Supplementary Online Content Hafeman DM, Merranko J, Goldstein TR, et al. Assessment of a person-level risk calculator to predict new-onset bipolar spectrum disorder in youth at familial risk. JAMA Psychiatry. Published online July 5, 2017. doi:10.1001/jamapsychiatry.2017.1763 emethods. Supplemental Methods etable 1. Decrement in Associated With Removing Pairs of Variables From the Risk Calculator etable 2. Decrement in Associated With Removing Variable Triplets From the Risk Calculator etable 3. Change in With Additional Variables This supplementary material has been provided by the authors to give readers additional information about their work.
emethods. Supplemental Methods Sample: Parents of children ages 6-18 years old were recruited between 2001 and 2007. Exclusion criteria were a pre-existing diagnosis of schizophrenia, mental retardation, or a mood disorder secondary to medical illness, substance or medication use. Follow-up is ongoing, and the retention rate thus far is 89%. Procedures: Participating biological co-parents (31%) were assessed by direct interview using the Structured Clinical Interview for DSM-IV. The psychiatric history of nonparticipating biological co-parents was obtained from the participant parent using the Family-History Research Diagnostic Criteria 1. All assessments of parents and children were performed by interviewers trained with the diagnostic instruments, and were reviewed by a child psychiatrist; all were blind to parental diagnoses. Kappa coefficients for all disorders were 0.70. Statistical Analysis: All analyses were conducted using R 2. We imputed missing data using multiple Multivariate Imputation by Chained Equations 3 (number of imputations=5). To account for within-family correlation, we also fit a hierarchical (Cox regression) model further accounting for this correlation, but the family covariance was estimated to be zero; thus our final model did not include this additional random component. As described in the main text, Harrell s algorithm for bootstrap optimism correction 4 was used for internal validation. Specifically, for each of 1000 bootstrap resamples, a model was trained on the bootstrap resample and then tested in the original sample. The mean train-test difference was considered an estimate of overfitting, and accounted for in the original estimate.
etable 1. Decrement in Associated With Removing Pairs of Variables From the Risk Calculator. All bootstrapped 95% confidence intervals included zero, meaning that there were no significant decrements. Removed Decrement Lower Upper Parental Age of Onset CGAS 0.06-0.002 0.12 Parental Age of Onset KMRS 0.05-0.01 0.12 Parental Age of Onset KDRS 0.04-0.02 0.1 Parental Age of Onset CALS 0.05-0.02 0.11 Parental Age of Onset SCARED 0.05-0.01 0.12 CGAS KMRS 0.01-0.04 0.07 CGAS KDRS 0.01-0.05 0.07 CGAS CALS 0.01-0.05 0.07 CGAS SCARED 0.01-0.05 0.06 KMRS KDRS 0.01-0.04 0.07 KMRS CALS 0.00-0.05 0.06 KMRS SCARED 0.00-0.06 0.06 CALS KDRS 0.00-0.05 0.06 CALS SCARED 0.01-0.05 0.06 SCARED KDRS -0.01-0.07 0.04
etable 2. Decrement in Associated With Removing Variable Triplets From the Risk Calculator. Significant decrements are in bold. Removed Decrement Lower Upper Parental Age of Onset CGAS KDRS 0.06-0.0005 0.12 Parental Age of Onset CGAS KMRS 0.07 0.01 0.13 Parental Age of Onset CGAS CALS 0.06-0.002 0.12 Parental Age of Onset CGAS SCARED 0.07 0.01 0.13 Parental Age of Onset SCARED KDRS 0.04-0.02 0.11 Parental Age of Onset KDRS CALS 0.04-0.02 0.1 Parental Age of Onset KMRS KDRS 0.06 0.0003 0.12 Parental Age of Onset KMRS SCARED 0.06-0.004 0.13 Parental Age of Onset KMRS CALS 0.05-0.01 0.11 Parental Age of Onset SCARED CALS 0.06 0.002 0.12 CGAS KDRS CALS 0.01-0.04 0.07 CGAS KDRS SCARED 0.01-0.05 0.07 CGAS KMRS CALS 0.01-0.04 0.06 CGAS KMRS SCARED 0.01-0.04 0.07 CGAS KMRS KDRS 0.04-0.02 0.1 CGAS SCARED CALS 0.02-0.04 0.07 KMRS KDRS CALS 0.01-0.04 0.07 KMRS SCARED CALS 0.00-0.05 0.06 KMRS SCARED KDRS 0.00-0.06 0.06 SCARED KDRS CALS -0.01-0.07 0.04
etable 3. Change in With Additional Variables Variable Added New Improvement Lower Upper Socioeconomic Status* 0.758-0.006-0.062 0.051 Major Depressive Disorder 0.770 0.006-0.051 0.059 Physical or Sexual Abuse 0.773 0.009-0.045 0.061 Live with Both Biological Parents 0.760-0.004-0.058 0.053 *Measured using the Hollingshead four factor index 5
References 1. Andreasen NC, Endicott J, Spitzer RL, Winokur G. The family history method using diagnostic criteria. Reliability and validity. Arch Gen Psychiatry. 1977;34(10):1229-1235. 2. R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.rproject.org/. 2017. 3. Raghunathan TE, Lepkowski JM, Van Hoewyk J, Solenberger P. A multivariate technique for multiply imputing missing values using a sequence of regression models. Survey methodology. 2001;27(1):85-96. 4. Harrell FE, Jr., Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Statistics in medicine. 1996;15(4):361-387. 5. Hollingshead AB. Four-Factor Index of Social Status. New Haven, Connecticut: Yale University Department of Sociology; 1975.