Proposal of an innovative approach to clarify the mechanism through which poor cognitive performance in adolescence impacts risk for schizophrenia (SZ).
To determine whether the developmental processes that predispose to SZ are better reflected by the observed cognitive performance in adolescence or the deviation of that performance from the individual’s familial cognitive aptitude (FCA).
Design, Setting, and Participants
A prospective cohort design. Risk for SZ and bipolar illness (BPI) are predicted by school achievement (SA) at age 16 years and IQ at ages 18 to 20 years and the deviation of that performance from an individual’s FCA. Familial cognitive aptitude is calculated from the SA, IQ, and educational attainment in biological relatives.
Main Outcomes and Measures
Diagnoses of SZ or BPI in the Swedish Hospital Discharge Register and the Swedish Outpatient Register.
Participants were 996 886 individuals with recorded SA and 106 187 individuals with recorded IQ born in Sweden between January 1, 1972, and December 31, 1990, with sufficient numbers of biological relatives to calculate their FCA. The first cohort is 48.7% female, and the second is all male. Risk for SZ was strongly predicted by the deviation of SA from the FCA (hazard ratio [HR], 0.56; 95% CI, 0.49-0.63) but not with the observed SA (HR, 1.01; 95% CI, 0.91-1.13). Similar results were obtained for IQ (HR, 0.53; 95% CI, 0.37-0.77 for the deviation from the FCA and HR, 1.07; 95% CI, 0.78-1.46 for the observed IQ). After matching SZ and control probands on cognitive performance, the siblings of the SZ probands had SA and IQs that did not differ from population means and were significantly higher in cognitive performance than for the siblings of control probands. Correlations in SA and IQs between the pre-SZ probands and their siblings were significantly lower than those observed between the matched control probands and their siblings. Risk for BPI was more weakly predicted by deviations from the FCA. No differences were found in the SA and IQs of siblings of BPI vs matched control probands.
Conclusions and Relevance
The neurodevelopmental processes that predispose to SZ are not well reflected by cognitive performance per se but rather are much better indexed by the deviation in cognitive ability from that expected from an individual’s FCA. The lowered cognitive performance in the pre-SZ probands appears to arise from qualitative developmental impairments rather than an unlucky combination of the genetic and environmental risk factors widely distributed within their family.
The neurodevelopmental model of schizophrenia (SZ) hypothesizes that deficits in cognition arise years before the emergence of clinical symptoms.1-3 Two major prospective designs have tested this hypothesis. The most common has been to examine the association between cognitive performance and risk for subsequent SZ. Such studies have consistently shown that poor cognitive performance predicts SZ risk.4-6 Rarer have been longitudinal designs with at least 2 measures of cognitive performance that can evaluate the degree to which the cognitive deficits predictive of SZ are static or progressive.7,8
We propose a third design. Relying on the strong familial resemblance for cognitive performance,9-11 we compare in a 16-year-old Swedish sample the risk of SZ from their observed school achievement (SA) and the difference between their observed SA and their familial cognitive aptitude (FCA) assessed from direct measures in their relatives. We extend these analyses to the prediction of SZ from a different measure of cognitive performance (IQ assessed at ages 18-20 years). We hope to clarify whether the neurodevelopmental processes that predispose to SZ are best indexed by poor cognitive performance or whether these pathogenic process are more accurately measured by the degree to which individuals fall short of their FCA.
We also want to know whether the lowered cognitive performance of pre-SZ individuals results from an unfortunate combination of the same genetic and environmental factors affecting cognitive aptitude in their siblings or from the developmental processes distinct from those operating in their close relatives. Therefore, we match pre-SZ individuals to controls on their level of cognitive performance and compare the SA and IQ in the 2 sets of siblings. If individuals destined for SZ just got an unlucky combination of the genes and environment distributed among their siblings, then the cognitive performance of the 2 groups of siblings should be similar. Furthermore, the correlation in cognitive performance between probands who develop SZ and their siblings should be the same as that seen between the matched control probands and their siblings. However, if the developmental factors impacting on the SZ proband were outside the range of those experienced by their relatives, then the cognitive performance of their siblings should be higher than that of the siblings of the matched controls, and the proband sibling correlations for SA and IQ should be lower in the SZ than the control sibships.
