Assessment of Mental Disorders

The Dunedin Study longitudinally ascertains mental disorders every few years, interviewing members about past-year symptoms (eAppendix 1 in the Supplement). Private structured interviews were conducted by health professionals blinded to participants’ prior data using the Diagnostic Interview Schedule¹⁹ at ages 18, 21, 26, 32, 38, and 45 years. We studied DSM–defined symptoms of 14 disorders: externalizing disorders (attention-deficit/hyperactivity disorder, conduct disorder, alcohol dependence, tobacco dependence, cannabis dependence, and other drug dependence), internalizing disorders (generalized anxiety disorder, depression, fears [including social phobia, simple phobia, agoraphobia, and panic disorder], eating disorders [including bulimia and anorexia], and posttraumatic stress disorder), and thought disorders (obsessive-compulsive disorder, mania, and schizophrenia) (eFigure 1 in the Supplement). Ordinal measures represented the number of possible DSM–defined symptoms associated with each disorder. Past-year prevalence rates of mental disorders in the Dunedin Study cohort are similar to prevalence rates in nationwide surveys of the US and New Zealand.²⁰

The method used to evaluate the structure of psychopathology in the Dunedin Study cohort has been described previously^6,21 (eAppendix 2 in the Supplement). Briefly, confirmatory factor analysis at the symptom level was used to test 2 standard models²¹: (1) a correlated-factors model and (2) a hierarchical or bifactor model. Using a correlated-factors model (eFigure 2 in the Supplement), 3 factors were tested, representing externalizing (with loadings from attention-deficit/hyperactivity disorder, conduct disorder, and alcohol, cannabis, tobacco, and other drug dependence), internalizing (with loadings from depression, generalized anxiety disorder, fears/phobias, posttraumatic stress disorder, and eating disorders), and thought disorders (with loadings from obsessive-compulsive disorder, mania, and schizophrenia). The model fit the data well (eTable 1 in the Supplement), confirming that 3 correlated factors (ie, internalizing, externalizing, and thought disorder) explain well the structure of the disorder symptoms. A hierarchical or bifactor model (eFigure 2 in the Supplement) established that symptom measures reflect both general psychopathology and narrower styles of psychopathology. General psychopathology (labeled p-factor) is represented by a factor that directly influences all of the diagnostic symptom factors. This model fit the data well (eTable 1 in the Supplement). The p-factor captures how cohort members differ from each other in the variety and persistence of many different kinds of symptoms from ages 18 to 45 years. Previous work has shown that Dunedin Study participants with higher p-factor scores experienced a younger age at onset, greater number of assessment ages with a disorder, and greater diversity of diagnoses (r = 0.76; 95% CI, 0.74-0.79) (eFigure 3 in the Supplement).⁶