The model is fully adjusted for year of conscription, socioeconomic status, age, parental serious mental illness, paternal age, alcohol misuse, illicit substance use, common mental disorder diagnosis, IQ score, and other personality measures at conscription. The average personality trait group served as reference (hazard ratio [HR], 1.00).
eMethods. Serious Mental Illness Diagnosis
eTable. Sensitivity Analysis
Customize your JAMA Network experience by selecting one or more topics from the list below.
Hayes JF, Osborn DPJ, Lewis G, Dalman C, Lundin A. Association of Late Adolescent Personality With Risk for Subsequent Serious Mental Illness Among Men in a Swedish Nationwide Cohort Study. JAMA Psychiatry. 2017;74(7):703–711. doi:10.1001/jamapsychiatry.2017.0583
Are adolescent personality traits (social maturity, mental energy, and emotional stability) associated with the risk for subsequent serious mental illness (bipolar disorder, schizoaffective disorder, schizophrenia, and other nonaffective psychoses)?
In this nationwide longitudinal cohort study of more than 1 million Swedish men, social maturity, mental energy, and emotional stability had a dose-response association with schizophrenia and other psychotic illnesses. Bipolar disorder had a unique U-shaped association with social maturity (with high and low levels increasing risk).
Individuals who develop serious mental illnesses have different patterns of premorbid personality, and genetic architecture for personality traits and serious mental illness may be shared.
Understanding of personality as an independent risk factor for serious mental illness (SMI) remains limited. Recently, overlap between the polygenic basis for specific personality traits and specific SMIs has been identified.
To determine the association of the adolescent personality domains of social maturity, mental energy, and emotional stability with later diagnosis of SMI.
Design, Setting, and Participants
This longitudinal cohort study enrolled Swedish male military conscripts aged 18 or 19 years from January 1, 1974, through December 31, 1997. The diagnosis of an SMI was extracted from the National Patient Register for all inpatient treatment episodes in Sweden from January 1, 1974, through December 31, 2011. Data were analyzed from May 3 to September 16, 2016.
Social maturity, mental energy, and emotional stability assessed at conscription interview.
Main Outcomes and Measures
Inpatient diagnoses of bipolar disorder, schizoaffective disorder, schizophrenia, and other nonaffective psychoses occurring until December 31, 2011.
Of the 1 017 691 men included in the cohort, 4310 developed bipolar disorder; 784, schizoaffective disorder; 4823, schizophrenia; and 5013, other nonaffective psychoses. After adjustment, with use of mean scores as a reference, low social maturity (hazard ratio [HR], 1.61; 95% CI, 1.48-1.74), low mental energy (HR, 1.34; 95% CI, 1.24-1.44), and low emotional stability (HR, 1.51; 95% CI, 1.40-1.63) were inversely associated with schizophrenia in a dose-dependent fashion. Other nonaffective psychoses displayed a similar pattern. Bipolar disorder was associated with high (HR, 1.21; 95% CI, 1.09-1.35) and low (HR, 1.12; 95% CI, 1.01-1.25) social maturity and low emotional stability (HR, 1.62; 95% CI, 1.46-1.78). Schizoaffective disorder was associated with low emotional stability (HR, 1.53; 95% CI, 1.26-1.85).
Conclusions and Relevance
Emotional stability is inversely associated with all SMI. Bipolar disorder has a unique U-shaped association with social maturity. Premorbid personality may reflect subtle changes in cerebral function, may combine with symptoms and other neurocognitive deficits to influence illness presentation, and/or may be owing to shared genetic architecture.
Kraepelin1 first raised the question of whether specific premorbid personality traits reflect an early stage of illness or a genetic disposition to serious mental illness (SMI). However, our understanding of personality as an independent risk factor for SMI remains limited. Recently, a polygenic basis for neuroticism and a shared genetic architecture with schizophrenia has been found,2,3 with inconsistent results for bipolar disorder.2-4 Furthermore, polygenic risk scores for bipolar disorder are associated with extraversion.5 This research has prompted renewed interest in the association between personality and SMI.6
Low extraversion and high neuroticism have been identified as features in patients with nonaffective psychosis7-11 and bipolar disorder.12 However, these studies do not elucidate the temporal association between personality and SMI. A small number of longitudinal cohort studies have found that low self-esteem and neuroticism are present before development of nonaffective psychotic symptoms13,14 and before diagnosis of schizophrenia.15-17 Personality-related risk factors for bipolar disorder and schizoaffective disorder have been less frequently investigated. One study found high extraversion was associated with bipolar disorder.15 In general, however, evidence of a premorbid diathesis in bipolar disorder is lacking.17-19 We used data from Swedish conscript interviews to examine whether premorbid social maturity (including extraversion), mental energy, and emotional stability (including neuroticism) were associated with development of bipolar disorder, schizoaffective disorder, schizophrenia, and other nonaffective psychotic disorders.
