Although it continues to run strong in some countries, the Covid pandemic is in retreat in many parts of the world, including much of the US. With that, it will likely be some time before the world returns to pre-2020 conditions. But even as countries and cities reopen, questions about how authorities handled coronavirus are beginning to surface – with the question of whether the total lockdown policies implemented by many governments were the best way to prevent the spread of Covid-19.
That question is not just academic; determining the effectiveness of lockdowns will help us prepare for future pandemics. But the answer is not at all clear; in fact, the forced closures of businesses of all kinds seemed to have had little effect on the death and infection rates in several US states. Florida and California offer good examples. During the height of the pandemic, both states averaged a Covid-19 case rate of around 8,900 per 100,000 residents, with death rates similar as well, according to the Centers for Disease Control and Prevention. This, despite the fact that there was no extensive statewide shutdown or forced mask-wearing mandate in Florida – while California was the first state to impose a statewide shutdown, and one of the last to lift it. The leaders of both states – as well as others with widely divergent Covid policies but similar outcomes – have explanations as to why they think their policy was more effective, but one thing is clear: Those policies were based on factors other than data on infections, causes, effects, level of infection, hospitalization, seriously ill, deaths, suicide and depression rates, comorbidities in Covid victims, best practices that lower infection rates, solid statistics on how effective social distancing really is.
Had that data been available, it’s possible – perhaps even likely – that governments would have implemented different policies. While most policy makers made “follow the data” a mantra, it turns out that most didn’t – not because they didn’t want to, but because they didn’t have access to the complete set of data they needed to accurately evaluate. Armed with the requisite data, governments would be able to decide on the appropriate policy that will keep infection/illness rates as low as possible, with the least damage to the economy, individual rights and lifestyle. The question then is – how could they acquire that data?
In order for governments to properly set policy, they need data from a wide variety of sources, including information on employment, bankruptcy filings, mortgage defaults, drug use levels, and more. Much of that data is compiled by representatives of government – courts, police, unemployment offices, state welfare agencies, etc. But as Covid-19 is a health crisis, policy makers also need health data, both for Covid and non-Covid patients, such as those suffering from depression.
With that data, governments can contract with companies that do advanced artificial intelligence (AI)-based analysis, understanding the relationship between thousands of data points to get an accurate picture of the impact of the pandemic on the population. Combined with data they already have access to, governments would be able to more fully understand the impact of Covid-19 on society, and the impact of policy on Covid spread rates.
But getting accurate and complete health data isn’t so simple. Similar to the GDPR, US states are significantly limiting the ability of organizations and businesses to share data without the express consent of individuals. The most advanced US state law in this regard is the California Consumer Privacy Act (CCPA), but numerous other states are working on laws with similar restrictions.
With that, many hospitals do share de-identified data on patients, including lab and imaging results, medication lists, and more. But once hospitals share data with outsiders, it opens itself to the possibility that hackers will gain access to data and use advanced methods for re-identification of sensitive patient data. Besides running afoul of government regulations on data privacy, such attacks could cost hospitals a lot of money.
And if the objective is to provide data for proper AI analysis, allowing that kind of access just isn’t enough to achieve the full potential embedded in the data. The data that is released is usually incomplete, and much of it is inaccessible, stored on various systems that are not compatible with each other or AI systems. The data policymakers need is there – but because it’s difficult or impossible to get to, it has only a fraction of the effectiveness it should have.
The challenges, then, are to ensure data security and to utilize as much of the data as possible, from as many hospitals as possible. One way to ensure security is by applying differential privacy policies to the data that is to be shared. DP algorithms enable the analysis of real and rich data, ensuring that reusing the data to identify any specific individual will be impossible.
That analysis could be done within the confines of the data lake belonging to the hospital that is in charge of the data – thus ensuring that it cannot be stolen in transit or directly from the outsiders, that don’t necessarily conform to the security policies of the healthcare institution. Only when strict control with granular policies and strong privacy techniques like differential privacy are used, the relevant data can be made accessible at scale to AI systems will crunch the data from that hospital, along with data from many others, to arrive at answers that will help policymakers accurately understand what they are facing.
The same system can be used for medical research in any area – the development of pharmaceuticals, cancer research, pediatrics, etc. With strong privacy guarantees – and practically impossible to re-identify, using advanced DP algorithms – and uploaded to a secure server where AI analysis can take place, hospitals, patients and regulators can gain confidence that the data is secure, while it is working to help solve health problems.
The Covid crisis is quickly fading, and it’s mostly due to the rapid development of the vaccines that are enabling us to emerge from the ravages of the pandemic – thanks in part to the judicious use of data to set policy, develop methods to protect society, and to research effective treatments. Our 2020 pandemic experience provided us with many lessons – including one on how we have a tremendous opportunity to better use existing data to speed up processes and design new treatments, diagnoses, and best practices in a faster, cheaper, more effective, and more accurate manner.
Photo: Asurobson, Getty Images