PO.SHP01.02 · 科学与健康政策
通过人工智能（AI）赋能的技术解决方案加强安全网医院中符合指南的诊疗服务
Strengthening guideline compliant care delivery in a safety net hospital through artificial intelligence (AI) enabled technology solutions
作者与单位 Authors & Affiliations
摘要 Abstract
中文摘要
背景：在资源受限的环境中，与基于指南的癌症诊疗的一致性往往较低，导致治疗不足或过度治疗以及诊疗延误。Navya Earthshot 是一个经过验证的 AI 赋能决策支持系统，可将患者病历与标准诊疗癌症治疗指南进行匹配，以生成循证推荐。本研究评估了 Navya Earthshot 在印度一家安全网医院的实施情况，以评估与国家癌症网络（NCG）指南一致性的改善及诊疗延误的减少。
方法：一项在印度公共卫生系统中的一家安全网癌症中心开展、历时六个月的前瞻性前后对照观察性研究。收集 Navya Earthshot 整合前的基线数据用于比较。关键结局包括治疗决策的一致性和每项治疗决策的平均住院就诊次数。Navya Earthshot 被整合到常规工作流程中。该系统将患者特异性信息与 NCG 指南匹配以生成治疗选项，由主治肿瘤科医生审核并在每日肿瘤委员会中讨论。通过取消非强制性检查来简化诊断检测，并通过数字化诊疗导航和由非专业癌症导航员进行的结构化随访加以强化。
结果：在研究期间，892 例患者到院进行首次癌症就诊。在需要主动治疗决策的 537 例患者中（290 例实施前；247 例实施后），411 例可评估（210 例实施前；201 例实施后）。指南一致性从基线的 67%（140/210；95% CI +/-6%）提高到 Navya Earthshot 整合后的 87%（175/201；95% CI +/-5%），绝对增幅为 20%。Navya Earthshot 的整合，结合数字化增强的诊疗导航，将治疗启动前患者的平均就诊次数从平均 3.45 次减少到 2.5 次（减少 0.95 次，降幅 27.5%），主要归因于取消了非强制性报告审核就诊。诊疗启动时间从基线的 6 周提前至干预后的 4 周（提前 2 周，降幅 33%）。
结论：Navya Earthshot 的实施改善了对标准诊疗指南的依从性，并缩短了资源受限环境中至治疗的时间。将经过验证的 AI 决策支持工具整合到肿瘤诊疗工作流程中，可加强癌症诊疗服务，并推动在资源受限的卫生系统中更广泛地采用基于指南的治疗。
查看英文原文 English abstract
Background: Concordance with guideline-based cancer care is often low in resource-constrained settings, leading to under- or over-treatment and delays in care. Navya Earthshot is a validated AI-enabled decision support system that matches patient records to standard-of-care cancer treatment guidelines to generate evidence-based recommendations. This study evaluated the implementation of Navya Earthshot at a safety-net hospital in India to assess improvement in concordance with National Cancer Grid (NCG) guidelines and reduction in delays to care.
Methods: A prospective pre-post observational study conducted over six months at a safety net cancer center in the public health system of India. Baseline data prior to Navya Earthshot integration were collected for comparison. Key outcomes included concordance of treatment decisions and average number of hospital visits per treatment decision. Navya Earthshot was integrated into routine workflows. The system matched patient-specific information to NCG guidelines to generate treatment options, which were reviewed by treating oncologists and discussed in daily tumor boards. Diagnostic testing was streamlined by eliminating non-mandatory investigations, reinforced through digital care navigation and structured follow-up by lay cancer navigators.
Results: During the study period, 892 patients presented to the hospital for a first cancer visit. Among 537 patients requiring active treatment decisions (290 pre-implementation; 247 post-implementation), 411 were evaluable (210 pre; 201 post). Guideline concordance improved from 67% (140/210; 95% CI +/-6%) at baseline to 87% (175/201; 95% CI +/-5%) after Navya Earthshot integration-a 20% absolute increase. Integration of Navya Earthshot, combined with digitally enhanced care navigation, reduced the average number of patient visits before treatment initiation of care from an average 3.45 to 2.5 visits (0.95 visits, 27.5% reduction), primarily due to elimination of non-mandatory report-review visits. Care initiation moved forward from 6 weeks at baseline down to 4 weeks after intervention (2 weeks, 33% reduction).
Conclusion: Implementation of Navya Earthshot improved adherence to standard-of-care guidelines and reduced time to treatment in a resource-constrained setting. Integrating validated AI decision support tools into oncology workflows can strengthen cancer care delivery and enable broader adoption of guideline-based treatment in resource-constrained health systems.
利益披露 Disclosure
V. Khanna, None..
N. Gampalahalli, None..
N. Ramarajan, None..
A. Bhagat, None..
A. Punpale, None..
P. K. Tike, None..
B. Zawar, None..
R. B, None..
N. Ghadyalpatil, None.