ORIGINAL ARTICLES原创文章

Does AI-Powered Clinical Documentation Enhance Clinician Efficiency? A Longitudinal Study

T.-L. Liu and Others

Abstract

BACKGROUND

Nuance’s Dragon Ambient eXperience (DAX) Copilot is an artificial intelligence (AI)–driven ambient clinical documentation software platform. Atrium Health, a large multisite academic learning health system, was the first to use DAX Copilot. This study evaluates outcomes for participating clinicians after DAX implementation.

METHODS

In this longitudinal study, 112 primary care clinicians using DAX were recruited between June and August 2023 along with a control group of 103 clinicians from similar practices not using DAX. Primary outcomes of electronic health record (EHR) use and financial impact were assessed over 180 days using linear mixed models. Within the DAX group were two subgroups: active users (who transferred ≥25% of DAX notes) and high users (who transferred ≥60% of DAX notes). We performed exploratory analyses to compare the control group with DAX subgroups, in addition to subgroup analyses stratified by patient volume and clinician specialty.

RESULTS

After controlling for length of intervention, age, gender, provider type, years of practice, and baseline outcome, we did not find statistical significance in the primary analyses of EHR and financial metrics. Exploratory analyses suggested that small decreases in documentation hours could result from high DAX usage (means ratio [MR] 0.93, 95% confidence interval [CI] 0.88 to 0.98) and from implementing DAX with low-volume clinicians (MR 0.91, 95% CI 0.83 to 0.99) and with family medicine clinicians (MR 0.91, 95% CI 0.85 to 0.98).

CONCLUSIONS

AI-powered ambient clinical documentation software has been promoted as a promising strategy to alleviate the documentation burden faced by outpatient clinicians. However, our findings suggest that the tool did not make clinicians as a group more efficient. Future studies can further investigate the utility of DAX for clinician subgroups and alternative implementations with improved clinical adoption. (Funded by Wake Forest University Health Sciences; ClinicalTrials.gov number, NCT06329427.)

DOI: 10.1056/AIoa2400659

全文链接：https://ai.nejm.org/doi/abs/10.1056/AIoa2400659

人工智能驱动的临床文档是否提高了临床医生的效率？一项纵向研究

T.-L. Liu 等人

摘要： 背景： Nuance的Dragon Ambient eXperience (DAX) Copilot是一个由人工智能（AI）驱动的临床文档软件平台。Atrium Health，一个大型多站点学术学习健康系统，是第一个使用DAX Copilot的机构。这项研究评估了DAX实施后参与临床医生的结果。 方法： 在这项纵向研究中，112名使用DAX的初级保健临床医生在2023年6月至8月之间被招募，以及一个由103名不使用DAX的类似实践的临床医生组成的对照组。主要结果电子健康记录（EHR）使用和财务影响在180天内使用线性混合模型进行评估。在DAX组内有两个子组：活跃用户（转移了≥25%的DAX笔记）和高用户（转移了≥60%的DAX笔记）。我们进行了探索性分析，比较对照组与DAX子组，以及按患者量和临床医生专业分层的子组分析。 结果： 在控制干预长度、年龄、性别、提供者类型、实践年数和基线结果后，我们没有发现EHR和财务指标的主要分析具有统计学意义。探索性分析表明，高DAX使用可能导致文档小时数小幅减少（均值比[MR] 0.93，95%置信区间[CI] 0.88至0.98），并且在实施DAX的低量临床医生（MR 0.91，95% CI 0.83至0.99）和家庭医学临床医生（MR 0.91，95% CI 0.85至0.98）中也是如此。 结论： 人工智能驱动的临床文档软件被推广为减轻门诊临床医生面临的文档负担的有希望的策略。然而，我们的发现表明，该工具并没有使临床医生作为一个群体更高效。未来的研究可以进一步调查DAX对临床医生子组的效用以及具有改进临床采用的替代实施。（由Wake Forest大学健康科学资助；ClinicalTrials.gov编号，NCT06329427。）

NEJM AI, Volume 1 No. 12 December 2024

译文来自于AI工具Kimi