Anxiety- and depressive disorders are common, debilitating and associated with a marked decrease in quality of life as well as a large societal cost. Psychological treatments are generally effective but there exists large unexplained variability in treatment effects as well as a general and global shortage of resources allocated to psychiatric care. This makes the identification of predictors that can aid in person centred care approaches not only desirable for increasing quality but necessary for managing resources effectively. Identifying predictors that might inform clinical decisions on who should be offered psychological treatment, how much resources should be allocated to each patient and if treatment should be discontinued has the potential of increasing the quality of care as well as managing resources. Different types of comorbid psychiatric disorders have been proposed as predictors of treatment effect but previous research has identified few consistently replicated predictors. This might be due to shortcomings in previous ways of doing research such as the risk of mass significance problems hampering how many disorders can be tested within the same sample, a general lack of studies testing quadratic effects, use of samples that are not representative of clinical sample, problems with replication due to overfitting of models to sample data and difficulties in testing complex models with many variables that individually have small effects. Method: We have used machine learning algorithm on a clinically gathered outcome data from three psychiatric outpatient clinics in Sweden in order to determine what comorbidities are predictive of treatment effect, if comorbidity profiles influences how many therapy hours are needed before remission and if comorbidity profiles predict occurrence of sudden gain, a well-known within therapy predictor of treatment effect. Results: Several comorbidities explained a significant amount of variance in treatment outcome, sessions needed for remission and occurrence of sudden gains.