This study used data from the 2009, 2010, and 2011 US National Health and Wellness Survey (NHWS) databases (Kantar Health, New York, NY). Data from these surveys were collected between January 1, 2009 and March 1, 2011. The NHWS is a cross-sectional, internet-based survey of adults (aged 18+). Using a stratified random sampling framework, potential respondents are recruited through internet panels such that the demographic composition of the NHWS is identical to that of the general population, as measured by the US Census. The reliability and validity of the NHWS, particularly as it relates to disease prevalence, has been assessed in prior research [11–13]. These studies have compared data collected from the NHWS and various sources including the US Census Bureau’s Current Population Survey, National Health Interview Survey, National Health and Nutrition Examination Survey, and the Medical Expenditure Panel Survey [11–13].
All participants of the NHWS agreed to participate in the survey explicitly and were awarded nominal incentives for their participation. The NHWS is approved each year by Essex Institutional Review Board (Lebanon, NJ). Of 1,157,205 who were invited to participate (from January 1, 2009 to March 1, 2011), 250,849 responded (21.68% response rate). Of those who responded, 175,000 (69.76%) met the study inclusion criteria (gave informed consent and were aged 18 or over) and completed the survey.
Bivariate analyses were conducted to describe and compare the demographic and health characteristics of patients being treated with aripiprazole versus other atypical antipsychotics. For categorical variables, chi-square tests were used to determine significant differences, while t-tests were used for continuous variables.
Although our original intent was to compare aripiprazole with each individual atypical antipsychotic, this was not possible due to small sample sizes of some of the comparators. As a result, we focused primarily on the comparison between aripiprazole and all other atypical antipsychotics pooled together. These main analyses compared those using aripiprazole with those using all other atypical antipsychotics on all HRQoL and health utility scores using t-tests. These differences were also examined using General Linear Models (GLM), controlling for the covariates described above in order to mitigate the effect of selection bias (i.e., the systematic reasons why a patient may be prescribed one atypical medication over another which also may influence HRQoL). Sub-analyses were also conducted to compare users of aripiprazole with users of each individual atypical antipsychotic, though only the comparison with quetiapine had sufficient sample size for any meaningful interpretation. Differences between the groups were first made on an unadjusted level using t-tests. Subsequent GLMs were then conducted to compare for group differences while controlling for the covariates mentioned above.
Adjusted means, which provide the mean HRQoL and utility scores when all covariates are set at the sample mean, were obtained through a least-squares algorithm for all GLMs (both in the main analysis and the sub-analysis). All models controlled for the following variables (reference variables are marked by an *): age (in years), gender (female*, male), ethnicity/race (non-Hispanic white*, non-Hispanic black, Hispanic, other), marital status (married or living with a partner, all else*), region (Northeast*, South, West, Midwest), household income (less than $25,000*, $25,000 to ≤ $50,000, $50,000 to ≤ $75,000, $75,000 or more), insurance coverage (yes, no*), year surveyed (2009, 2010*, 2011), exercise behavior (none in the past month*, 1 to 9 times in the past month, 10 or more times in the past month), alcohol consumption (none in the past month*, once or more in the past month), BMI (underweight or normal weight*, overweight, obese), comorbidity burden (Charlson Comorbidity Index as a continuous variable), prescriber type (psychiatrist, all else*), time since diagnosis (less than 4 years*, 5 to 10 years, 11 or more years), selective serotonin reuptake inhibitor use (yes, no*), selective norepinephrine reuptake inhibitor (yes, no*), tricyclic antidepressant (yes, no*), other antidepressant use (yes, no*).
Sensitivity analyses were also conducted to assess the robustness of the findings. Rather than entering all covariates simultaneously in the model, a series of hierarchical multiple regressions were conducted to predict all summary and domain HRQoL scores. For each outcome, the following process was initiated separately. In the first step, all demographic variables were entered (age, gender, ethnicity, region, marital status, household income, health insurance) into a multiple regression model. For the second step, only the significant variables from the prior step as well as all health characteristic information (smoking behavior, exercise behavior, alcohol consumption, BMI, Charlson Comorbidity Index, prescriber type, time since diagnosis, and concomitant treatments) were entered into a multiple regression model. For the third step, all significant variables from the prior step, as well as atypical antipsychotic use (aripiprazole versus other atypical antipsychotics) were entered into a general linear model. Only the adjusted means from the third, and final model, are reported. All analyses were conducted in SAS v9.1 (Cary, NC) and the a priori cutoff for statistical significance was p<.05.