Multinomial Logistic Regression Modeling to Determine the Antenatal Factors Affecting Birth Weight – a Case of Tigoni Level 4 Hospital, Kiambu County

Abstract

Background: The birth weight is one of the important determinants of the chances of a child surviving, having good physical and mental health. On average normal birth weight is between 2.5 kilograms (Kgs) -3.5 Kgs. Among the causes of high mortality and morbidity rates, LBW has been singled out as a significant cause. In 2013, an estimated 16 per cent of all babies born globally that year, had low birth weight, this is nearly 22 million babies. In Kenya studies have reported different prevalence of LBW ranging from 7-16%. Understanding the factors that are associated with birth weight is important in reducing the incidence of LBW and HBW and improving perinatal outcomes Objective: This study sought to determine and model the antenatal factors that affect the birth weight of babies by use of multinomial logistic regression model in one level 4 hospital (Tigoni Hospital, in Kiambu County) Methods: This was across-sectional descriptive study that used data that was retrieved from the maternity register (MOH 333 Version 25 (August 2016). Maternity records of a sample of 306 women who delivered between January and December 2018 was obtained and their information retrieved by use of a data retrieval form, to identify their profile. Stratified random sampling method was used to identify the sample: using the three categories of the outcome variable as the strata. Data was analyzed using STATA version 13. Results: The range of the age of the subjects was 17-44years; the median age was 26 years. Most of the women were having a second pregnancy and had made 3 visits to the ANC before visiting the hospital for delivery. 26.14 %( 80) of the women had made less than four ANC visits prior to delivery, 53.92% (165) of the mothers had given birth to male infants. The prevalence of LBW was high in women below the age of 20 years and those above the age of 35 years, in the latter age category the prevalence of HBW was considerably high as well (23.6%). LBW was more prevalent in single women compared to married women. The prevalence of LBW was comparable between women who had made at least four ANC visits and those who had less than four visits prior to delivery (35.39%, 36.25%) as was the case in the prevalence of HBW (27.87%, 26.25%).The prevalence of HBW was higher in male infants compared to the female infants (30.90%, 23.40%) while in the latter category, the prevalence of LBW was high (39.71%, 32.12%).The mean birth weight was 3151 grams. The prevalence of HBW, LBW and NBW was 3.09%, 7.26 % and 89.65% respectively (121, 284 and 3506 given N=3911). There was no enough statistical evidence (significance level = 0.05) that any of the predictor was individually associated with birth weight. However, age category, No. of ANC visits and gravida had close statistical association (p-value 0.05, 0.07, 0.19) respectively. Mothers who had an age of less than 20 years were less likely to give birth to high birth weight infants by 71.9% compared to middle aged mothers of age 20 – 35 years (OR: 0.281, p-value: 0.04, 95% CI: 0.07,0.98). Adjusting for all other variables, a unit increase in the number of ANC visits was shown to increase the likelihood of having a high birth weight infant by 38.1 % compared to normal birth weight (OR: 1.381, p-value: 0.01, CI:1.05,1.81). Conclusion: The study concluded that mothers who had an age of less than 20 years and those with an age of above 35 years were more likely to give birth to low birth weight infants whereas those in the latter category are also more likely to have HBW infants, an increase in the number of ANC visits was shown to increase the likelihood of having a high birth weight infant. The number of ANC visits, maternal age, gravida, and sex of new born are associated with birth weight. The data captured in the maternity register (MOH 333) may not be enough to adequately answer this question; investigators who may wish to answer such a question should consider a prospective study and explore more variables. The Multinomial Logistic Regression Model was a good model for answering the questions in the study, it is recommended to other researchers who may wish to model birth weight in more than two categories.