AI has advantages for COVID-19 vaccine rollout, but potential dangers too

人工智能对新冠肺炎疫苗的推出具有优势,但也存在潜在危险

2021-02-12 05:01:53 Healthcare IT News

本文共2083个字,阅读需6分钟

The question of who should get access to COVID-19 vaccines first has varied from state to state, with some governments prioritizing those with high-risk conditions and others lowering the age of eligibility.  One South Dakota-based system, Sanford Health, is using a machine learning model to identify which individuals are at greatest risk of having severe COVID-19 outcomes – and applying the algorithm to eligible groups.   "With [those] 85,000 people what we can do is take a real-time picture that evolves over time, using computer learning to tell us what patients or what people in the Midwest get the sickest from COVID-19," said Sanford chief physician Dr. Jeremy Cauwels to Minnesota Public Radio. Cauwels told MPR that he believes an artificial intelligence approach is more equitable than random choice for administering the vaccine.  Sanford isn't alone. Experts say AI has big potential to assist with the COVID-19 vaccine rollout.   "The pace and scale of the vaccine rollout is unprecedented, and we are seeing AI play a role," said Lori Jones, chief revenue officer and president for the provider market at Olive, an AI-as-a-service vendor.   Rather than using AI to identify at-risk patients, Jones noted its potential to promote efficiency within existing workflows. "The biggest areas of focus for organizations that we’re working with have all related to managing the organization, scheduling, preregistration and communications activity around the testing and vaccines themselves, with additional automation activity to streamline patient communications and drive better vaccine efficacy by ensuring patients are aware, prepared and present to receive second doses," Jones explained.   "We’ve got an important mission ahead of us still, and if we can’t expand the capacity of organizations delivering the vaccines to take on more patients faster, then there is a very real risk that this process could take years, not months," she said.   Jones pointed to chatbots as a prime example of the way AI can be used in conjunction with other tools, specifically when it comes to patient engagement.   "AI-enabled digital call centers are helping organizations manage the significant level of interest in key vaccine information," she said. "FAQs can be converted into chatbots to refresh the available information to be COVID-19-specific." "If the healthcare industry continues to rely on paper forms, phone calls, mobile apps, portals and email campaigns, process bottlenecks will create long lines, confusion and frustration," agreed Greg Johnsen, CEO of LifeLink, which powers conversational solutions for healthcare organizations.    "Additional complexities around new documentation, specific follow-up vaccination windows and an influx of people that are new patients could overwhelm current intake and scheduling processes," said Johnsen. "Building a handful of digital assistants versus training thousands of individuals is also a key consideration when it comes to efficiency and cost."    That said, there are unmistakable downsides to relying on AI – namely, expecting the technology to be infallible.    "Pursuing AI strategies can certainly bring about adoption challenges, and adoption is critical to any AI strategy," said Jones. "One roadblock to AI adoption is understanding that AI tools aren’t replacing human healthcare workers: They’re actually empowering them and helping them work better, faster."   There are also the ever-present dangers of reproducing bias or faulty algorithms. In December of this past year, Stanford Medical Center came under fire for prioritizing administrators over frontline health workers due to an error in the rule-based formula it was using to help calculate who would get vaccinated first. Sanford, in South Dakota, is also not using race or ethnicity as a factor in its algorithms, theorizing that individuals with higher rates of chronic disease will be elevated in the prioritization.    But given the disproportionate effect of COVID-19 on patients of color – especially Black, Latinx and Native people – other health systems in nearby states say it's important to take those demographics into account.   