Incidence of neonatal mortality and its predictors among live births in Ethiopia: Gompertz gamma shared frailty model

Italian Journal of Pediatrics, Volume 46, No. 1, Year 2020

Interpretation

Background: Neonatal mortality remains a serious public health concern in developing countries including Ethiopia. Ethiopia is one of the countries with the highest neonatal mortality in Africa. However, there is limited evidence on the incidence and predictors of neonatal mortality at the national level. Therefore, this study aimed to investigate the incidence of neonatal mortality and its predictors among live births in Ethiopia. Investigating the incidence and predictors of neonatal mortality is essential to design targeted public health interventions to reduce neonatal mortality. Methods: A secondary data analysis was conducted based on the 2016 Ethiopian Demographic and Health Survey (EDHS) data. A total weighted sample of 11,022 live births was included in the analysis. The shared frailty model was applied since the EDHS data has hierarchical nature, and neonates are nested within-cluster, and this could violate the independent and equal variance assumption. For checking the proportional hazard assumption, Schoenfeld residual test was applied. Akakie Information Criteria (AIC), Cox-Snell residual test, and deviance were used for checking model adequacy and for model comparison. Gompertz gamma shared frailty model was the best-fitted model for this data since it had the lowest deviance, AIC value, and the Cox-Snell residual graph closet to the bisector. Variables with a p-value of less than 0.2 were considered for the multivariable Gompertz gamma shared frailty model. In the multivariable Gompertez gamma shared frailty model, the Adjusted Hazard Ratio (AHR) with a 95% confidence interval (CI) was reported to identify significant predictors of neonatal mortality. Results: Overall, the neonatal mortality rate in Ethiopia was 29.1 (95% CI: 26.1, 32.4) per 1000 live births. In the multivariable Gompertz gamma shared frailty model; male sex (AHR = 1.92, 95% CI: 1.52, 2.43), twin birth (AHR = 5.22, 95% CI: 3.62, 7.53), preceding birth interval less than 18 months (AHR = 2.07, 95% CI: 1.51, 2.85), small size at birth (AHR = 1.64, 95% CI: 1.24, 2.16), large size at birth (AHR = 1.53, 95% CI: 1.16, 2.01) and did not have Antenatal Care (ANC) visit (AHR = 2.10, 95% CI: 1.44, 3.06) were the significant predictors of neonatal mortality. Conclusion: Our study found that neonatal mortality remains a public health problem in Ethiopia. Shorter birth interval, small and large size at birth, ANC visits, male sex, and twin births were significant predictors of neonatal mortality. These results suggest that public health programs that increase antenatal care service utilization should be designed to reduce neonatal mortality and special attention should be given for twin births, large and low birth weight babies. Besides, providing family planning services for mothers to increase birth intervals and improving accessibility and utilization of maternal health care services such as ANC is crucial to improve neonatal survival. A secondary data analysis was conducted based on the EDHS 2016 data. The 2016 EDHS survey was the fourth survey conducted in Ethiopia, situated in the Horn of Africa. Ethiopia has 9 regional states (Afar, Amhara, Benishangul-Gumuz, Gambela, Harari, Oromia, Somali, Southern Nations, Nationalities, and People’s Region (SNNPR) and Tigray) and two Administrative Cities (Addis Ababa and Dire-Dawa). The EDHS used a stratified two-stage cluster sampling technique selected in two stages using the 2007 Population and Housing Census (PHC) as a sampling frame. Stratification was achieved by separating each region into urban and rural areas. In total, 21 sampling strata have been created. In the first stage, 645 Enumeration Areas (EAs) (202 in the urban area) were selected with probability selection proportional to the EA size and independent selection in each sampling stratum. In the second stage, on average, 28 households were systematically selected. A total weighted sample of 11,022 live births within 5 years preceding the survey were included. The detailed sampling procedure was presented in the full EDHS 2016 report [32]. The outcome variable for this study was neonatal survival status categorized as being alive (coded as 0) or died (coded as 1). Neonatal mortality is defined as the death of live birth within 28 days of life. Age at death was recorded in days if the child died within 28 days of delivery. The independent variables considered for this study were categorized as socio-demographic and economic variables (residence, region, religion, maternal education, husband education, maternal occupation, sex of household head, distance to the health facility and wealth status), child-related factors (sex of neonate, type of birth, preceding birth interval and birth size), and maternal healthcare services related factors (ANC visit, place of delivery, mode of delivery, parity, birth order, wanted pregnancy, and health insurance coverage). The data were weighted using sampling weight, primary sampling unit, and strata before any statistical analysis to restore the representativeness of the survey and take into account the sampling design to get reliable statistical estimates. The sampling statisticians determine how many samples are needed in each stratum to get reliable estimates, in EDHS, some regions were oversampled, and some regions were under-sampled. So, to get statistics that are representative of Ethiopia, the distribution of neonate in the sample need to be weighted (mathematically adjusted) such that it resembles the true distribution in Ethiopia by using sampling weight (v005), primary sampling unit (v021) and strata (v022). The descriptive and summary statistics were conducted using STATA version 14 software. The EDHS data has a hierarchical structure, and therefore neonates are nested within a cluster/EAs. This violates the traditional regression model assumption, which is the independence of observations and equal variance across clusters. We have checked whether there was clustering or not by running the frailty model (random effect survival model). EA was used as a random effect (clustering variable). The theta parameter was used to assess whether there is significant clustering or not. The EDHS data were collected at two-level at individual and at the community level. Therefore, neonates in the same cluster are more of share similar characteristics than neonates in another cluster. The theta (frailty parameter) was significant at the null model (θ = 0.45, 95% CI: 0.22, 0.83). It showed that there was unobserved heterogeneity or shared frailty that needs to be taken into account to get a reliable estimate. Schoenfeld residual test was applied to check the Proportional Hazard (PH) assumptions, and the PH assumption was fulfilled (p-value> 0.05). For model selection, log-likelihood ratio test, deviance (−2LL), Akaike Information Criteria (AIC), and Cox-Snell residual plot were applied. Cox-Snell Residual test is the difference between an observed data point and a predicted or fitted value. A Cox-Snell residual considers the distribution and estimated parameters from the lifetime regression model. A model with the highest values of log-likelihood and the lowest value of AIC was the best-fitted model. Nested parametric models in generalized gamma (Exponential, Weibull, lognormal) were compared based on deviance, and non-nested models (Gompertz and log-logistic) were compared using AIC. Deviance, AIC, and Cox-Snell residual graph showed that the Gompertz gamma shared frailty model had the lowest value and the closet graph to the bisector. Therefore, the Gompertz gamma shared frailty model was the best-fitted model for the data. Variable with a p-value less 0.20 in the uni-variable gamma shared frailty analysis were included in the multivariable analysis. We estimate the hazard ratio and 95% confidence interval. In the multivariable analysis, the Adjusted Hazard Ratio (AHR) with 95% Confidence Interval (CI) was used to declare significant predictors of neonatal mortality. Permission for data access was obtained from major demographic and health survey through an online request from http://www.dhsprogram.com. The data used for this study were publicly available with no personal identifier.