We repeated all these analyses for bipolar illness (BPI) as an important comparison group. Prior research is mixed as to the relationship of BPI with premorbid cognitive functioning.7,8,12-15
Box Section Ref ID
Question Are the developmental processes that predispose to schizophrenia better indexed by observed cognitive performance in adolescence or the deviation of that performance from the individual’s familial cognitive aptitude?
Findings The risk for schizophrenia was strongly predicted by the deviation of school achievement from the familial cognitive aptitude but not at all by observed school achievement. Similar results were obtained for IQ.
Meaning The neurodevelopment processes that predispose to schizophrenia are not well reflected by cognitive performance per se but rather are much better indexed by the deviation in ability from that expected given the cognitive potential of the family.
We analyzed information on individuals from Swedish population-based registers with national coverage. These registers were linked using each person’s unique identification number replaced by a serial number to preserve confidentiality. We secured ethical approval for this study from the Regional Ethical Review Board of Lund University.
The National School Registry contains educational achievement (ie, grade point average) for all students in grade 9 (usually age 16 years) from 1988 to 2012. Students had an incentive to perform well because those scoring higher were more likely to gain admission to the desirable secondary schools. For each year and sex, we standardized grade score and called this variable SA. From 1988 to 1997, the score was expressed on a scale between 1 and 5 (mean, 3.2), and students were assessed by a peer-referencing system.16 Under this system, grades had minimal inflation over time and were normally distributed. From 1998 onward, scores were expressed on scale between 10 and 320 using a criterion-referenced system and were not standardized across schools.
The Military Conscription Register (1990-2008) includes, for almost all men in Sweden 18 to 20 years old, a full medical assessment, including IQ measured by 4 subtests representing logical, spatial, verbal, and technical abilities. During these years, the examination was legally required. Only those with serious medical conditions or disabilities were excused (approximately 4.2%). The global IQ, a summation of the 4 subtests, was standardized to give a gaussian score between 1 and 9. For each year, we standardized the score.
From the Swedish Hospital Discharge Register and the Swedish Outpatient Register, we identified all individuals registered for SZ and BPI. Schizophrenia was defined by a lifetime diagnosis of the following International Classification of Diseases (ICD) codes: ICD-9 codes 295.1, 295.2, 295.3, 295.6, and 295.9 and ICD-10 codes F20.0, F20.1, F20.2, F20.3, F20.5, and F20.9. We excluded cases of acute, latent, and schizoaffective SZ.
Bipolar illness was defined by a lifetime diagnosis of the following ICD codes: ICD-9 codes 296.0, 296.1, 296.4, 296.5, 296.6, 296.7, and 296.8 and ICD-10 codes F31 and F30. We applied a lifetime hierarchical diagnostic approach so that individuals diagnosed as having BPI never received a SZ diagnosis.
We began with all individuals born in Sweden between January 1972 and December 1990 who had not died or migrated before age 16 years (n = 1 861 934). From the Multigenerational Register, we also selected all siblings, cousins, and parents for these individuals. The first analysis included individuals with recorded SA at age 16 years and at least 1 sibling and 1 cousin with measures of SA and both parents with educational status (996 886 [53.5% of those eligible]). Educational status among biological parents was categorized into the following 5 age groups and standardized: (1) younger than 9 years, (2) 9 years, (3) 10 to 11 years, (4) 12 years, and (5) older than 12 years. We then performed a linear regression analysis with proband SA as the outcome. The predictor variable, which we called FCA, included the mean SA in siblings, mean SA in cousins, and educational status in the mother and father. The resulting residual was called the deviation in the FCA. We then used Cox proportional hazards regression models to investigate risk for SZ (and BPI) in individuals as a function of their SA, deviation in the FCA, and (in an additional model) their interaction. We then constructed and analyzed in the same way a second subsample of 106 187 men with recorded IQ and SA and IQs available on at least 1 sibling and 1 cousin and educational status of both parents from calculation of a second FCA. The relevant correlations in relatives were parent educational attainment of +0.49, sibling SA and IQ of +0.49 and +0.43, respectively, and cousin SA and IQ of +0.23 and +0.18, respectively.