The University of Wisconsin-Madison did use a race-based algorithm to prioritize employee vaccines in its initial distribution, Shiva Bidar-Sielaff, chief diversity officer at UW Health, told MPR.  "It's incredibly important to realize that all data points to the fact that, unfortunately, race and ethnicity have been shown to create a much higher risk of hospitalization and death for COVID-19," said Bidar-Sielaff.    "So when we looked at our algorithm, we saw that if you add age and SVI, which has that component of race and ethnicity, it's a multiplier effect in how much higher risk an individual is at for hospitalization and death," she said. Some companies are stressing the need for caution when it comes to allocating vaccines.  Representatives from Salesforce, which launched its Vaccine Cloud tool this past month to assist clients with managing vaccine administration, said they were working to ensure equitable distribution.   "Vaccine Cloud can deliver integrated and customized solutions for our customers, including the ability to use data and insights to support [the] distribution, management and administration of vaccines," a Salesforce spokesperson told Healthcare IT News via email.   "However, our Principles for the Ethical Use of COVID-19 Vaccine Technology Solutions explicitly state that AI should not be used to predict personal characteristics or beliefs that would affect a person’s or group’s prioritization for access to vaccines, and we work closely with our partners and teams on this guidance."   Still, it's clear that AI – when deployed responsibly – may be able to make the COVID-19 vaccine rollout faster and more effective for at-risk patients. "The vaccine rollout is the ultimate test for AI to showcase the breadth of time-saving and efficacy capabilities, and demonstrate its full value for healthcare leaders," said Jones. "When organizations emerge from the COVID crisis, we see AI becoming an integral part of their digital strategy." Kat Jercich is senior editor of Healthcare IT News.Twitter: @kjercichEmail: kjercich@himss.orgHealthcare IT News is a HIMSS Media publication.
谁应该首先获得新冠肺炎疫苗的问题因州而异,一些政府优先考虑高危人群,另一些政府则降低了获得疫苗的年龄。 一个位于南达科他州的系统,桑福德健康,正在使用一个机器学习模型来识别哪些人有最大的风险出现严重的COVID-19结果,并将算法应用于符合条件的群体。 桑福德大学的主治医师杰里米·考尔斯博士对明尼苏达州公共广播电台说:“对于这8.5万人,我们所能做的就是拍摄一张随时间变化的实时照片,利用计算机学习来告诉我们哪些患者或中西部地区哪些人感染新冠肺炎最严重。” Cauwels告诉MPR,他相信人工智能方法比随机选择疫苗更公平。 桑福德并不孤单。专家表示,人工智能在帮助新冠肺炎疫苗研发方面具有巨大潜力。 “疫苗推出的速度和规模是史无前例的,我们看到人工智能正在发挥作用,”人工智能即服务供应商Olive的首席收入官兼提供商市场总裁洛瑞·琼斯说。 Jones没有使用人工智能来识别有风险的病人,而是指出了它在现有工作流程中提高效率的潜力。 琼斯解释说:“我们正在与之合作的组织的最大重点领域都与管理组织、日程安排、预注册以及围绕测试和疫苗本身的沟通活动有关,并通过额外的自动化活动简化患者沟通,通过确保患者意识到、准备好并准备好接受第二剂疫苗,从而提高疫苗效力。” 她说:“我们面前还有一个重要的任务,如果我们不能扩大提供疫苗的机构的能力,以更快地接诊更多的病人,那么这个过程可能需要几年,而不是几个月,这是一个非常真实的风险。” Jones指出,聊天机器人是人工智能与其他工具结合使用的一个主要例子,特别是当涉及到病人参与时。 她说:“支持人工智能的数字呼叫中心正在帮助组织管理对关键疫苗信息的重大兴趣。”“常见问题可以转换成聊天机器人,刷新可用信息,使其与新冠疫情相关。” “如果医疗保健行业继续依赖纸质表格、电话、移动应用程序、门户网站和电子邮件活动,流程瓶颈将导致排长队、混乱和沮丧,”为医疗保健机构提供对话解决方案的LifeLink首席执行官格雷格·约翰森表示同意。 Johnsen说:“新的文件、特定的后续疫苗接种窗口以及新患者的大量涌入可能会使目前的摄入和计划流程不堪重负。”“在效率和成本方面,建立少数几个数字助理,而不是培训数以千计的个人,也是一个关键的考虑因素。” 话虽如此,但依赖人工智能的缺点是显而易见的--即期望这项技术是万无一失的。 Jones说:“追求人工智能战略肯定会带来采用方面的挑战,而采用对于任何人工智能战略都是至关重要的。”“采用人工智能的一个障碍是理解人工智能工具并没有取代人类医疗保健工作者:它们实际上是在赋予他们权力,帮助他们更好、更快地工作。” 此外,还存在着重现偏见或错误算法的危险。去年12月,斯坦福医疗中心因将管理人员优先于一线卫生工作者而遭到抨击,原因是该中心用来帮助计算谁将首先接种疫苗的规则公式出现错误。 南达科他州的桑福德也没有将种族或族裔作为算法的一个因素,他认为慢性病发病率较高的个人将在优先排序中得到提升。 但鉴于COVID-19对有色人种患者--尤其是黑人、拉丁裔和本地人--的影响过大,附近各州的其他卫生系统表示,将这些人口统计因素考虑在内很重要。 威斯康星大学麦迪逊分校确实使用了一种基于种族的算法,在最初的分配中为员工的疫苗排定优先级,威斯康星大学卫生部门的首席多样性官Shiva Bidar-Sielaff告诉MPR。 Bidar-Sielaff说:“认识到所有数据都表明,不幸的是,种族和族裔导致新冠肺炎住院和死亡的风险要高得多,这一点非常重要。” 她说:“因此,当我们研究我们算法时,我们发现,如果你把年龄和SVI(包含种族和民族因素)相加,就会产生乘数效应,即一个人住院和死亡的风险会增加多少。” 一些公司强调在分配疫苗时需要谨慎。 Salesforce的代表说,他们正在努力确保公平分配。Salesforce上个月推出了疫苗云工具,帮助客户管理疫苗接种。 Salesforce发言人通过电子邮件告诉Healthcare IT News:“疫苗云可以为我们的客户提供集成和定制的解决方案,包括使用数据和洞察力来支持疫苗的分配、管理和行政的能力。” “然而,我们关于COVID-19疫苗技术解决方案的伦理使用原则明确指出,不应使用人工智能来预测会影响个人或群体获得疫苗的优先次序的个人特征或信念,我们与合作伙伴和团队在这一指导上密切合作。” 不过,很明显,人工智能--如果负责任地部署--可能能够使新冠疫苗更快地推出,对高危患者更有效。 Jones说:“疫苗的推出是对人工智能的最终测试,它将展示其节省时间和功效的能力,并展示其对医疗保健领导者的全部价值。”“当组织从COVID危机中走出来时,我们看到AI成为他们数字化战略的一个组成部分。” Kat Jercich是Healthcare IT News的高级编辑。Twitter:@KjercicheMail:Kjercich@HIMSS.orghealthCare IT News是HIMSS的媒体出版物。

以上中文文本为机器翻译,存在不同程度偏差和错误;偶尔因源网页结构局限,内容无法一次完整呈现。请理解并参考原站原文阅读。

阅读原文