AI Digest

Study Justification:

– Neonatal mortality is a significant public health concern in developing countries, including Ethiopia.
– Ethiopia has one of the highest neonatal mortality rates in Africa.
– Limited evidence exists on the incidence and predictors of neonatal mortality at the national level in Ethiopia.
– Investigating the incidence and predictors of neonatal mortality is crucial for designing targeted public health interventions to reduce neonatal mortality.

Study Highlights:

– A secondary data analysis was conducted using the 2016 Ethiopian Demographic and Health Survey (EDHS) data.
– The study included a total weighted sample of 11,022 live births.
– The shared frailty model was applied to account for the hierarchical nature of the data and the clustering of neonates within clusters.
– The Gompertz gamma shared frailty model was found to be the best-fitted model for the data.
– Significant predictors of neonatal mortality were identified, including male sex, twin birth, short birth interval, small and large size at birth, and lack of antenatal care (ANC) visits.

Study Recommendations:

– Public health programs should focus on increasing antenatal care service utilization to reduce neonatal mortality.
– Special attention should be given to twin births and babies with large or low birth weight.
– Family planning services should be provided to mothers to increase birth intervals.
– Accessibility and utilization of maternal health care services, such as ANC, should be improved to improve neonatal survival.

Key Role Players:

– Ministry of Health: Responsible for implementing and coordinating public health programs.
– Health care providers: Involved in delivering antenatal care services and providing support for neonatal care.
– Community health workers: Play a crucial role in promoting antenatal care utilization and educating mothers on neonatal care.
– Non-governmental organizations (NGOs): Can provide support and resources for implementing targeted interventions to reduce neonatal mortality.

Cost Items for Planning Recommendations:

– Training and capacity building for health care providers and community health workers.
– Development and implementation of awareness campaigns to promote antenatal care utilization.
– Provision of family planning services and contraceptives.
– Improvement of infrastructure and accessibility to health care facilities.
– Monitoring and evaluation of the implemented interventions to assess their effectiveness.
Please note that the cost items provided are general suggestions and may vary depending on the specific context and resources available in Ethiopia.