In Cox proportional hazards regression models, robust standard errors were used to adjust the 95% CIs to reflect that the analytic sample contained individuals from the same family. Furthermore, we weighted each individual based on the number of siblings and cousins that were included in the FCA, weighted 0.5 and 0.125, respectively. Follow-up time in years was measured from the year of the SA (or IQ) until the year of the first registration for SZ (or BPI), death, emigration, or end of follow-up (end of 2012), whichever came first. In all models, we investigated the proportionality assumption, which examines whether the hazard ratio (HR) is constant over follow-up. When it was not fulfilled, we divided the effect into the following 3 periods: the first 4 years of follow-up, years 5 to 14, and after 14 years.
In the next step, we selected all possible full-sibling pairs from the source population (n = 1 765 096). We then selected all pairs where the proband was registered for SZ and his or her sibling was not (n = 2859). To each pair, we matched 3 pairs (n = 8577) where the control proband was matched on SA, year of birth, and sex, while the control sibling only had the same year of birth and sex as the proband sibling. The control proband could not have been registered for SZ. We then tested the difference in SA between the proband sibling and the control sibling using hierarchical linear model to account for the matching. All analyses were repeated using IQ for the male SZ probands (n = 396) and matched control probands (n = 1188). We then repeated these analyses with the same methods for BPI first matched on SA (9997 probands and 29 991 controls) and IQ (624 probands and 1872 controls). Statistical analyses were performed using a software program (SAS, version 9.3; SAS Institute Inc).17
Prediction of Risk for SZ and BPI From SA at Age 16 Years and Deviation From the FCA
The prevalence rates of SZ and BPI among our first cohort were 0.16% (1,570/996,889) and 0.58% (5750/996,865), respectively. The strongest predictors of FCA were, in order, SA in siblings, SA in cousins, and parents’ educational status (Table 1). The mean deviations from the FCA for individuals with SZ and BPI were −0.404 and −0.239, respectively. The HR for SZ was predicted robustly by the deviation from the FCA (HR, 0.56; 95% CI, 0.49-0.63) but not by the observed SA (HR, 1.01; 95% CI, 0.91-1.13). For BPI, the HR was significantly and equally predicted by the deviation from the FCA (HR, 0.83; 95% CI, 0.78-0.89) and by the observed SA (HR, 0.84; 95% CI, 0.79-0.88).
The predicted HRs for SZ as a function of SA and FCA, relative to an individual with a mean SA (0 SD) from a family with a mean FCA (0 SD), are shown in Figure 1. This figure can be usefully viewed in 2 ways. Focusing on level of FCA, in individuals with an average FCA, the HRs for SZ were estimated at 3.63, 1.75, 0.68, and 0.55, respectively, if their observed SA was substantially or moderately below the mean (ie, −2 or −1 SD) or moderately or substantially above the mean (ie, +1 or +2 SDs). Alternatively, we can attend to the observed SA. Taking those with an average SA (the white columns in Figure 1), the HRs for SZ were 0.35, 0.59, 1.68, and 2.83, respectively, if the FCA was substantially or moderately below the mean (ie, −2 or −1 SD) or moderately or substantially above the mean (ie, +1 or +2 SDs). Of note, an individual with low FCA (−2 SD) who performed as predicted from their family background (ie, had an SA −2 SDs below the population mean) had no significant increased risk for SZ (HR, 0.91). However, an individual with above-average SA (+1 SD) was at a 62% increased risk for SZ if their FCA was even higher (+2 SDs).
For both SZ and BPI, a significant interaction was observed between the deviation from the FCA and SA, indicating that this relationship was strongest for individuals with low observed SA. The proportionality assumption for the Cox proportional hazards regression model was not met for the deviation from the FCA because it more strongly predicted risk of illness onset in the earlier than in the later years of follow-up (Table 1).
SA in Unaffected Siblings of Individuals With SZ and BPI and Matched Controls
As shown in Figure 2, the siblings of the SZ probands had SA that did not differ from the population mean and was significantly higher than that of the siblings of the matched control probands (P = .003). By contrast, for BPI, the SA of the siblings of the probands and controls was almost identical (P = .67). Pearson product moment correlations in SA between the pre-SZ probands and their siblings (+0.43) was significantly lower than that observed between the matched control probands and their siblings (+0.52) (P < .001). Parallel figures for BPI were +0.45 and +0.51 (P < .001).
