^{1}

^{2}

^{3}

Newborn with Sepsis condition has become a major cause of deaths in newborn babies in the world especially in developing countries. This study considered a total of 1019 cases diagnosed with sepsis at the Tamale Teaching Hospital covering the period January 2010 to June 2015. The data was modeled using the Binary logistic regression model to ascertain the variables that contributed to mortality of newborn with sepsis condition at the Tamale Teaching Hospital. Preliminary analysis revealed that pregnant women who mostly failed to attend the antenatal are most affected. Further analysis shows cost of treatment and location of patients as significant contributing variables that contributed to mortality of newborn with sepsis condition at the facility. As a result, it was recommended that facilities must be improved at the districts health centers to cut traveling time of pregnant women to the bigger health centers as well as checking the rising cost of treatment so as to minimize or eradicate newborn with sepsis condition.

An acceptable definition of newborn sepsis is lacking according to reports on the expert meeting on neonatal and pediatric sepsis of EMA (2010). It can be defined by the presence of at least two clinical symptoms in the presence of suspected baby [

In April, 2018, a paper was written on the topic, “Predictors of Neonatal Deaths in Ashanti Region of Ghana” Neonatal mortality continues to be a public health problem, especially in sub-Saharan Africa. Their study was conducted to assess the maternal, neonatal, and health system related factors that influence neonatal deaths in the Ashanti Region, Ghana. Out of the 222 mothers, there were 115 (51.8%) whose babies did not survive. Majority, 53.9%, of babies died within 1 - 4 days, 31.3% within 5 - 14 days, and 14.8% within 15 - 28 days. The study found out that, asphyxia, low birth weight, congenital anomalies, infections, and respiratory distress syndrome. Neonatal deaths were influenced neonatal factors (birth weight, gestational period, sex of baby, and Apgar score), and health related factors (health staff attitude, supervision of delivery, and hours spent at labor ward). Their research concluded that, there was a high level of neonatal deaths in the Ashanti Region of Ghana. According to [

An article on the topic, “Risk Factors Associated with Neonatal Sepsis” according to the study, neonatal sepsis accounts for an estimated 26% of under-five deaths, with sub-Saharan Africa having the highest mortality rates. It remains a notable hindrance to the progress in the decline of cause-specific mortality rates especially in sub-Saharan Africa. This study aimed at examining the risk factors of neonatal sepsis at the Trauma and Specialist Hospital, Winneba. The study found both maternal and neonatal factors to have a strong association with the risk of developing neonatal sepsis [

According to the [

This comprises of the data management, which included the study facility, the region of study, the population of the area the coding of variables and how the data was extracted.

The data used in this study was a secondary data from Tamale Teaching Hospital. Tamale Teaching Hospital is located in the capital of Northern Region-Ghana. Ghana is located in West Africa. It has a total population of over twenty four million people (24,000,000) The sample size was one thousand and nineteen (1019). It covers the period from January 2010 to June 2015. This data was extracted from among other ailments from the District Health Information Management System (DHIMS). Tamale has a total population of 2,468,557 people (2010 population census) [

Both discriminant analysis and logistic regression can be used to determine the categorical probability of an event occurring given a selected number of continuous and categorical variables. However, discriminant analysis requires that the distribution of independent variables in the model follow a normal distribution [

A limitation of logistic regression is that it is sensitive to variables that have very high correlations with each other. Variables that are highly collinear often produce very large standard errors and inflated regression estimates (Tabachnick & Fidell, 2013). Therefore, the collinearity between the independent variables in the model had to be observed. A standard procedure that allows for this is the calculation of tolerance for each variable. The tolerance statistic is the calculation of the variance of each of the independent variables in the model not explained by all of the other independent variables in the model. A higher tolerance value suggests low levels of collinearity. Menard (2010) suggests that a tolerance of less than 0.2 is alarming. Although logistic regression software does not typically offer a tolerance function, Menard (2010) suggests that the model be calculated as linear regression to observe the relationship among independent variables [

1) Preliminary classification shows from the decision variable in

2) The mathematical concept of logistic regression is to express the relationship between outcome variable and predictor variables (independent variables) in terms of logit: the natural logarithm of odds. Considering a case where Y is a dichotomous outcome variable categorized as “ 1” and “ 0” and X is a continuous predictor variable. Logistic regression facilitates a situation by logit transforming on the outcome variable Y. The logistic regression model can be written as:

logit ( y ) = ln μ 1 − μ = β 0 + β 1 X (1)

here μ is the probability of occurring the outcome Y and ( μ 1 − μ ) is the odds of success; the ratio of the probability of occurring the outcome Y and the probability of not occurring the outcome Y. β0 and β1 are called intercept and slope (regression coefficient) respectively. By taking antilog on both sides of Equation (1) we can estimate the probability of the occurrence of outcome Y for a given value of predictor

X : μ = p ( Y X = x ) = e β 0 + β 1 X 1 + e β 0 + β 1 X (2)

We can extend the logistic model in Equation (2) for more than one predictor as in the case of this study

log i t ( Y ) = ln µ 1 − µ = β 0 + β 1 X 1 + β 2 X 2 + β 3 X 3 + β 4 X 4 + β 5 X 5 (3)

Observed | Predicted | ||
---|---|---|---|

Outcome of Treatment | Percentage Correct | ||

Discharge | Died | ||

Discharge | 0 | 111 | 0.0 |

Died | 0 | 906 | 100.0 |

Total | 89.1 |

Step | Chi-square | Df | Sig. |
---|---|---|---|

1 | 20.979 | 8 | 0.007 |

Sig. (p < 0.05), Ho: “The model is not adequate for the data”, H1: The model is adequate for the data”.

