keywords: Mixture model, Age, length of stay, EM algorithms, maternity, complications
In literature one of the risk factors for maternity length of stay (MLOS) is age. Women are mostly fertile at age less than 25 years and it greatly reduced thereafter with associated complications. The population of the study comprises of 701 pregnant women that visited 4 hospitals in Funtua, Katsina State, Nigeria for childbirth. We considered hospital type, educational qualification, occupation, number of procedures, number of diagnosis, mode of delivery, parity, location, mother’s weight and child’s weight as risk factors for MLOS and proposed a mixture of gamma regression model to unearth the heterogeneous in MLOS data for two age groups of women: age less than 25 years and 25 years and above. The expectation maximization (EM) algorithm was used to estimate the model parameters and employed posterior probability for classification of objects into components. The results of the analysis show existence of two components in each of the two groups corresponding to short-stay patient and long-stay patient components. The proportions of patients in short-stay and long-stay components are 0.79 and 0.21 with average length of stay 1.34 days and 5.36 days, respectively for women in age group less the 25 years. Similar trend is observed in age group 25 years and above. There are 9 risk factors (hospital type, location, educational qualification, model of delivery, mother weight, baby weight and number of procedures,) that affect long-stay compare to four factors (educational qualification, model of delivery and number of procedures) that affect short-stay in age group 25 years and above. Only four factors (hospital type, educational qualification, model of delivery and mother weight) affect the short-stay and long-stay in the other group (ageless 25 years). Therefore, more resource allocations will be needed in terms of intervention for older pregnant women.