Scholarlink

Background <p>An infant’s birth weight, typically measured within the first hours after birth, is crucial for assessing their health. Low birth weight (LBW) can result from intrauterine growth restriction, preterm birth, or a combination of both factors. This study employs Bayesian propensity score (BPS) methods to estimate the causal effect of midwife-led continuity care (MLCC) on LBW in Ethiopia’s North Shoa Zone, Amhara Regional State. Using quasi-experimental data, we address covariate imbalance between MLCC and other professional groups, with simulations validating the robustness of our findings.</p> Methods <p>A prospective non-randomized (quasi-experimental study design) was employed from August 2019 to September 2020 in the North Shoa Zone, Amhara Regional State, Ethiopia. Markov Chain Monte Carlo algorithms were employed in Bayesian causal inference approaches to estimate the average treatment effect. A simulation study was conducted to evaluate the performance of the standard nearest-neighbor Propensity score and the weighting BPS methods for estimating the average treatment effect (ATE) on LBW.</p> Results <p>Our analysis showed that the MLCC reduced the risk of LBW by 24% (ATE: −2.39, 95% CI: −12.4 to 7.63). Bayesian methods estimated ATE more precisely (2.065, SE = 0.875) than Frequentist ones (2.156, SE = 0.86), due to better noise handling and use of more data. Although bias differences weren’t significant, Bayesian estimates proved more reliable. Key LBW predictors include uterine height, emergency cesarean, and nutrition. These findings support MLCC as a valuable strategy for improving neonatal outcomes in low-resource settings.</p> Conclusion <p>The counterfactual estimate of the Bayesian method indicates that the Bayesian credible interval effect of the LBW of newborn babies reliably estimates the true causal relationship. The sensitivity analysis demonstrated the robustness of our findings, with minimal variation in adjusted odds ratio and ATE across alternative priors and matching methods, and showed the reliability of the BPS approach in producing consistent and credible causal inferences. More attention should be given to pregnant women with an abortion history, with high blood pressure, with more folic acid, with an emergency cesarean, women with premature babies, and women with low nutritional status.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Bayesian propensity score approaches for balancing covariates associated with low birth weight in North Shoa Zone, Ethiopia

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月17日

查看全文
  • Wudneh Ketema Moges,
  • Awoke Seyoum Tegegne,
  • Aweke A. Mitku,
  • Esubalew Tesfahun,
  • Solomon Hailemeskel

摘要

Background

An infant’s birth weight, typically measured within the first hours after birth, is crucial for assessing their health. Low birth weight (LBW) can result from intrauterine growth restriction, preterm birth, or a combination of both factors. This study employs Bayesian propensity score (BPS) methods to estimate the causal effect of midwife-led continuity care (MLCC) on LBW in Ethiopia’s North Shoa Zone, Amhara Regional State. Using quasi-experimental data, we address covariate imbalance between MLCC and other professional groups, with simulations validating the robustness of our findings.

Methods

A prospective non-randomized (quasi-experimental study design) was employed from August 2019 to September 2020 in the North Shoa Zone, Amhara Regional State, Ethiopia. Markov Chain Monte Carlo algorithms were employed in Bayesian causal inference approaches to estimate the average treatment effect. A simulation study was conducted to evaluate the performance of the standard nearest-neighbor Propensity score and the weighting BPS methods for estimating the average treatment effect (ATE) on LBW.

Results

Our analysis showed that the MLCC reduced the risk of LBW by 24% (ATE: −2.39, 95% CI: −12.4 to 7.63). Bayesian methods estimated ATE more precisely (2.065, SE = 0.875) than Frequentist ones (2.156, SE = 0.86), due to better noise handling and use of more data. Although bias differences weren’t significant, Bayesian estimates proved more reliable. Key LBW predictors include uterine height, emergency cesarean, and nutrition. These findings support MLCC as a valuable strategy for improving neonatal outcomes in low-resource settings.

Conclusion

The counterfactual estimate of the Bayesian method indicates that the Bayesian credible interval effect of the LBW of newborn babies reliably estimates the true causal relationship. The sensitivity analysis demonstrated the robustness of our findings, with minimal variation in adjusted odds ratio and ATE across alternative priors and matching methods, and showed the reliability of the BPS approach in producing consistent and credible causal inferences. More attention should be given to pregnant women with an abortion history, with high blood pressure, with more folic acid, with an emergency cesarean, women with premature babies, and women with low nutritional status.