecklists). European Archives of Psychiatry and Clinical Neuroscience 242,218–223.

Huber, G., Gross, G., Schüttler, R., Linz, M., 1980. Longitudinal studies of schizo-phrenic patients. Schizophrenia Bulletin 6, 592–605.

Jäger, M., Becker, T., Haack, S., Frasch, K., 2011. Schizoaffective disorder—an ongoingchallenge for psychiatric nosology. European Psychiatry 26, 159–165.

Jäger, M., Riedel, M., Schmauß, M., Laux, G., Pfeiffer, H., Naber, D., Schmidt, L.G.,Gaebel, W., Klosterkötter, J., Heuser, I., Kühn, K.U., Lemke, M.R., Rüther, E.,Buchkremer, G., Gastpar, M., Bottlender, R., Strauss, A., Möller, H.J., 2009.Prediction of symptom-remission in schizophrenia during inpatient treatment.World Journal of Biological Psychiatry 10, 426–434.

Jäger, M., Riedel, M., Obermeier, M., Schennach-Wolff, R., Seemüller, F., Messer, T.,Laux, G., Pfeiffer, H., Naber, D., Schmidt, L.G., Gaebel, W., Klosterkötter, J.,Heuser, I., Kühn, K.U., Lemke, M.R., Rüther, E., Klingberg, S., Gastpar, M.,Bottlender, R., Möller, H.J., 2010. Time course of antipsychotic treatmentresponse in schizophrenia: results from a naturalistic study in 280 patients.Schizophrenia Research 118, 183–188.

Jung, T., Wickrama, K.A.S., 2008. An introduction to latent class growth analysis andgrowth mixture modeling. Social and Personality Psychology Compass 2,302–317.

Kaufman, A, Lichtenberger, E., 2002. Assessing Adolescent and Adult Intelligence.John Wiley & Sons, Hoboken.

Kay, S.R., 1991. Positive and negative syndromes in schizophrenia: assessment andresearch. Clinical and experimental psychiatry monography no. 5 Brunner/Mazel, New York.

Kilian, R., Steinert, T., Schepp, W., Weiser, P., Jaeger, S., Pfiffner, C., Frasch, K.,Eschweiler, G.W., Messer, T., Croissant, D., Becker, T., Längle, G., 2012. Effec-tiveness of antipsychotic maintenance therapy with quetiapine in comparisonwith risperidone and olanzapine in routine schizophrenia treatment: results ofa prospective observational trial. European Archives of Psychiatry and ClinicalNeuroscience 262, 589–598.

Kim, J.H., Kim, S.Y., Lee, J., Oh, K.J., Kim, Y.B., Cho, Z.H., 2012. Evaluation of the factorstructure of symptoms in patients with schizophrenia. Psychiatry Research 197,285–289.

Längle, G., Steinert, T., Weiser, P., Schepp, W., Jaeger, S., Pfiffner, C., Frasch, K.,Eschweiler, G.W., Messer, T., Croissant, D., Becker, T., Kilian, R., 2012. Effects ofpolypharmacy on outcome in patients with schizophrenia in routine psychia-tric treatment. Acta Psychiatrica Scandinavica 125, 372–381.

Lang, F.U., Kösters, M., Lang, S., Becker, T., Jäger, M., 2013. Psychopathological long-term outcome of schizophrenia—a review. Acta Psychiatrica Scandinavica 127,173–182.

Leonhard, K., 1979. Classification of Endogenous Psychoses and Their DifferentiatedEtiology. Irvington Publishers, Inc., New York. (Charles H. Cahn, Trans.).

Levine, S.Z., Leucht, S., 2010. Elaboration on the early-onset hypothesis of anti-psychotic drug action: treatment response trajectories. Biological Psychiatry 68,86–92.

Levine, S.Z., Leucht, S., 2013. Psychometric analysis in support of shortening theScale for the Assessment of Negative Symptoms. European Neuropsychophar-macology 23, 1051–1056.