We conducted a cohort study using longitudinal Swedish population registers. Personality data from a psychologist-conducted semistructured interview were derived from the Military Service Conscript Register and the Military Archives of Sweden. Interviews were conducted with nearly all male members of the Swedish population from January 1, 1974, through December 31, 1997, when aged 18 or 19 years. This register also includes IQ test results, physician physical health review, psychiatrist mental health review, and other sociodemographic data. The diagnosis of an SMI was extracted from the National Patient Register for all inpatient treatment episodes in Sweden from January 1, 1974, through December 31, 2011. Parents of conscripts were identified in the Multi-Generational Register and linked to the National Patient Register to identify parental SMI. Parental socioeconomic status (SES) data were derived from census information from January 1, 1970, through December 31, 1990, and linked through the Multi-Generational Register. The study was approved by the research ethics committee at the Karolinska Institute, Stockholm, Sweden, which waived the need for informed consent for these publicly available data.
The personality traits assessed by military psychologists were social maturity, mental energy, and emotional stability, each on a 5-point Likert scale. Psychologists were instructed to assign grades based on a quasi-normal distribution. For the purposes of this study, we categorized each exposure as low (1 or 2), average (3), or high (4 or 5). Assessment was done through semistructured interviews lasting approximately 20 to 30 minutes. Interviews addressed topics such as adjustment problems, psychosomatic ailments, managing stress, successes, and initiative. Interrater reliability was assessed repeatedly throughout the use of the interview tool and was consistently found to be high.20-22 The interview schedule is still in use today in the Swedish military and other government agencies.20 The psychological concepts assessed have been defined previously.23,24 In brief, social maturity reflects extraversion and sense of responsibility. Mental energy is defined as the ability to take initiative, persevere, and be task oriented. Emotional stability is the ability to control and channel nervousness, tolerate stress, and manage anxiety. Low emotional stability in these conscripts has previously been identified as a risk factor for future suicide,25,26 criminality,27 peptic ulcer disease,28 unemployment,29 and coronary heart disease.30 Definitions of the 3 traits agree well with the original version of the Minnesota Multiphasic Personality Inventory, which was used to identify combat fighters in a number of countries.31 With relation to the Big Five personality traits, high values on the 3 personality variables used in this study are analogous to high levels of conscientiousness (social maturity and mental energy), extraversion (social maturity), agreeableness (social maturity and emotional stability), openness to experience (mental energy), and low levels of neuroticism (emotional stability).32,33
Bipolar disorder, schizoaffective disorder, schizophrenia, and other nonaffective psychotic illness were coded according to Swedish versions of the International Classification of Diseases, Eighth Revision (ICD-8), International Classification of Diseases, Ninth Revision (ICD-9), and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) (eMethods in the Supplement).34 Diagnosis date was defined as the date of the first psychiatric inpatient record. If an individual received multiple SMI diagnoses, we used the latest, because at that time the treating clinician would have the most information about the course of the individual’s illness.
Data were analyzed from May 3 to September 16, 2016. We assessed the association between personality measures using polychoric correlations. Cox proportional hazard regression analyses compared SMI rates in individuals with high, average, and low levels of social maturity, mental energy, and emotional stability (with the average levels as a reference). We investigated the proportionality assumption using Schoenfeld residuals.35 We performed a complete case analysis in which individuals missing psychological or IQ assessments at baseline were excluded. Individuals who had admissions for SMI before their conscription interview were excluded from the analysis, as were individuals who were assessed as having psychosis at the conscription interview (this assessment was completed by a military psychiatrist). For each outcome, individuals were followed up from the conscription interview until the date that they developed the outcome or were censored at the date of receiving another SMI diagnosis, date of death, date of migration from Sweden, or December 31, 2011, whichever was earliest.
