Scholarlink

<p>Regularly updating repositories of air pollution impacts on health is essential for evidence-based policies, interventions, and progress monitoring. However, this is often time-consuming and labour-intensive. Automation can ensure up-to-date evidence; reduce retrieval, screening time, and costs; and make information accessible to non-academic stakeholders. This study investigated whether a machine learning approach can perform any stages of a traditional scoping review on the health impacts, policies, and interventions related to air pollution due to domestic waste burning. A traditional approach was conducted in parallel with machine learning methods enabling a comparison of the efficiency and quality of the partially automated approach against the manual review. Findings from the two approaches were compared and the final set of included articles were considered for (1) reported impacts of waste burning on health outcomes and (2) recommendations on solutions and interventions to prevent/reduce adverse effects on health from waste burning. We found a range of health impacts associated with waste burning, including low birth weight, hypertensive disorders of pregnancy, adverse respiratory outcomes like asthma and wheeze, cancer risk, and mortality. Few studies proposed solutions or evaluated the effectiveness of interventions. The ML approach showed a tendency towards false positives, which are preferable to false negatives (where relevant papers were excluded). Results showed that the model can conduct initial searches and decisions for the review. However, the articles included in the model should be screened manually for final acceptance. Therefore, we propose a hybrid approach be used until the automated model can be further refined.</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

Air pollution-related health impacts from domestic waste burning and associated interventions: the merits of a traditional versus machine learning scoping review methodology

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月18日

查看全文
  • Thandi Kapwata,
  • Candice Webster,
  • Wen Guan Gavin Chen,
  • Muhammad Ibrahim Suder,
  • Devon Jarvis,
  • Alex Vogt,
  • Nomfundo Mahlangeni,
  • Tracey Laban,
  • Natasha Naidoo,
  • Pierpaolo Mudu,
  • Kerolyn Shairsingh,
  • Caradee Y. Wright

摘要

Regularly updating repositories of air pollution impacts on health is essential for evidence-based policies, interventions, and progress monitoring. However, this is often time-consuming and labour-intensive. Automation can ensure up-to-date evidence; reduce retrieval, screening time, and costs; and make information accessible to non-academic stakeholders. This study investigated whether a machine learning approach can perform any stages of a traditional scoping review on the health impacts, policies, and interventions related to air pollution due to domestic waste burning. A traditional approach was conducted in parallel with machine learning methods enabling a comparison of the efficiency and quality of the partially automated approach against the manual review. Findings from the two approaches were compared and the final set of included articles were considered for (1) reported impacts of waste burning on health outcomes and (2) recommendations on solutions and interventions to prevent/reduce adverse effects on health from waste burning. We found a range of health impacts associated with waste burning, including low birth weight, hypertensive disorders of pregnancy, adverse respiratory outcomes like asthma and wheeze, cancer risk, and mortality. Few studies proposed solutions or evaluated the effectiveness of interventions. The ML approach showed a tendency towards false positives, which are preferable to false negatives (where relevant papers were excluded). Results showed that the model can conduct initial searches and decisions for the review. However, the articles included in the model should be screened manually for final acceptance. Therefore, we propose a hybrid approach be used until the automated model can be further refined.