Levine, S.Z., Lurie, I., Kohn, R., Levav, I., 2011. Trajectories of the course ofschizophrenia: from progressive deterioration to amelioration over threedecades. Schizophrenia Research 126, 184–191.

Levine, S.Z., Rabinowitz, J., 2010. Trajectories and antecedents of treatmentresponse over time in early-episode psychosis. Schizophrenia Bulletin 36,624–632.

Levine, S.Z., Rabinowitz, J., Case, M., Ascher-Svanum, H., 2010. Treatment responsetrajectories and their antecedents in recent-onset psychosis: a 2-year prospec-tive study. Journal of Clinical Psychopharmacology 30, 446–449.

Levine, S.Z., Rabinowitz, J., Faries, D., Lawson, A.H., Ascher-Svanum, H., 2012.Treatment response trajectories and antipsychotic medications: examinationof up to 18 months of treatment in the CATIE chronic schizophrenia trial.Schizophrenia Research 137, 141–146.

Lo, Y., Mendell, N.R., Rubin, D.B., 2001. Testing the number of components in anormal mixture. Biometrika 88, 767–778.

Matschinger, H., Kilian, R., Angermeyer, M.C., 2006. Detecting specific patterns ofchange for two outcome scores in a mental health study by means of linearmixture models. Methodology 2, 113–123.

Muthén, B.O., 2004. Mplus Technical Appendices. Muthén & Muthén, Los Angeles.Muthén, L.K., Muthén, B.O. Mplus 6, 2010 Unser's Guide. Statistical analysis with

latent variablesSixth ed.. Muthén & Muthén, Los Angeles.

M. Jäger et al. / Psychiatry Research 215 (2014) 274–279278

Parnas, J., Bovet, P., 1995. Research in psychopathology: epistemiologic issues.Comprehensive Psychiatry 36, 167–181.

Picardi, A., Viroli, C., Tarsitani, L., Miglio, R., de Girolamo, G., Dell'acqua, G., Biondi,M., 2012. Heterogeneity and symptom structure of schizophrenia. PsychiatryResearch 198, 386–394.

Roick, C., Kilian, R., Matschinger, H., Bernet, S., Mory, C., Angermeyer, M.C., 2001.German adaptation of the Client Sociodemographic and Service ReceiptInventory. Psychiatrische Praxis 28, S84–S90. (Suppl. 2).

Schennach, R., Meyer, S., Seemüller, F., Jäger, M., Schmauss, M., Laux, G., Pfeiffer, H.,Naber, D., Schmidt, L.G., Gaebel, W., Klosterkötter, J., Heuser, I., Maier, W.,Lemke, M.R., Rüther, E., Klingberg, S., Gastpar, M., Musil, R., Möller, H.J., Riedel,M., 2012. Response trajectories in "real-world" naturalistically treated schizo-phrenia patients. Schizophrenia Research 139, 218–224.

Spellmann, I., Riedel, M., Schennach, R., Seemüller, F., Obermeier, M., Musil, R.,Jäger, M., Schmauß, M., Laux, G., Pfeiffer, H., Naber, D., Schmidt, L.G., Gaebel, W.,Klosterkötter, J., Heuser, I., Maier, W., Lemke, M.R., Rüther, E., Klingberg, S.,Gastpar, M., Möller, H.J., 2012. One-year functional outcomes of naturalisticallytreated patients with schizophrenia. Psychiatry Research 198, 378–385.

Stauffer, V., Case, M., Kollack-Walker, S., Ascher-Svanum, H., Ball, T., Kapur, S.,Kinon, B., 2011. Trajectories of response to treatment with atypical antipsycho-tic medication in patients with schizophrenia pooled from 6 double-blind,randomized clinical trials. Schizophrenia Research 130, 11–19.

Watt, D.C., Katz, K., Shepherd, M., 1983. The natural history of schizophrenia. A 5-yearprospective follow-up of a representative sample of schizophrenics by means of astandardized clinical and social assessment. Psychological Medicine 13, 663–667.

M. Jäger et al. / Psychiatry Research 215 (2014) 274–279 279