We adjusted for year of conscription, age at conscription, SES using Statistics Sweden’s socioeconomic classification (defined as the highest level of employment of either parent when the individual was aged 15 years as unemployed, worker, white collar worker, business owner, or missing), paternal age (in 10-year age ranges), parental SMI diagnosis at any time, IQ score (grouped as <74, 74-81, 82-89, 90-95, 96-104, 105-110, 111-118, 119-126, or >126), diagnosis of common mental disorder (CMD) (anxiety or depression diagnosis made by the military psychiatrist or before conscription interview), and alcohol and illicit drug use disorders at conscription. These confounders were chosen a priori based on previous research. In addition, we adjusted for each of the personality measures (social maturity, mental energy, and emotional stability) in multivariable regression. Therefore, each of the effect estimates is independent of the other personality measures and confounders. From this fully adjusted model, we calculated the proportional reduction in SMI in the population if each personality measure was average, rather than low or high; this is the population-attributable fraction (PAF). The PAF assumes that the association between the personality trait and SMI is causal. We conducted a sensitivity analysis excluding individuals who developed SMI within 5 years after their conscript interview to remove individuals who were potentially experiencing personality changes due to prodromal symptoms before receiving a diagnosis. All analyses were completed using Stata software (version 14; StataCorp).36
From 1974 through 1997, 1 017 691 men without preexisting SMI attended enlistment interviews that included personality and IQ assessments. Of these, 570 individuals were excluded because of preexisting SMI. The distribution of personality traits is presented in Table 1. Only 3.7% of the cohort had missing SES or paternal age data. During the follow-up period, 4310 individuals developed bipolar disorder; 784, schizoaffective disorder; 4823, schizophrenia; and 5013, other nonaffective psychoses. The rate of incident SMI was 53.36 per 100 000 person-years at risk (95% CI, 52.48-54.26 per 100 000 person-years at risk). None of the regression models violated proportionality assumptions or gave evidence of multicollinearity.
Associations with increased bipolar disorder were found for low (hazard ratio [HR], 1.67; 95% CI, 1.54-1.82) and high (HR, 1.12; 95% CI, 1.03-1.22) deviations from average social maturity (Table 2). Associations with increased rates were also found for low mental energy (HR, 1.51; 95% CI, 1.39-1.65) and low emotional stability (HR, 1.91; 95% CI, 1.77-2.07) (Table 2). After adjustment for the other personality measures, year of conscription, age at conscription, SES, parental SMI, paternal age, IQ score, CMD, and alcohol and illicit drug use disorders at conscription, the HR for bipolar disorder remained elevated for low (HR, 1.12; 95% CI, 1.01-1.25) and high (HR, 1.21; 95% CI, 1.09-1.35) compared with average social maturity. The association between mental energy and rate of bipolar disorder was attenuated such that mental energy was no longer associated with bipolar disorder (Table 2 and Figure). Low emotional stability continued to be associated with increased bipolar disorder rates (HR, 1.62; 95% CI, 1.46-1.78) (Table 2 and Figure). The multivariable adjusted PAF for bipolar disorder was 7% (95% CI, 4%-11%) for social maturity and 12% (95% CI, 9%-15%) for emotional stability. After excluding individuals who developed bipolar disorder within 5 years after their conscription interview, the associations remained (eTable in the Supplement).
We found a dose-response association for each personality domain in unadjusted analysis, such that individuals in the high category had reduced rates of schizoaffective disorder compared with average, and those with low scores had increased rates (Table 2). After adjustment, only low emotional control was associated with increased rates of schizoaffective disorder (HR, 1.53; 95% CI, 1.26-1.83) (Table 2 and Figure). This finding is equivalent to a PAF of 11% (95% CI, 5%-17%). The association between emotional stability and schizoaffective disorder remained after excluding individuals who developed schizoaffective disorder within 5 years after the personality assessment (eTable in the Supplement).
Unadjusted results suggested dose-response effects similar to those for schizoaffective disorder but with effect estimates further from unity. After adjustment, low social maturity (HR, 1.61; 95% CI, 1.48-1.74), mental energy (HR, 1.34; 95% CI, 1.24-1.44), and emotional stability (HR, 1.51; 95% CI, 1.40-1.63) were independently associated with schizophrenia (Table 2 and Figure). Reduced rates of schizophrenia were found in individuals with high social maturity (HR, 0.86; 95% CI, 0.77-0.95), mental energy (HR, 0.87; 95% CI, 0.76-0.95), and emotional stability (HR, 0.80; 95% CI, 0.71-0.90). If all individuals scored within the average range, PAFs would be 15% (95% CI, 12%-18%) for social maturity, 9% (95% CI, 6%-11%) for mental energy, and 14% (95% CI, 11%-16%) for emotional stability. After exclusion of individuals who developed schizophrenia within 5 years after the personality assessment, results remained similar (eTable in the Supplement).
