Scholarlink

Background <p>Malaria transmission in West Africa is governed by intricate interactions among climatic variability, temporal epidemiologic memory, and health-system determinants. Nevertheless, the majority of investigations model incidence in isolation and presuppose linear climatic effects, thereby leaving unresolved the manner in which climate influences the entire cascade from incidence to severity and mortality. The present study employs a hybrid structural equation modelling (SEM) approach together with SHAP-enhanced machine-learning methods to quantify phase-specific climatic and endogenous drivers of malaria outcomes in Tarkwa-Nsuaem, Ghana (2013–2023).</p> Methods <p>Monthly DHIMS-2 malaria surveillance data were linked with gridded rainfall and temperature datasets obtained from the Global Climate Monitor. A dynamic structural equation model (SEM) was employed to estimate directional pathways for incidence and severity, incorporating lagged climatic and epidemiologic predictors. Mortality, characterized by sparse counts, was analyzed using negative binomial, Poisson, and logistic regression approaches. Random forest models with SHAP interpretation were applied to quantify non-linear contributory effects.</p> Results <p>Incidence was substantially influenced by lagged rainfall (β = 0.436, p = 0.001) and nonlinear rainfall saturation (β = − 0.346, p = 0.006), in conjunction with pronounced epidemiological memory, as indicated by incidence<sub>t–1</sub> (β = 0.450, p &lt; 0.001). Severity was predominantly endogenous, with current incidence suppressing disease progression (odds ratio [OR] = 0.108; 95% CI: 0.068–0.168), whereas lagged severity markedly elevated the likelihood of severe outcomes (OR = 2.304; 95% CI: 1.972–2.717). Mortality, although infrequent, exhibited modest associations with temperature<sub>t–1</sub> (IRR = 1.774; p = 0.034) and severe<sub>t–1</sub> (IRR = 1.015; p = 0.023).</p> Conclusion <p>Precipitation effects concentrate at the incidence phase, whereas severity is consistent with intrinsic clinical dynamics, and mortality is consistent with delayed progression to severe disease compounded by thermal stress. Integrating rainfall-triggered early warning with severity-focused clinical surge preparedness could substantially reduce under-five mortality from malaria.</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

Lagged climatic drivers of malaria incidence severity and mortality using SEM and SHAP hybrid modelling in Ghana

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月1日

查看全文
  • Senyefia Bosson-Amedenu,
  • Abdulzeid Yen Anafo

摘要

Background

Malaria transmission in West Africa is governed by intricate interactions among climatic variability, temporal epidemiologic memory, and health-system determinants. Nevertheless, the majority of investigations model incidence in isolation and presuppose linear climatic effects, thereby leaving unresolved the manner in which climate influences the entire cascade from incidence to severity and mortality. The present study employs a hybrid structural equation modelling (SEM) approach together with SHAP-enhanced machine-learning methods to quantify phase-specific climatic and endogenous drivers of malaria outcomes in Tarkwa-Nsuaem, Ghana (2013–2023).

Methods

Monthly DHIMS-2 malaria surveillance data were linked with gridded rainfall and temperature datasets obtained from the Global Climate Monitor. A dynamic structural equation model (SEM) was employed to estimate directional pathways for incidence and severity, incorporating lagged climatic and epidemiologic predictors. Mortality, characterized by sparse counts, was analyzed using negative binomial, Poisson, and logistic regression approaches. Random forest models with SHAP interpretation were applied to quantify non-linear contributory effects.

Results

Incidence was substantially influenced by lagged rainfall (β = 0.436, p = 0.001) and nonlinear rainfall saturation (β = − 0.346, p = 0.006), in conjunction with pronounced epidemiological memory, as indicated by incidencet–1 (β = 0.450, p < 0.001). Severity was predominantly endogenous, with current incidence suppressing disease progression (odds ratio [OR] = 0.108; 95% CI: 0.068–0.168), whereas lagged severity markedly elevated the likelihood of severe outcomes (OR = 2.304; 95% CI: 1.972–2.717). Mortality, although infrequent, exhibited modest associations with temperaturet–1 (IRR = 1.774; p = 0.034) and severet–1 (IRR = 1.015; p = 0.023).

Conclusion

Precipitation effects concentrate at the incidence phase, whereas severity is consistent with intrinsic clinical dynamics, and mortality is consistent with delayed progression to severe disease compounded by thermal stress. Integrating rainfall-triggered early warning with severity-focused clinical surge preparedness could substantially reduce under-five mortality from malaria.