Strength of Evidence

The strength of evidence for this abstract is 7 out of 10.
The evidence in the abstract is rated 7 because it is based on a secondary data analysis using the 2016 Ethiopian Demographic and Health Survey (EDHS) data. The study used a shared frailty model to account for the hierarchical nature of the data. The model selection process and statistical tests were conducted to ensure the validity of the results. However, the abstract does not provide information on the representativeness of the sample and potential limitations of the study. To improve the evidence, the abstract could include details on the sampling procedure, sample representativeness, and potential limitations such as selection bias or missing data.

Abstract

A secondary data analysis was conducted based on the EDHS 2016 data. The 2016 EDHS survey was the fourth survey conducted in Ethiopia, situated in the Horn of Africa. Ethiopia has 9 regional states (Afar, Amhara, Benishangul-Gumuz, Gambela, Harari, Oromia, Somali, Southern Nations, Nationalities, and People’s Region (SNNPR) and Tigray) and two Administrative Cities (Addis Ababa and Dire-Dawa). The EDHS used a stratified two-stage cluster sampling technique selected in two stages using the 2007 Population and Housing Census (PHC) as a sampling frame. Stratification was achieved by separating each region into urban and rural areas. In total, 21 sampling strata have been created. In the first stage, 645 Enumeration Areas (EAs) (202 in the urban area) were selected with probability selection proportional to the EA size and independent selection in each sampling stratum. In the second stage, on average, 28 households were systematically selected. A total weighted sample of 11,022 live births within 5 years preceding the survey were included. The detailed sampling procedure was presented in the full EDHS 2016 report [32]. The outcome variable for this study was neonatal survival status categorized as being alive (coded as 0) or died (coded as 1). Neonatal mortality is defined as the death of live birth within 28 days of life. Age at death was recorded in days if the child died within 28 days of delivery. The independent variables considered for this study were categorized as socio-demographic and economic variables (residence, region, religion, maternal education, husband education, maternal occupation, sex of household head, distance to the health facility and wealth status), child-related factors (sex of neonate, type of birth, preceding birth interval and birth size), and maternal healthcare services related factors (ANC visit, place of delivery, mode of delivery, parity, birth order, wanted pregnancy, and health insurance coverage). The data were weighted using sampling weight, primary sampling unit, and strata before any statistical analysis to restore the representativeness of the survey and take into account the sampling design to get reliable statistical estimates. The sampling statisticians determine how many samples are needed in each stratum to get reliable estimates, in EDHS, some regions were oversampled, and some regions were under-sampled. So, to get statistics that are representative of Ethiopia, the distribution of neonate in the sample need to be weighted (mathematically adjusted) such that it resembles the true distribution in Ethiopia by using sampling weight (v005), primary sampling unit (v021) and strata (v022). The descriptive and summary statistics were conducted using STATA version 14 software. The EDHS data has a hierarchical structure, and therefore neonates are nested within a cluster/EAs. This violates the traditional regression model assumption, which is the independence of observations and equal variance across clusters. We have checked whether there was clustering or not by running the frailty model (random effect survival model). EA was used as a random effect (clustering variable). The theta parameter was used to assess whether there is significant clustering or not. The EDHS data were collected at two-level at individual and at the community level. Therefore, neonates in the same cluster are more of share similar characteristics than neonates in another cluster. The theta (frailty parameter) was significant at the null model (θ = 0.45, 95% CI: 0.22, 0.83). It showed that there was unobserved heterogeneity or shared frailty that needs to be taken into account to get a reliable estimate. Schoenfeld residual test was applied to check the Proportional Hazard (PH) assumptions, and the PH assumption was fulfilled (p-value> 0.05). For model selection, log-likelihood ratio test, deviance (−2LL), Akaike Information Criteria (AIC), and Cox-Snell residual plot were applied. Cox-Snell Residual test is the difference between an observed data point and a predicted or fitted value. A Cox-Snell residual considers the distribution and estimated parameters from the lifetime regression model. A model with the highest values of log-likelihood and the lowest value of AIC was the best-fitted model. Nested parametric models in generalized gamma (Exponential, Weibull, lognormal) were compared based on deviance, and non-nested models (Gompertz and log-logistic) were compared using AIC. Deviance, AIC, and Cox-Snell residual graph showed that the Gompertz gamma shared frailty model had the lowest value and the closet graph to the bisector. Therefore, the Gompertz gamma shared frailty model was the best-fitted model for the data. Variable with a p-value less 0.20 in the uni-variable gamma shared frailty analysis were included in the multivariable analysis. We estimate the hazard ratio and 95% confidence interval. In the multivariable analysis, the Adjusted Hazard Ratio (AHR) with 95% Confidence Interval (CI) was used to declare significant predictors of neonatal mortality. Permission for data access was obtained from major demographic and health survey through an online request from http://www.dhsprogram.com. The data used for this study were publicly available with no personal identifier.