Prediction of Risk for SZ and BPI From IQ at Ages 18 to 20 Years and the Deviation From the FCA
The prevalence rates of SZ and BPI among our second cohort were 0.2% (180/106187) and 0.3% (339/106185), respectively (Table 2). The strongest predictors of FCA were, in order, IQ in siblings, educational status in parents, SA in siblings, and IQ and SA in cousins. The HR for SZ was strongly predicted by the deviation from the FCA (HR, 0.53; 95% CI, 0.37-0.77) but not by the observed IQ (HR, 1.07; 95% CI, 0.78-1.46). The HR for BPI was more weakly associated with the deviation from the FCA (HR, 0.77; 95% CI, 0.59-1.00) and not at all with the observed IQ (HR, 1.03; 95% CI, 0.83-1.28). For neither SZ nor BPI was there a violation of the proportionality assumption. A significant interaction was seen between IQ and the deviation from the FCA in the prediction of BPI but not SZ.
IQ in Unaffected Siblings of Individuals With SZ and BPI and Matched Controls
We conducted parallel analyses of IQ in the well siblings of the probands and matched controls with SZ and BPI (Figure 2). For SZ, the results were qualitatively similar to those obtained with SA but with a larger effect size. The siblings of the SZ probands had a mean IQ at the population average, while the siblings of the matched controls had mean IQs of −0.18 SD (P = .002). The sibling IQs for the BPI and control probands did not differ (P = .66). Pearson product moment correlation in IQ between the pre-SZ probands and their siblings (+0.40) was significantly lower than that observed between the matched control probands and their siblings (+0.48) (P < .05). Similar results of +0.39 and +0.47 were seen for BPI (P < .05).
We propose an innovative approach to clarify the underlying nature of the association between premorbid cognitive performance and risk for SZ that permitted us to determine the degree to which risk for SZ is reflected by (1) an individual’s observed cognitive performance in adolescence and (2) the degree to which the performance deviates from that predicted from their relatives (ie, their FCA). We conducted these analyses using the following 2 measures of cognitive performance: SA in both sexes at age 16 years and IQ in men at ages 18 to 20 years. Risk for SZ was strongly predicted by the deviation from the FCA and not at all by the actual SA. The same patterns of findings were observed for IQ. We are not the first to examine an FCA-like construct in SZ. Keefe et al18 considered a “cognitive function decrement” for SZ defined as current cognitive functioning compared with that predicted from reading level (an index of prior intellectual function) and maternal education.
We then sought to determine whether the lowered cognitive performance in the pre-SZ probands results largely from an unlucky combination of the genetic and environmental risk factors distributed within their sibship or from a qualitatively different process. Our results favored the second hypothesis because the SA of the siblings of the SZ probands was significantly higher than that of the siblings of the matched controls and only slightly different from the population mean. Using IQ, we found a similar but even more robust pattern. These results are not entirely consistent with most (but not all19) prior studies that report moderate decreases in cognitive performance in siblings of SZ probands vs controls.20,21 Finally, the proband sibling correlations in SA and IQ were significantly lower in families with a pre-SZ vs control proband.
These results have 2 major implications for the developmental theory of SZ. First, they suggest that the lowered cognitive performance or the neurobiological developmental precursors thereof do not, in and of themselves, impact strongly on risk for SZ. Rather, the neurodevelopmental process that increases risk for SZ is better indexed by a decline in cognitive performance from what would be expected given an individual’s FCA. Our cross-sectional findings analyzed using a co-relative design are thus congruent with prior longitudinal studies7,8 that find declining cognitive functioning over development to predict SZ risk, including recent evidence for a similar process in 22q11.2 deletion carriers.22 Second, we found consistent, albeit indirect, evidence that the pathological neurodevelopmental processes that predispose to SZ arise from etiologic factors that are distinct from the more normative factors that impact on cognition in these families.
