Use Cases, Goals, Challenges

Wednesday, April 8, 2020

COVID-19 Use Cases, Goals, Challenges, Impact

AI Use Cases to Address the COVID-19 Pandemic

First of all, we want to take a moment to thank all of the Healthcare workers on the front line of this pandemic.  You are the real heroes and we appreciate everything you are doing.  

At CognitiveScale we are eager to do our small part in helping to build out solutions across the public health system, including detection, engagement, care delivery, and service insights - delivered with trust and assurance.  

CognitiveScale has been engaged in a number of areas related to the COVID-19 pandemic.  There are numerous use cases for Artificial Intelligence (AI) that we wanted to touch on - and we will expand on this in a longer format white paper, AI Use Cases and the COVID-19 Pandemic_ Goals, Challenges, and Impact, available soon on the Cognitive Scale website. 

AI Use Cases in Public Health

AI-powered solutions are impacting Healthcare processes from high-level public health authority insights to more personalized insights, for example:

  • Location-based hot-spot detection and prediction
  • Patient risk scores - combining location-based risk scores with patient specifics (e.g. age and comorbidity information)
  • Hospital capacity predictors
  • Smart devices (e.g. app-powered thermometers, with over a million in use in the USA) provide hot-spot detection insights
  • Curated news feeds for specific high risk locations
  • Natural Language Processing (NLP) solutions like COVID-19 hotline support for routing calls 
  • Epidemiology use cases like spread simulators
  • Supply chain predictors e.g. for drug supplies and personal protective equipment

There are many more, thankfully, and more to come.

Challenges with AI-Powered COVID-19 and Public Health Solutions

At a high level, challenges with the deployment of AI solutions in the public health arena include:

  1. Personalization: how to bring aggregated data and predictions (e.g. high risk areas) to a hyper-personal level
  2. Accuracy: For personalized solutions like determining individual risk scores and treatment guidelines, these are life-and-death decisions, so solutions must be accurate obviously
  3. Data: access to data is a common challenge for AI solutions, but in a public health setting there are even more disparate systems across broader networks
  4. Speed: traditional sources of data in some healthcare settings, like claims, don’t always arrive in time to help in this context (e.g. to provide a signal or pattern in data like a diagnosis or service code), or, systems might not be architected to collect and analyze massive amounts of data and connections in real time
  5. Trust, Ethics and Assurance:  can models be trusted, are they accurate, are there bias / fairness / explainability issues, and will providers overcome skepticism about machines that are making decisions (vs. augmenting intelligence)?  Bringing government, community or health plan level data to an individual level - and then sharing it across a public health system - can involve privacy issues and compliance challenges. Trusted AI is particularly challenging in public health settings.

CognitiveScale Use Cases

Leveraging prior work that we have done, our platform and software, and even solutions we have developed in areas like Service Experience and Care Management, CognitiveScale has been working with some large Healthcare organizations to leverage these existing assets for public health use cases, including response to the COVID-19 pandemic.  For example, we are working on:

  • Guided help and call center support based on personalized member-patient profiles - with the goal of “the right call center agent (customer service vs. telemedicine or COVID-19-trained), with the right directions at the right time.”  This type of solution requires_
    • Integration of disparate data sources_ government (e.g. CMS, NIH, CDC), claims, lab results, prescriptions, member benefit data and demographics, 
    • Orchestration of multiple models including location- and community-based risk scores + individual risk score models = personalized insights
    • Multi-channel delivery of insights_ surfacing insights to call center employees (on numerous IT systems) and eventually to self-service outlets (in app, patient/member portals)
    • Prescriptions like next best action, e.g. directions on the nearest testing centers, public health resources, or hospitals with capacity
  • Intelligent telemedicine guidance based on hyper-personalized patient or member profiles - with the goal of augmenting the intelligence of caregivers working through symptom checkers or trying to provide personalized direction to member-patients.  Again, this kind of solution requires integration of numerous disparate data sources, orchestration of multiple models, and delivery insights to multiple channels. 
  • Intelligent interventions with monitoring.  Smart devices, in home devices, and apps can provide insights into public health issues like COVID-19 as long as patients opt in and participate in data sharing - at least offering to share their data.  Trust and compliance are particularly important in monitoring solutions.

“Waze for Public Health”: Enabling Collaboration Across Public Health Systems

Many drivers would be lost without a map app telling them how to get from place to place - and some appreciate the intelligence of Waze, a particularly smart map and directions app.  What makes Waze special is mass collaboration - the fact that end users participate in improving the performance of the solution. Waze drivers tell others where there are impediments to speed, for example (construction, accidents, police) - often while they are driving (not recommended) - and this insight, in turn, improves performance - often resulting in other drivers getting better recommendations on their route, for example.

AI-powered public health use cases get ‘cognitive’ when they learn and improve over time - sometimes over seconds and minutes.  This depends on data, of course, and in the case of public health crises, this can mean data across a massive network of sources.  Mass collaboration - from data sharing to trusted use of data - will certainly help improve the overall intelligence of public health systems thanks to AI.

In the case of this COVID-19 pandemic, near-future flare-ups, or other viruses, the need for speed can save lives, obviously.   When everyone using an AI-powered public health solution - member-patient-citizens, providers, government officials, etc. -can improve the intelligence of the solution, community- and individual-level AIs will begin to protect the public and save lives.  Fortunately there are a number of intermediate steps we can take to improve underlying detection and prediction, engage the population better, and offer better guidance.

More to Come Soon…

CognitiveScale’s “intelligent intervention” work in Healthcare is all based on enabling collaboration of data sources and model developers, orchestrating trusted and scalable solutions, and augmenting the intelligence of Healthcare workers.  Watch for a white paper that will elaborate on this topic_ AI Use Cases and the COVID-19 Pandemic_ Goals, Challenges, and Impact, available soon on the Cognitive Scale website.

And thanks again, sincerely, to all of the Healthcare workers on the front line.

Download the Whitepaper

About the Author

Jeffrey Eyestone Aug09 bw web

Tags for this resource
Share this page