| dc.description.abstract | Background: The length of stay (LOS) is an important indicator of the efficiency of hospital management. Reduction in the number of inpatient days, results in decreased risk of infection and medication side effects, improvement in the quality of treatment, and increased hospital profit with more efficient bed management.
Objective: The aim of this study was to examine the use of count data analysis modelling to determine predictors of length of stay for neonates admitted in NICU in Kawempe NRH, in order to help the hospital manage the duration of inpatient stay more efficiently.
Methods: I analyzed secondary data from the KNRH database for neonatal admissions. It comprised of babies admitted after delivery with some life threatening conditions like prematurity, extreme low birth weight, fetal abnormalities, meconium aspiration, convulsions, respiratory distress syndrome, and breathing difficulties. The registered admissions (4,685 babies) in the KNRH Neonatal Intensive Care Unit (NICU) between January 2021 and December 2021, was explored and a sample of 1,406 babies among were analysed for length of stay and its associated factors that determined its distribution using count data analysis techniques. Several strategies were applied to select the best count fit model that included the Generalized Linear Models Poisson, Negative Binomial Regression, Zero-inflated Poisson, Zero-inflated Negative Binomial Regression model and Lasso model. AIC & BIC were used to determine the best fit model, which was also subjected to goodness of fit tests. A univariate, bivariate and multivariable analysis approaches were performed to determine relative counts (RC) of days of stay by levels of covariates and 95% confidence intervals (CI). A p-value ≤0.05 were considered to be significant.
Results: The mean LOS was 4.2 days and median LOS was 3 days. In the multivariable statistical analysis, factors associated with longer duration of stay included referred-in babies (RR=1.113, [CI=1.01 - 1.23, p-value =0.029]) ; prematurity (RR=1.202, [CI=1.07 - 1.38, p-value =0.002]); gestation age (RR=0.972 [CI=0.96 – 0.98, p-value <0.0001]), baby’s age (RR=1.032 [CI=1.01 – 1.06, p-value <0.046]), birth weight (RR=0.9999, [CI=0.9999 – 1.0, p-value =0.01]), and convulsions (RR=1.563, [CI=1.08 – 2.32, p-value =0.022]). The Negative Binomial regression model was the best fit to the data
Conclusion Overall, the results of the negative binomial was the best model choice, multivariable regression analysis suggested that: prematurity, gestation age, baby’s age, birth weight convulsion, and referral-in were statistically significant predictors of a newborn baby staying in the hospital.
Recommendation: I strongly recommend the use of count data analysis models especially the Negative Binomial regression model to model the factors that can predict the length of stay of neonates in NICU. I also recommend to clinical teams to use these results as benchmarks to improve care for neonates in NICU. | en_US |