Equation (3) is the specific form of logistic regression model for five numbers of predictors. Regression parameter βs can be estimated by either maximum likelihood (ML) method or weighted least square method. The value of regression coefficients β1 … β5 indicate the relationship between X’s and logit of Y. Coefficient value bigger than 0 indicates an increase in logit of Y with an increase in X and coefficient value smaller than 0 indicates a decrease in logit of Y with an increase in X. When the coefficient value is 0, it indicates there is no linear relationship among logit of Y and predictors X.

3) The predictive variables which consist of dataset of gender (categorical variable), the cost in Ghana cedis (continuous variable), the location in kilometers (categorical variable), insurance (categorical variable) and duration (continuous variable) depending on the outcome of treatment of a dichotomous variable (Discharged and Died), the model can be fitted as:

Logit ( y = 1 ) = β 0 + β 1 Gender + β 2 Cost + β 3 Location + β 4 Insurance + β 5 Duration (4)

The null hypothesis of the overall model states that all regression coefficients (β0, β1, β2, β3, β4, β5) are zero. Rejection of this null hypothesis will imply that at least one regression coefficient is non-zero meaning the logistic regression Equation in (4) predicts from the results in

logit ( outcome ) = 0.82 + 0.05 gender + 0.02 cost + 0.5 location + 19.28 insurance + 0.01 duration (5)

All the predictors “gender”, “cost”, “location”, “insurance” and “duration” are positively related to the log of odds of the independent variable (outcome) as seen in

It can be observed that, with an odds ratio of 1.049, male newborns are more likely to die of sepsis as compared to their female counterparts. Again, the odds

Variables | B | S.E. | Wald | Df | Sig. | Exp(B) |
---|---|---|---|---|---|---|

Gender | 0.048 | 0.205 | 0.055 | 1 | 0.814 | 1.049 |

Cost | 0.018 | 0.004 | 15.761 | 1 | 0.000 | 1.018 |

Location | 0.548 | 0.205 | 7.142 | 1 | 0.008 | 1.730 |

Insurance | 19.277 | 11,274.163 | 0.000 | 1 | 0.999 | 235,473,798.287 |

Duration | 0.011 | 0.022 | 0.261 | 1 | 0.609 | 1.011 |

Constant | 0.816 | 0.307 | 7.060 | 1 | 0.008 | 2.261 |

Sig. (p < 0.05).

ratio for location, 1.730, indicates that newborn with sepsis who travels at least 10km to access health care have higher chance of being dead as compared to those who resides within less than 10 km radius of the Tamale Teaching hospital. Meanwhile, as cost of treatment increases there is likelihood of increases in death of newborns with sepsis.

It can generally be seen that, the cost of treatment and the location of the patient were the major contributing variables that, causes mortality among newborns with sepsis at the Tamale Teaching hospital within the period under study and therefore considered to be significant with 0.000 and 0.008 respectively. The other variables gender, insurance and duration of stay of patients contributed but their contributions according to the analysis were not significant since their significance values were greater than the p value of 0.05.

Neonatal sepsis is an acute infection on newborns which can be categorized into two: early-onset or late-onset. Early onset is mostly common since those infections are contracted through the birth canal of the mother during deliveries which mostly occur within 24 hours. Late-onset occurs after 24 hours contracted within the environment [

The Binary Logistic Regression model was utilized for this study. It was realized that the cost of treatment and the location of patients were statistically significant in contributing to the mortality in newborn with sepsis with their p-values less than 0.05. It can be explained that the value 0.816 is seen to be the log of odds for outcome of treatment (Discharged and Died) status when all the variables in question are zero. It was observed that male newborns are more likely to die of sepsis as compared to their female counterparts. Again, it was indicated that the newborn with sepsis who travels at least 10 km to access health care have higher chance of being dead as compared to those who stayed within less than 10 km radius of the Tamale Teaching Hospital. The location of patients when other variables are constant will have a unit increase in kilometers by 54.8% whiles keeping other variables constant, a unit of time in days will increase the outcome of treatment by 1.1%.

The data used required no permission from the patients since the identities of the patients were not in known and therefore not applicable here.

Not applicable.

This work was carried out in collaboration between all authors. Author ALA worked on the background, methodology and analysis of the data. Author MS worked on the interpretation and literature review. Author MY worked on discussions and the conclusion. All authors read and agreed on the final manuscript.

The authors declare no conflicts of interest regarding the publication of this paper.

Alhassan, A.L., Sulemana, M. and Yahaya, M. (2019) Determinants of Mortality in Newborn with Sepsis Condition Using Binary Logistic Model at the Tamale Teaching Hospital. Open Access Library Journal, 6: e5409. https://doi.org/10.4236/oalib.1105409