Factor Analysis of the Prediction of the Postpartum Depression Screening Scale.

Postpartum depression (PPD), a severe form of clinical depression, is a serious social problem. Fortunately, most women with PPD are likely to recover if the symptoms are recognized and treated promptly. We designed two test data and six classifiers based on 586 questionnaires collected from a county in North Carolina from 2002 to 2005. We used the C4.5 decision tree (DT) algorithm to form decision trees to predict the degree of PPD. Our study established the roles of attributes of the Postpartum Depression Screening Scale (PDSS), and devised the rules for classifying PPD using factor analysis based on the participants' scores on the PDSS questionnaires. The six classifiers discard the use of PDSS Total and Short Total and make extensive use of demographic attributes contained in the PDSS questionnaires. Our research provided some insightful results. When using the short form to detect PPD, demographic information can be instructive. An analysis of the decision trees established the preferred sequence of attributes of the short form of PDSS. The most important attribute set was determined, which should make PPD prediction more efficient. Our research hopes to improve early recognition of PPD, especially when information or time is limited, and help mothers obtain timely professional medical diagnosis and follow-up treatments to minimize the harm to families and societies.