stub How AI Can Help Us Prepare for the Second Wave - Thought Leaders - Unite.AI
Connect with us

Artificial Intelligence

How AI Can Help Us Prepare for the Second Wave – Thought Leaders

Updated on

By Eric Paternoster, CEO of Infosys Public Services

So far, existing data science models haven’t done the best job at predicting the ease of transmission of COVID-19, the extent of its development, and outbreaks in new hot spots. Many were developed in a rush, with limited information.

An AI model, however, would be adaptive, built to scale, and automated, crunching together sociological, economic, and COVID related health data to enable economies to reopen successfully should another wave occur.

The data used in this model should be both accurate and statistically significant. It must also be reliable. So far, things like R-values, herd immunity levels, and fatality rates have been very difficult to estimate across geographies, especially in places without a coherent testing and contact tracing strategy. Another problem has been that, even when good testing has been carried out, there have been wild differences in sensitivity and specificity rates, caused by variance in immunodiagnostic test types and specimen collection techniques.

Not only is the data lackluster, but the models themselves have been flawed. The model used by the White House, built by the Institute for Health Metrics and Evaluation, failed to take into account differences in key regional parameters and assumed that the virus would affect the population in the same way it did in China, Spain, and Italy. Of course, the U.S. has very different population characteristics, levels of quarantine, and testing availability.

Other models, often developed by leading universities around the world, did a little better. They incorporated estimates of contagion, along with factors that increase the risk of serious illness or death. But even these were based on inaccurate assumptions, leading to errors in the working model. For instance, the model initially developed by the Imperial College London failed to infer the obvious change in population behavior that would still arise in the absence of government-mandated interventions. It also lacked understanding as to how the virus reproduction (R0) number would change due to this behavior.

No wonder then that so much confusion has resulted, especially in the U.S. and the UK. Easing controls without preparation for the fallout has been costly, even when the disease is likely to return. Measures must be taken now to inform decision-making at a more granular level. Populations must be stratified to determine who emerges from lockdown first. A strategy must be implemented to enable contact tracing at scale and ensure that health care is sufficient in the future.

To help in this, artificial neural networks and deep learning techniques should be used, augmenting existing epidemiological models and making them more dynamic and responsive in real time. This AI model would use semi-supervised or unsupervised learning and could work even with limited input from large-scale testing reports. It would be self-sustaining and require a reduced amount of data to learn and predict, compared with current models. By continuously adjusting input parameters and continuously learning, the model would generate predictions that wouldn’t suffer from inevitable adjustment delays.

With deep learning, AI could discover complex patterns, self-learn, and self-heal automatically. It can auto-detect anomalies and would also be able to judge the accuracy of variables, producing much more reliable results than existing COVID data science models.

Key parameters in this AI model would draw from clinical test reports, contact tracing data and large regional data sets, and include regional population characteristics, socioeconomic status, and risk factors such as smoking, drug dependencies, and obesity. The number of infected individuals who quarantined and could no longer spread the infection would be incorporated into the model.

This would give task force leaders the insights needed to stem this dangerous disease in a proactive way, enabling them to make rational decisions in next-to-real time, providing world economies with a robust and well-informed exit strategy.

Eric Paternoster is Chief Executive Officer of Infosys Public Services, an Infosys subsidiary focused on public sector in US and Canada. In this role, he oversees company strategy and execution for profitable growth, and advises public sector organizations on strategy, technology and operations. He also serves on the Boards of Infosys Public Services and the McCamish subsidiary of Infosys BPM.

Eric has over 30 years of experience in public sector, healthcare, consulting and business technology with multiple firms. Prior to his current role, he was Senior Vice President and Head of Insurance, Healthcare and Life Sciences business unit, where he grew the business from $90 million to over $700 million with 60+ clients across Americas, Europe and Asia. Eric joined Infosys in 2002 as Head of Business Consulting for Eastern US and Canada.