It is helpful to interpret these results in the context of early studies23,24 that examined IQs of siblings of probands with very low vs moderately low IQ. Counterintuitively, the sibling IQs were substantially higher in the first than the second group. This finding is likely because the developmental processes impacting on very low IQ probands were typically chromosomal anomalies, birth trauma, or recessive mendelian disorders that were not present in their siblings. We suggest that a similar process may be occurring in SZ. Some pathogenic developmental insult of substantial effect size is preventing normal cognitive development in the pre-SZ individual but not, at least to a first approximation, in other members of a sibship. This result would explain both why the cognitive functioning of the unaffected siblings in such sibships is substantially greater than that in the siblings of the matched controls because those unaffected siblings were not exposed to the developmental insult. A major impact on cognitive performance in the pre-SZ probands but not the siblings would also cause the observed decrease in the proband sibling correlations in those families. While not a focus in this study, it is noteworthy that our findings are consistent with prior evidence that little of the association between SA, IQ, and SZ in the Swedish cohorts we have studied can be explained by transmissible genetic factors impacting on both traits.25-27
We conducted an identical set of analyses on risk for BPI to help contextualize our SZ findings. Consistent with prior findings of the association with premorbid cognitive performance and BPI,7,8,12-15 some results for BPI were similar to those found for SZ and others quite different. Like SZ, risk for BPI was predicted by the deviation from the FCA (whether measured using SA or IQ), although the magnitude of the association was substantially weaker. Unlike SZ, risk for BPI was predicted by the observed SA even after accounting for the deviation from the FCA. Unlike SZ, the cognitive performance of the siblings of BPI probands was indistinguishable from that of the siblings of the matched controls. Like SZ, correlations within siblings for SA and IQ were attenuated in the pairs containing an affected proband. These results suggest both quantitative and qualitative differences in the association of premorbid cognitive performance and disease risk for BPI vs SZ.
These results should be interpreted in the context of 5 methodological limitations. First, our sample was restricted to Sweden and may not extrapolate to other populations. Second, because our analyses required probands with information on SA or IQ in siblings and cousins and educational attainment in parents, our sample was not entirely representative of the Swedish population, having a slightly higher SA and IQ and a slightly lower risk for SZ and BPI.
Third, we relied on hospital diagnoses for SZ and BPI. Two studies have found using record reviews28 and diagnostic interviews29 that 96% and 94%, respectively, of Swedish cases with hospital diagnoses of SZ met DSM-IV criteria for SZ.30 Hospital diagnoses of BPI in Sweden have also been validated by record review.31
Fourth, the prevalence rates for SZ were quite low in our sample. This finding was likely a result of our selected cohort, narrow definition of illness, and limited years of follow-up (mean age at follow-up, 30 years).
Fifth, SA was evaluated differently in Sweden from 1987 to 1997 and after 1998. To test for the equivalence of the association with psychiatric outcomes between the 2 measures of FCA, we added a dummy variable for SA measured in 1987 to 1997 vs 1998 onward. We then examined the interactions between the dummy variable and the deviation from the FCA in predicting risk for SZ and BPI. None of these interactions were significant, indicating that the 2 measures of SA were equally predictive of our key outcomes.
Using a large population-based prospective cohort design, we found that the neurodevelopmental processes that impact on SZ risk are not well indexed by cognitive performance measured in adolescence per se. The risk process is much better reflected by the deviation in performance from that expected given the cognitive abilities of close relatives. Our results further suggest that the lowered cognitive performance of individuals who go on to develop SZ arises largely from a qualitative developmental impairment rather than just an unlucky combination of the genetic and environmental risk factors already present in their families.
Submitted for Publication: November 3, 2015; final revision received January 11, 2016; accepted January 13, 2016.
Corresponding Author: Kenneth S. Kendler, MD, Virginia Institute for Psychiatric and Behavioral Genetics, Virginia Commonwealth University, PO Box 980126, Richmond, VA 23298-0126 (email@example.com).
Published Online: March 30, 2016. doi:10.1001/jamapsychiatry.2016.0053.
Author Contributions: Drs Kendler and K. Sundquist had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Kendler, Ohlsson, K. Sundquist.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Kendler.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Kendler, Ohlsson.
Obtained funding: Kendler, J. Sundquist, K. Sundquist.
Administrative, technical, or material support: J. Sundquist, K. Sundquist.
Study supervision: Kendler.
Conflict of Interest Disclosures: None reported.
Funding/Support: This project was supported by the Swedish Research Council, the Swedish FORTE, and ALF funding from Region Skåne.
