Extracting relevant predictors of the severity of mental illnesses from clinical information using regularisation regression models

Abstract Mental disorders are common non-communicable diseases whose occurrence rises at epidemic rates globally. The determination of the severity a mental illness has important clinical implications and it serves as prognostic factor for effective intervention planning management. This paper aims to identify relevant predictors illnesses (measured by psychiatric rating scales) from wide range variables consisting information on both laboratory test results factors. collectively indicate measurements 23 components derived vital signs blood tests evaluation complete count. 8 factors known affect considered, viz. family history, course onset an illness, etc. Retrospective data 78 patients diagnosed with behavioural were collected Lady Hardinge Medical College & Smt. S.K, Hospital in New Delhi, India. observations missing imputed using non-parametric random forest algorithm. multicollinearity is detected based variance inflation factor. Owing presence multicollinearity, regularisation techniques such ridge regression extensions least absolute shrinkage selection operator (LASSO), adaptive group LASSO used fitting model. Optimal tuning parameter λ obtained through 13-fold cross-validation. It was observed that coefficients quantitative extracted comparable regression.