The HRs were similar to those for schizophrenia. Again, we found an inverse gradient with elevated hazards in those with low social maturity (HR, 1.54; 95% CI, 1.42-1.67), mental energy (HR, 1.25; 95% CI, 1.15-1.35), and emotional stability (HR, 1.40; 95% CI, 1.30-1.51) (Table 2 and Figure). The PAFs were 13% (95% CI, 10%-16%) for social maturity, 6% (95% CI, 3%-9%) for mental energy, and 11% (95% CI, 8%-14%) for emotional stability. After excluding individuals who developed nonaffective psychosis within 5 years after the assessment, the association between personality traits and SMI was maintained (eTable in the Supplement).
Late adolescent social maturity, mental energy, and emotional stability were inversely associated with schizophrenia and other nonaffective psychotic illnesses in a dose-dependent fashion after accounting for important confounders. Bipolar disorder had a U-shaped association with social maturity and was associated with low emotional stability. Schizoaffective disorder was solely associated with low emotional stability. Associations remained after excluding men who became ill within 5 years after their personality assessment, suggesting that these results are not purely attributable to prodromal changes in personality.
To our knowledge, we have presented the largest population-based longitudinal study to examine personality traits and risk for subsequent SMI. In keeping with previous research,13-17 we found that individuals with low emotional stability (ie, high neuroticism) had increased rates of schizophrenia and other nonaffective psychosis. We believe that this study is the first to find that neuroticism is associated with increased rates of bipolar and schizoaffective disorder. Previously, a study found that premorbid introversion is associated with schizophrenia and nonaffective psychosis.15 Our measure of social maturity includes extraversion as a central characteristic, and social maturity was similarly associated with schizophrenia and other nonaffective psychotic disorders. This study also found that premorbid extraversion was associated with an increased risk for subsequent bipolar disorder.15 However, ours is the first to show that low and high social maturity are associated with bipolar disorder. Owing to the way that bipolar disorder diagnosis was recorded (using ICD-8, ICD-9, and ICD-10), we were unable to examine whether differences in social maturity were linked to polarity of illness or psychotic vs nonpsychotic bipolar disorder, which could potentially explain the U-shaped association. A previous study involving patients with preexisting bipolar disorder37 found that extraversion was associated with mania and introversion with depression. Social maturity and mental energy are likely to capture elements of the hyperthymic temperament,38 which has been found to be common in individuals with bipolar disorder and their apparently healthy siblings.39 Therefore, the lack of association between mental energy and bipolar disorder is of interest. Schizoaffective disorder appears to be intermediate to bipolar disorder and schizophrenia in terms of associations with personality measures. We might expect this finding because of the mix of affective and psychotic symptoms in these patients. Schizoaffective disorder often has cross-sectional symptomatologic similarities with schizophrenia and longitudinal illness course similarities with bipolar disorder.40 In each analysis, no evidence of strong confounding by age, year, IQ score, paternal age, CMD, alcohol or illicit drug use disorders, or parental SMI was found.
Our results suggest that neuroticism is a wide-ranging risk factor for the 4 categories of SMI that we defined, perhaps because it signals vulnerability to stressors,41 influences cognitive interpretation of events, or reflects subtle changes in cerebral function42,43 or purely because of shared genetic risk for emotional instability and SMI.2,44 Neuroticism has been found to be associated with reduced glucose metabolism in several prefrontal regions (including the dorsal medial prefrontal cortex and inferior and middle frontal gyrus) that are implicated in mood disorders and schizophrenia.45,46 The PAFs calculated in this study suggest that, if the association were causal, modifying neuroticism so that all individuals scored within the normal range would reduce the prevalence of each SMI by approximately 12%. Low emotional stability may be modifiable with specific forms of psychotherapy47 and serotonergic drugs48,49 and thus may represent an appropriate target for interventions.50 Addressing neuroticism could also have a positive effect on the duration of untreated psychosis,51 risk for CMD,52 and physical health outcomes.53,54
Similarly, reduced social maturity and extraversion may reflect an inability to manage stressors leading to increased SMI rates, or a shared genetic risk for SMI and the personality trait may be present. However, our results suggest that the association may be specific to nonaffective psychosis and that high and low social maturity are associated with bipolar disorder. Extraversion is positively correlated with resting glucose metabolism in the striatum and posterior temporal region, reflecting high dopamine transmission.45,55 High levels of dopamine transmission have been related to emotions such as elation, euphoria, and desire,56 symptoms consistent with a diagnosis of mania. Again, abnormalities in social maturity may be amenable to psychopharmacologic57 and psychotherapeutic50 intervention to reduce SMI incidence on a population level. Potentially normalizing social maturity would reduce rates of bipolar disorder by 7%, schizophrenia by 15%, and other nonaffective psychosis by 13%. Cross-diagnostic and quantifiable constructs associated with neural metabolism and shared genetic risk factors for SMI, neuroticism, and extraversion may be useful measures in the National Institute of Mental Health research domain criteria to develop a biologically driven psychiatric classification system.58
