Comprehensive evaluation of the risk of lactational mastitis in Chinese women: combined logistic regression analysis with receiver operating characteristic curve
Yongshuo Yin, Zhiyong Yu, Min Zhao, Yuemei Wang, Xiao Guan
Abstract
OBJECTIVE: To identify the potential risk factors for acute mastitis during lactation comprehensively. Subsequently, to evaluate logistic regression model in predicting the risk of lactational mastitis in Chinese women by applying receiver operating characteristic (ROC) curve. METHODS: A case-control study among Chinese women enrolled 652 patients with mastitis and 581 healthy women with breastfeeding experience as control. The retrospective information was obtained by questionnaires that included medical history of pregnancy, delivery, puerperium and breastfeeding behaviors. Univariate analysis and multivariate logistic regression model were performed to investigate the relationship between these factors and the occurrence of lactational mastitis. Using ROC curve to evaluate the prognostic value of these selected indicators in the risk of acute mastitis. RESULTS: The multivariate logistic regression analysis showed that the primiparity (P < 0.001), mastitis in previous breastfeeding (P < 0.001), nipple's heteroplasia (P < 0.001), cracked nipple (P < 0.001), breast trauma by external force (P = 0.002), lateral position (P = 0.007), breast pump (P = 0.039), nipple sucking (P = 0.007), sleep with sucking (P = 0.007), and tongue-tie (P = 0.013) were risk variables independently and significantly related with mastitis. While vaginal delivery (P = 0.015), clean nipple before breastfeeding (P = 0.015), first contact with child within 1 h (P = 0.027) were protective factors. The ROC analysis demonstrated that the area under the curve of model 2 was 0.8122 (95%CI = 0.7885-0.8360), which stated that the model presented a high sensitivity and specificity. CONCLUSION: By means of collecting and summarizing the risk factors associated with the occurrence of breast mastitis in Chinese women, we established risk discriminant model to identify and warn the individuals susceptible to acute mastitis early, which will allow practitioners to provide appropriate management advice and effective individual care.