A full report of the methods employed within the NCS-R can be found in Kessler et al. [26]: The NCS-R undertaken by the World Mental Health team as “a nationally representative community household survey of the prevalence and correlates of mental disorders in the US” [26]. Data was collected between February 2001 and April 2003 [26]. Participants were selected from a nationally representative multi stage clustered area probability sample of households [26]. The study was subdivided into two sections, with all respondents receiving part 1 (N = 9282), with reported sampling strategies and criteria for selecting participants to receive part two (N = 5692) [26]. However, over and above these criteria, to reduce financial cost and participant burden, certain subsections (including family burden) were only administered to a 30% subsample where it was deemed that the data analysis goals could be achieved by administering the section to a probability subsample of respondents [26].
The recruitment, consent and field procedures within the NCS-R were approved by the Human Subjects Committees of both the Harvard Medical School and the University of Michigan [26]. The principles align with those of the declaration of Helsinki for ethical principles for medical research involving humans. Recruitment protocol took the form of an advance letter and study fact brochure, followed several days later by interviewer contact [26]. Interviewers used a standardised method to select a random respondent within each household, and obtained verbal informed consent [26]. Respondents were given a minimum $50 for participation [26]. If the initially selected participant declined, the invitation was extended to another person within the household [26]. The number of occasions on which this happened is not reported [26]. Reports are not given on how many households were contacted but then did not uptake participation in the survey [26], however persuasion letters were sent; and 60 days before the end of the closeout period, a special effort was made by sending a letter offering an increased financial incentive to complete an abbreviated interview either in person orby telephone [26]. It is however reported that interviews were only broken off by 107 out of 9389 initial NSCR respondents, and that the overall response rate was 74.6%, with 9282 completed interviews [26].
For methodological reasons, interviews were administered face to face in the homes of respondents using laptop assisted personal interview methods by professional survey interviewers [26]. Minimum survey completion time was 90 minutes (when no lifetime disorders reported), with average completion time being 2 hours 30 minutes, and stretching to 6 hours where complex history was present [26]. Interviewers gauged participant fatigue and suggested breaks where necessary, with time length of interviews varying from days to weeks depending on the complexity of the participants’ history [26]. Quality control procedures checked accuracy of responses recorded by interviewers and showed no evidence of any problems, with diagnoses being determined by computer algorithms [26].
Samples
The overall NCS-R sample consisted of 9,282 individuals [26]. However, the measures concerning family burden were only administered to a random 30% subsample [27]. Only this subsample (N = 3,192; 1,371 males and 1,821 females) (mean age = 46.1 years old (sd = 18.1 years) are analysed within this paper. Age ranges were 18–98 years old, with 34 years old being the modal age.
Mental disorder status
The structured interview administered was the version of the Composite International Diagnostic Interview (CIDI) that was developed for the WHO World Mental Health (WMH) (WMH-CIDI) [26, 27]. These disordersa included anxiety disorders (panic disorder, agoraphobia without panic, specific phobia, social phobia, generalized anxiety disorder, post-traumatic stress disorder, obsessive compulsive disorder, adult separation anxiety disorder, any other anxiety disorder), mood disorders (major depressive disorder, dysthymia, bipolar I-II sub disorders, any mood disorder), impulse-control disorders (oppositional-defiant disorder, conduct disorder, attention-deficit hyperactivity disorder, intermittent explosive disorder and any impulse control disorder), and substance disorders (alcohol abuse with/without dependence, drug abuse with/without dependence, nicotine dependence, any substance disorder). Disorders were assessed using the definitions and criteria of the DSM-IV. In the current analysis no account was taken of co-morbidity, thus an individual may have had one disorder or several.
Family burden
Family burden was assessed within section 42 of the NCS-R [28]. The instrument used to assess family burden was constructed for the purposes of the World Mental Health study, and considered elements of both objective (practical) and subjective (emotional) burden. No details of the psychometrics of the questionnaire are available. Herein, participants were asked a series of questions concerning their close family members (parents, siblings, children, spouse/partner), and the health problems of these individuals in terms of 12 illnesses. The list included both physical illnesses (cancer, serious heart problems, permanent physical disability such as blindness or paralysis, or any other serious chronic physical illness) as well as mental illnesses (serious mental problems like senility or dementia, mental retardation, alcohol or drug problems, depression, anxiety, schizophrenia or psychosis, manic depression, or any other serious chronic mental problem). These were all included within the mental health category given their inclusion in DSM. Participants responded indicating which of their kin experienced each condition.
If a participant indicated that any of their first degree relatives had any of the above conditions, they were then asked about family burden. This was indicated through a question in which participants rated the extent to which their own life was affected by the health problems of their relative. For analysis purposes, this was recoded as a dichotomous variable with those who reported that the health of their relative impacted upon their life ‘a lot’ or ‘some’ deemed as experiencing family burden, as opposed to those who responded as ‘a little’ or ‘not at all’. Using a yes/no format, those reporting such family burden were then asked further details concerning both objective burden (e.g. the provision of help with practical tasks such as washing, getting around or housework, or spending more time keeping company or giving more emotional support to ill relatives than they would otherwise) and subjective burden (e.g. psychological distress such as worry, anxiety or depression, or embarrassment). These details regarding format of burden type were not included in the current analyses due to the fact that the questions were only asked to those individuals who reported perceiving burden.
Data coding and analytic methods
In terms of considering the actual health profile of the relatives’ variables were (a) 12 count variables (0–4) representing the number of kin types with each of the 12 health conditions, and (b) four count variables (0–12) representing the number of condition types experienced by each kinship type (parents, siblings, spouse/partner, and children). The grouping of physical and mental illnesses has been outlined above. Demographics such as age, sex, marital status and education and household income were also included for consideration, as were the earlier described measures of family burden and personal mental health. Household income was recoded as those above and below the mean household income.
Descriptive statistics are reported, using crosstabs to obtain the appropriate level of detail. Binary logistic regressions were used to examine both hypotheses. In subsequent reporting of these, the outcome/criterion variables and the predictor variables are clearly stated within each analysis. All analyses were conducted within SPSS. The minimum level for statistical significance was .05 within all analyses. As the data to be analysed include some variables from part 1 and others from part 2, the part 2 sample and weights are used [26].