SBIR Phase I: Harnessing Natural Language Processing for Scalable Text-based Behavioral Health Care

这个小企业创新研究（SBIR）第一阶段项目的更广泛的影响/商业潜力是解决系统性的行为健康和物质使用提供者短缺，具有成本效益地提高患者在治疗中的保留率，并积极治疗慢性行为健康患者。如果不加以治疗，这些疾病每年造成的损失超过1万亿美元，并导致无数人过早死亡。同伴支持是一种有效的工具，可以让不愿意或无法获得临床护理的患者参与，特别是在边缘化人群中。 然而，扩大同行支持是具有挑战性的，目前的在线支持论坛充斥着拖钓和滥用。我们的自然语言处理（NLP）工具可以推断短信的情感，自动标记临床相关或关键内容。这使得临床医生可以通过将时间集中在最需要的患者身上来轻松地调节群体，而同行则可以生成日常参与所需的接触点。这个小企业创新研究（SBIR）第一阶段项目将大大提高临床医生在支持小组内跟踪患者心理健康的能力。目前，管理同龄人群体的一个挑战是确定一个群体作为一个整体的健康状况-有些群体可能比其他群体更具建设性。考虑到在线支持小组中生成的消息量，以及护理经理和同伴支持专家的预期工作量，他们可能管理数十个小组，如果没有技术的帮助，这是一项不可能完成的任务。为了实现这一目标，我们专注于三个主要领域：1）通过开发新技术来识别和跟踪可能在组中同时发生的多个对话，从而提高我们现有NLP算法的性能，2）开发一种通过分析同伴之间的互动来跟踪一个群体的整体健康和稳定性的方法;设计新的界面，有效地显示算法产生的所有见解。这些NLP工具将为一个平台提供动力，让患者获得更多的支持。提供者将有机会获得一个新的高保真数据源，以更好地分诊外展和个性化护理。这个奖项反映了NSF的法定使命，并已被认为是值得通过使用基金会的智力价值和更广泛的影响审查标准进行评估的支持。

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is to address a systematic behavioral health and substance use provider shortage, cost-effectively improve patient retention in treatment, and proactively treat chronic behavioral health patients. Left untreated, these conditions cost over $1 trillion annually and result in countless early deaths. Peer support is an effective tool to engage patients unwilling or unable to access clinical care, particularly in marginalized populations. However, scaling peer support is challenging, with current online support forums rife with trolling and abuse. Our Natural Language Processing (NLP) tools can extrapolate the emotional sentiment of text messages, automatically flagging clinically relevant or critical content. This allows clinicians to easily moderate groups by focusing their time on the patients most in need, while peers generate the touchpoints necessary for day-to-day engagement. This Small Business Innovation Research (SBIR) Phase I project will greatly enhance the ability of clinicians to track the mental health of patients within a support group. Currently, a challenge in managing peer groups is identifying the health of a group as a whole - some groups can be far more constructive than others. Given the volume of messages generated in an online support group, together with expected caseloads for care manager and peer support specialists, who may be managing dozens of groups, this is an impossible task without the aid of technology. To achieve this goal we focus on three main areas: 1) improving the performance of our existing NLP algorithms by developing novel techniques to identify and track multiple conversations that might be co-occurring in the group, 2) developing a method of tracking the overall health and stability of a group by analyzing interactions among peers and 3) design new interfaces that effectively display all of the insights generated by the algorithms. These NLP tools will power a platform to give patients more access to support. Providers will have access to a novel high-fidelity data source to better triage outreach and personalize care.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.