HANDLING OVER DISPERSION WITH NEGATIVE BINOMIAL AND GENERALISED POISSON REGRESSION MODELS IN ANALYSIS OF ANTENATAL VISITS UTILISATION IN NIGERIA

Abstract

Majority of pregnancy related complications and deaths can be prevented through the utilization of antenatal care (ANC) services. Utilisation of ANC service is measured through number of visits made to health facilities. The number of visits to an ANC facility is a count data and it is often, modeled by the Poisson regression. However it is sometimes erroneously used in situations where over dispersion of the response variable exist i.e. variance exceeds the mean. Negative Binomial and Generalised Poisson regression models are alternative models for estimating regression parameters in the presence of over dispersion. This study examined the pattern of ANC xiiutilization, determined the factors that affect the optimum utilisation of ANC visits and compared Poisson, Negative Binomial and Generalized Poisson regression models to determine the best statistical model which describes the utilisation of ANC visits. A nationally representative sample of women within reproductive age 15-49 years within households in communities was obtained from the National Demographic and Health Survey (NDHS) of 2013. The number of ANC visits and optimal utilization defined as four or more ANC visits were outcome variables and other explanatory variables include age, region. employment status. wealth index. husband's/partner's employment status, husband's/partner's educational level. place of residence and place of ANC. Binary logistic regression was used to determine the factors associated with optimal utilization, while the Poisson, Negative Binomial and Generalized Poisson regression models were also used to determine the factors associated with number of ANC visits. The best model was selected based on the values of-2logL, AIC and BIC selection test/criteria. Analysis was done on SPSS version 20 and STA TA version 12. The mean age of the women was 29 years (SD=7.0), 49.2% had no formal education and 23.5% belonged to the poorest wealth quintile. About 65% had at least one visit while 52.5% achieved optimum utilisation. Several factors were found to affect optimal utilization some of which include age; (35-49 years) (OR=l .354, 95% CI: 1.093, 1.678), geopolitical zone; south west (OR= 4.396, 95% Cl: 3.493. 5.533), higher educational level (OR= 1.883, 95% Cl: 1.359. 2.610). Of the three regression models Generalized Poisson regression had the least; -2logL of 81230.048, AIC=81282.048 and BIC=81475.433. Age; (35-49 years) (IRR= 1.142, 95% CI: 1.053, 1.240). south west geopolitical zone; (IRR=l.681. 95% Cl: 1.589, 1.781) and rural place of residence; (IRR=0.910. 95% CI: 0.874. 0.947) were amongst other factors as determinants of the number of ANC visits. The factors associated with number of ANC visits include age, geopolitical zone, educational status, wealth index, husband's/partners educational and employment status. In the presence of over dispersion, Generalised Poisson Regression was found to be the best regression model to estimate the factors which affect the number of ANC visits.