文献速递技术与安全CTMRUSPET

人工智能利用自由文本检查申请录入促进联邦要求的临床决策支持:回顾性评估

Retrospective Evaluation of Artificial Intelligence Leveraging Free-Text Imaging Order Entry to Facilitate Federally Required Clinical Decision Support.
PubMed 文献 · J Am Coll Radiol · 2021-10-01

本研究评估了一种商用人工智能工具,该工具基于开单医生输入的自由文本检查原因预测结构化适应症,以促进临床决策支持。回顾性分析实施后前3个月的40,053份高级影像检查申请,结果显示开单医生更倾向于选择自由文本-AI方法而非直接搜索结构化适应症,AI工具在91.7%的自由文本申请中给出了预测适应症,医生在57.7%的提供预测适应症的病例中选择了AI预测的适应症。

要点

  • 开单医生显著更倾向于使用自由文本-AI方法进行临床决策支持(58.9% vs 41.1%,P < .001)。
  • AI工具在91.7%的自由文本申请中成功预测适应症,医生最终采纳了57.7%的AI预测适应症。
  • 该工具可能通过减少结构化适应症搜索负担来促进临床决策支持的依从性。

局限性

  • 单中心回顾性研究,结果外推需谨慎。
  • 未报告AI预测适应症的准确性(如与人工编码金标准对比)。
  • 研究仅覆盖实施后前3个月,长期效果未知。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

英文原文(摘要)· 点击展开

OBJECTIVE: The Protecting Access to Medicare Act mandates clinical decision support (CDS) at imaging order entry, necessitating the use of structured indications to map CDS scores. We evaluated the performance of a commercially available artificial intelligence (AI) tool leveraging free-text order entry to facilitate provider selection of the necessary structured indications.

METHODS: Our institution implemented an AI tool offering predicted structured indications based upon the ordering provider's entry of a free-text reason for examination. Providers remained able to order via the traditional direct search for structured indications. Alternatively, they could take the new free-text-AI approach allowing them to select from AI-predicted indications, perform additional direct searches, indicate no matching indication, or exit CDS workflow. We hypothesized the free-text-AI approach would be elected more often and the AI tool would be successful in facilitating selection of structured indications. We reviewed advanced imaging orders (n = 40,053) for the first 3 months (February to May 2020) since implementation.

RESULTS: Providers were more likely (P < .001) to choose the free-text-AI approach (23,580; 58.9%) to order entry over direct search for structured indications (16,473; 41.1%). The AI tool yielded alerts with predicted indications in 91.7% (n = 21,631) of orders with free text. Ultimately, providers chose AI-predicted indications in 57.7% (n = 12,490) of cases in which they were offered by the tool.

DISCUSSION: Providers significantly more often elected the new free-text-AI approach to order entry for CDS, suggesting provider preference over the traditional approach. The AI tool commonly predicted indications acceptable to ordering providers.

来源:J Am Coll Radiol,查看原文。版权归原作者所有。

引用

  1. Gish DS, Ellenbogen AL, Patrie JT, et al. Retrospective Evaluation of Artificial Intelligence Leveraging Free-Text Imaging Order Entry to Facilitate Federally Required Clinical Decision Support. J Am Coll Radiol. 2021. PMID: 34600896. doi:10.1016/j.jacr.2021.08.021

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。