Materials

N/A

Innovations

Based on the provided information, it is difficult to identify specific innovations for improving access to maternal health. The information provided is focused on the methodology and findings of a study on neonatal mortality in Ethiopia, rather than innovations or recommendations for improving maternal health. To provide recommendations for improving access to maternal health, it would be helpful to have information on the current challenges and context of maternal health in Ethiopia.

AI Innovations Description

The recommendation to improve access to maternal health based on the study findings is to design targeted public health interventions that focus on increasing antenatal care (ANC) service utilization. The study identified that not having ANC visits was a significant predictor of neonatal mortality. Therefore, efforts should be made to improve accessibility and utilization of ANC services in Ethiopia. This can be achieved through various strategies such as increasing the number of ANC clinics, ensuring availability of skilled healthcare providers, and promoting awareness and education about the importance of ANC among pregnant women and their families.

Additionally, the study found that shorter birth intervals, small and large size at birth, male sex, and twin births were also significant predictors of neonatal mortality. To address these factors, it is important to provide family planning services to mothers to increase birth intervals and promote healthy spacing between pregnancies. Special attention should also be given to twin births, as they were found to have a higher risk of neonatal mortality. This can include providing specialized care and support for mothers carrying twins, as well as ensuring access to appropriate medical interventions during pregnancy and childbirth.

Overall, the recommendation is to focus on improving access to and utilization of maternal healthcare services, particularly ANC, while also addressing specific risk factors such as birth intervals and twin births. By implementing these recommendations, it is expected that maternal health outcomes, including neonatal mortality, can be improved in Ethiopia.

AI Innovations Methodology

Based on the provided information, here are some potential recommendations to improve access to maternal health:

1. Increase Antenatal Care (ANC) utilization: Implement programs and initiatives to encourage pregnant women to attend ANC visits regularly. This can be done through community outreach, education campaigns, and incentives for attending ANC appointments.

2. Improve accessibility of maternal health care services: Ensure that maternal health care services, including prenatal care, delivery services, and postnatal care, are easily accessible to all women, especially those in rural and remote areas. This can be achieved by establishing more health facilities, mobile clinics, and transportation services.

3. Enhance maternal health education: Provide comprehensive education and information to women and their families about the importance of maternal health care, including the benefits of ANC visits, safe delivery practices, and postnatal care. This can be done through community health workers, health education programs, and mass media campaigns.

4. Strengthen health systems: Invest in improving the overall health system infrastructure, including training and capacity building for health care providers, ensuring the availability of essential medicines and supplies, and implementing quality assurance mechanisms to ensure safe and effective maternal health care.

To simulate the impact of these recommendations on improving access to maternal health, a methodology could include the following steps:

1. Define the indicators: Identify specific indicators that measure access to maternal health, such as the percentage of women attending ANC visits, the percentage of women delivering in health facilities, and the neonatal mortality rate.

2. Collect baseline data: Gather data on the current status of access to maternal health care services, including the indicators identified in step 1. This can be done through surveys, interviews, or analysis of existing data sources.

3. Develop a simulation model: Create a mathematical or statistical model that simulates the impact of the recommendations on the identified indicators. This model should take into account various factors, such as population demographics, geographic distribution, and existing health infrastructure.

4. Input the recommended interventions: Incorporate the recommended interventions into the simulation model. Assign specific values or parameters to each intervention based on available evidence or expert opinion.

5. Run the simulation: Execute the simulation model using the baseline data and the inputted interventions. This will generate simulated data on the expected changes in the indicators of access to maternal health care.

6. Analyze the results: Analyze the simulated data to assess the impact of the recommended interventions on improving access to maternal health. Compare the simulated indicators with the baseline data to determine the magnitude of change.

7. Validate the results: Validate the simulation results by comparing them with real-world data, if available. This will help assess the accuracy and reliability of the simulation model.

8. Refine and iterate: Based on the results and validation, refine the simulation model and interventions if necessary. Repeat the simulation process to further explore different scenarios or variations of the interventions.

By following this methodology, policymakers and stakeholders can gain insights into the potential impact of the recommended interventions on improving access to maternal health care and make informed decisions on implementing them.

Author

Dima Health

Statistics:

Citations: 6

Identifiers:

DOI: 10.1186/s13052-020-00893-6

ISSN: 17208424

Research Areas:

Community Interventions, Disparities, Environmental, Genetics and Genomics, Health System and Policy, Maternal Access, Maternal and Child Health, Quality of Care, Sexual and Reproductive Health, Social Determinants

Study Countries:

Ethiopia

Study Design:

Cohort Study, Cross Sectional Study

Study Approach:

Quantitative

Participants Gender:

Male