Interpretation is hindered by lack of validation of the rating of psychological traits and an understanding of how social maturity and mental energy map to more contemporary personality concepts. This lack may reduce comparability with other studies. However, the fact that the entire population of men received the same assessment allows standardization of scores and aids interpretation and generalizability. In addition, evidence suggests that semistructured interviews of this type have high validity and interrater reliability,21 and the interview remains in use.20 We were unable to test the stability of the personality measures because each individual received assessment only 1 time. However, the interviews focused on persistent personality traits rather than recent deviations. Evidence suggests that personality traits increase in stability during adolescence and remain relatively stable after that.59
Use of hospital records for SMI diagnoses captures only the most severely unwell individuals. Outcome misclassification in schizophrenia is unlikely because more than 90% of individuals with the diagnosis will be admitted to the hospital at some point during their illness.60 However, individuals with bipolar disorder may be less representative of all cases in the population because hospitalization is likely to be more variable. If this misclassification were nondifferential with regard to each of the personality traits, our results would represent an underestimate of the true association between personality and development of SMI. If high levels of social maturity, mental energy, and emotional stability represent improved life and coping skills, we could hypothesize that they would differentially reduce rates of hospitalization; however, we did not see this in bipolar disorder, and thus selection bias seems to have been unlikely. Misclassification because of the use of clinical diagnoses of SMI, rather than operationally defined research diagnoses, is also possible. If this misclassification were random with respect to each of the personality measures, it would tend to reduce differences among SMI groups. However, differential misclassification is possible, for example, if individuals with high social maturity were more likely to receive a diagnosis of bipolar disorder than schizophrenia because of their apparently high social functioning.
We included all individuals who had complete personality and IQ tests and excluded 75 647 men who attended interviews but had incomplete data. The excuded group had a rate of SMI (58.37 per 100 000 person-years at risk; 95% CI, 53.78-62.07 per 100 000 person-years at risk) similar to that of included individuals; thus, missingness (ie, the manner in which data are missing from a sample of the population) does not appear to be associated with outcome. Until 2000, conscription was mandatory and only 3% to 5% of Swedish men were exempt from attending because of severe functional or intellectual disability,61 which suggests that this cohort is representative of all men in Sweden. The only confounders with missing data were SES and paternal age, and no evidence suggests that 3.7% missing data would result in biased results from complete case analysis.62
The potential for residual and unmeasured confounding in our fully adjusted model remains. For instance, we did not include measures of urban living or educational attainment, which could be associated with personality and SMI. However, for these (or any other) covariates to have important effects on the results presented, they would have to be strongly associated with the personality measures and with SMI and be independent of other confounders included in the regression model.63,64
This study indicates clearly and robustly that low social maturity, mental energy, and emotional stability are associated with nonaffective psychotic illnesses in general. However, individuals who developed bipolar disorder had different patterns of these personality traits. Personality dimensions may therefore be useful in understanding SMI endophenotypes, and future research should focus on shared genetic influences on personality and SMI.
Corresponding Author: Joseph F. Hayes, MB, ChB, MSc, Division of Psychiatry, University College London, 6th Floor Maple House, 149 Tottenham Court Rd, London W1T 7NF, England (email@example.com).
Accepted for Publication: February 25, 2017.
Published Online: May 24, 2017. doi:10.1001/jamapsychiatry.2017.0583
Author Contributions: Dr Hayes had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Hayes, Osborn, Dalman, Lundin.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Hayes.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Hayes, Lundin.
Obtained funding: Hayes, Dalman.
Administrative, technical, or material support: Hayes, Lundin.
Study supervision: Hayes, Osborn, Lewis, Dalman.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was supported by grant MR/K021362/1 from the Medical Research Council and grant 523-2010-1052 from the Swedish Research Council.
Role of the Funder/Sponsor: The funding sources 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.
Create a personal account or sign in to: