Putting corona victims’ data to good use

As of the end of April, there are more than 1 million corona infections in the United States and the global death toll has crossed 200,000. We propose a mechanism to save lives through the use of the data from the infected, hospitalized, and deceased individuals.

Corona virus infection does not lead to health complications or to death in the same manner for all individuals. People with certain co-morbidities are at much higher risk (many times, compared to the general population) for developing complications, hospitalization, and death. A recent JAMA article (see: www.the-scientist.com/news-opinion/nearly-all-nyc-area-covid-19-hospitalizations-had-comorbidities-67476?) shows that 94% of the covid patients hospitalized in NYC had a chronic health problem, and 88 percent had two or more. WHO has identified 9 co-morbidities (age, sex, uncontrolled diabetes, hypertension, cardiovascular disease, lung disease, chronic liver and kidney diseases, immune deficiency) and data from China, US, Italy, India, and England has shown these to be relevant.

As India has shown, lock downs and social distancing can contain the spread even in poorer nations. But what happens when the lockdowns are lifted? It is very difficult to isolate everyone from the virus for long periods of time. The most optimistic predictions are that a suitable vaccine is 12 to 18 months away. Models have predicted that from 30 to 60% of the world’s population will be exposed to the virus before the first wave is over. Estimates from antibody tests in NYC show that about 25% of the population has already been exposed to the virus. Rather than locking everyone in, a more effective way is to focus on the high risk individuals. These people can be isolated when the rest of the population develops herd immunity. Some countries like Sweden is already pursuing this strategy with some success (see www.nytimes.com/2020/04/28/world/europe/sweden-coronavirus-herd-immunity.html and www.nbcnews.com/nightly-news/video/sweden-adopts-different-approach-to-battling-coronavirus-82633285580)

For selective isolation to be effective, we need to accurately determine each individual’s risk for developing complications, hospitalization, and death from a corona virus infection. Recently Artificial Intelligence (AI) models have surpassed the ability of humans in many areas (e.g., playing GO, recognizing faces, understanding speech) by learning from data. With over 3 million cases and over 200,000 deaths world-wide, there is enough data to develop accurate AI models for calculating an individual’s risk. Using these AI models, high risk individuals can be identified through surveys that ask about the 9 co-morbidities. For people with access to smart phones, these surveys can be through web-based tools or Apps, and in villages, the surveys can be door-to-door. We have preliminary success in both.

There are many published summaries of co-morbidities in communities (For example, see NY state summary at: https://covid19tracker.health.ny.gov/views/NYS-COVID19-Tracker/NYSDOHCOVID-19Tracker-Fatalities?). But there is no individual-level anonymized co-morbidity data available to researchers to build good AI risk models. A co-ordinated effort is needed among major hospital systems and healthcare organizations to rectify this. Once such data is available, a co-ordinated effort is also needed among AI experts to build accurate risk models. The 200K deaths would then become useful in saving the lives of those remaining!

K P Unnikrishnan

Chief Scientist and Co-founder at eNeuroLearn

Newsletter Subscription

Sign Up to enjoy top Finance Articles and Insights Delivered to your inbox.
  • Programic Asia Pvt Ltd may keep me informed with eMails about cfoengage.com & finance newsletters. See our Privacy policy for more details