Quantifying family history for mental illness

Family history, in the context of medicine, is made up of information about various disorders which have been suffered by blood relatives of a patient. Family history is a strong predictor in the development of many chronic diseases. It can provide information (both genetic and environmental) which can guide prevention and treatment with relatively low expense. One of the main issues, however, is the method by which one may generate a family history score for the purpose of analysis. A dichotomous measure (which takes the value of 1 if any relative of the patient is affected with the disorder, 0 otherwise), count of affected relatives, proportion of relatives that are affected within a family and scores involving prevalence and expectation of developing a disorder have been proposed in the literature. This project compares six existing methods of quantifying family history and proposes one new method, and outlines the advantages and disadvantages of each. A longitudinal psychological data set on bipolar disorder (BD) is used to illustrate the methods. The data set consists of individuals at high-risk of BD since one of their parents has BD. The methods are assessed and compared through exploratory data analysis and Cox proportional hazards modeling. The different score methods produced varying results emphasizing the importance of weighting, family size and age of relatives as variables within family score calculation. Implications and directions for future study are discussed.