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
DR. Implications of normal brain development for the pathogenesis of schizophrenia. Arch Gen Psychiatry
. 1987;44(7):660-669.PubMedGoogle ScholarCrossref
RS. Schizophrenia is a cognitive illness: time for a change in focus. JAMA Psychiatry
. 2013;70(10):1107-1112.PubMedGoogle ScholarCrossref
B. Intelligence in schizophrenia: meta-analysis of the research. Schizophr Bull
. 1984;10(3):430-459.PubMedGoogle ScholarCrossref
LJ. Premorbid IQ in schizophrenia: a meta-analytic review. Am J Psychiatry
. 2008;165(5):579-587.PubMedGoogle ScholarCrossref
PB. A quantitative meta-analysis of population-based studies of premorbid intelligence and schizophrenia. Schizophr Res
. 2011;132(2-3):220-227.PubMedGoogle ScholarCrossref
et al. Static and dynamic cognitive deficits in childhood preceding adult schizophrenia: a 30-year study. Am J Psychiatry
. 2010;167(2):160-169.PubMedGoogle ScholarCrossref
et al. Decline in cognitive performance between ages 13 and 18 years and the risk for psychosis in adulthood: a Swedish longitudinal cohort study in males. JAMA Psychiatry
. 2013;70(3):261-270.PubMedGoogle ScholarCrossref
JC. Familial resemblance for general cognitive ability in the Hawaii Family Study of Cognition. Behav Genet
. 1985;15(4):401-406.PubMedGoogle ScholarCrossref
Jr. Genetic and environmental influences on human behavioral differences. Annu Rev Neurosci
. 1998;21:1-24.PubMedGoogle ScholarCrossref
et al. A longitudinal study of premorbid IQ score and risk of developing schizophrenia, bipolar disorder, severe depression, and other nonaffective psychoses. Arch Gen Psychiatry
. 2004;61(4):354-360.PubMedGoogle ScholarCrossref
et al. Premorbid intellectual functioning in bipolar disorder and schizophrenia: results from a cohort study of male conscripts. Am J Psychiatry
. 2005;162(10):1904-1910.PubMedGoogle ScholarCrossref
et al. Evidence for early-childhood, pan-developmental impairment specific to schizophreniform disorder: results from a longitudinal birth cohort. Arch Gen Psychiatry
. 2002;59(5):449-456.PubMedGoogle ScholarCrossref
et al. Excellent school performance at age 16 and risk of adult bipolar disorder: national cohort study. Br J Psychiatry
. 2010;196(2):109-115.PubMedGoogle ScholarCrossref
SAS Institute Inc. SAS/STAT User’s Guide. Version 9.3. Cary, NC: SAS Institute Inc; 2011.
et al. Sex differences in familiality effects on neurocognitive performance in schizophrenia. Biol Psychiatry
. 2013;73(10):976-984.PubMedGoogle ScholarCrossref
CS. Cognitive deficits in unaffected first-degree relatives of schizophrenia patients: a meta-analytic review of putative endophenotypes. Schizophr Bull
. 2006;32(1):179-194.PubMedGoogle ScholarCrossref
et al. Relative risk for cognitive impairments in siblings of patients with schizophrenia. Biol Psychiatry
. 2001;50(2):98-107.PubMedGoogle ScholarCrossref
et al; International Consortium on Brain and Behavior in 22q11.2 Deletion Syndrome. Cognitive decline preceding the onset of psychosis in patients with 22q11.2 deletion syndrome. JAMA Psychiatry
. 2015;72(4):377-385.PubMedGoogle ScholarCrossref
F. A population-based study of shared genetic variation between premorbid IQ and psychosis among male twin pairs and sibling pairs from Sweden. Arch Gen Psychiatry
. 2012;69(5):460-466.PubMedGoogle ScholarCrossref
K. IQ and schizophrenia in a Swedish national sample: their causal relationship and the interaction of IQ with genetic risk. Am J Psychiatry
. 2015;172(3):259-265.PubMedGoogle ScholarCrossref
J. A Swedish national prospective and co-relative study of school achievement at age 16, and risk for schizophrenia, other nonaffective psychosis, and bipolar illness. Schizophr Bull
. 2016;42(1):77-86.PubMedGoogle Scholar
PF. Recurrence risks for schizophrenia in a Swedish national cohort. Psychol Med
. 2006;36(10):1417-1425.PubMedGoogle ScholarCrossref
et al. Evaluation of diagnostic procedures in Swedish patients with schizophrenia and related psychoses. Nord J Psychiatry
. 2005;59(6):457-464.PubMedGoogle ScholarCrossref
American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 4th ed. Washington, DC: American Psychiatric Association; 1994.
N. Validity of bipolar disorder hospital discharge diagnoses: file review and multiple register linkage in Sweden. Acta Psychiatr Scand
. 2011;124(6):447-453.PubMedGoogle ScholarCrossref