Big Data & Public Health: Think Big, Think Small, and Think Once is Enough

Earlier this week, thanks to a program from the USF College of Public Health led by Dr. Marissa Levine, I was fortunate to be invited to present to a group of 20 public health officials who came from all over the state of Florida. My session was on “Big Data, Machine Learning and AI for Health”.

I presented many data science examples from both health and non-health domains to his group, with the goal of generating ideas for discussion on how big data can be used in the domain of public health.

In the discussion that followed a few things emerged.

(1)  Public health officials are close to the community, close to specific health issues, and close to action and policy that can move the needle.

(2)  They often are very specific in the questions they ask and study – e.g. measuring HIV incidence rates over time accurately in their communities. But of course they ask many such “small" (meaning, specific) questions, each of which is important and "its own thing". In my own work I have repeatedly seen that asking very small/specific questions is the way to go to demonstrate value.

(3)  Data is available in their world, but data quality and sharing are issues, so the use is perhaps less than it could be.

(4)  Budgets are limited, and their abilities to hire data scientists and build data-oriented teams is limited at the local level. It is better at the state level and even better at the federal level but still significantly less than say what a toothpaste company might be spending today analyzing how people use toothpaste across the world to help the company better make toothpaste and target the right customers. Toothpaste is important. But so is public health.

(5)  Translating what they do into action involves coordinating with a lot of other agencies and policy makers. So to move the needle they need a “big tent” approach to involve all the players.

(6)  There were many powerful opportunities to use data better, but this would necessarily need such a “big tent” approach that involves: (a) public health agencies at the local/state levels (b) hospitals and insurance companies that see patients and pay for them (c) federal agencies (like CDC and others) (d) universities that can help do the heavy lifting on the data science side and (e) companies like Google, Facebook, Verizon, Nielsen that may be able to offer unique data to help the very specific questions being asked.

This essentially is the heart of the "Thing Big (tent) and Think Small (problems)" framework as a way forward to embrace data-driven ideas in the areas of public health.

The Coronavirus and COVID-19. Once is Enough.

Coincidentally, thanks to Ramesh Saligame, today I came across an outstanding, specific example of this, that has worked in containing the novel coronavirus: https://jamanetwork.com/journals/jama/fullarticle/2762689

As brilliantly shown in the JAMA article above, Taiwan, with its proximity to China, has 45 detected cases of COVID-19 as of today.

Yes. Forty Five. Not 450, or 4,500 or 45,000. Taiwan is about as close to the epicenter of this outbreak as any place can get. There is a lot of travel from the Wuhan area to places in Taiwan. Despite this they have 45 cases (and not because they are not testing people).

 How did they do this?

As the JAMA article shows, learning from SARS and prior epidemics, Taiwan has over the last many years built a country-wide big data-driven methodology to tackle a very specific (“think small”) problem: managing outbreaks before they affect their population.

The good thing about people who travel a lot is that they bring a lot back. The bad thing about people who travel a lot is that they bring a lot back. We need the ideas and not the pathogens. By integrating travel data of their citizen with data in their electronic health tracking systems, Taiwan is able to tell when a patient reports to a hospital what the level of severity could be, and what the patient needs to be tested for based on where all they've been. Patients who tested positive can be tracked through their phones to ensure that quarantine is maintained, and to perhaps use that data to see who else might be vulnerable and needed to be tested. There is a lot more data integration and use as well.

It was not all “big data”. There was also a “big tent” approach that included efforts in fighting disinformation and involving the right agencies for that. The JAMA article is brilliant in its clear description of what the other agencies did and how it all was a planned and systematic effort.

Can we do this in the US?

Maybe. There will be some privacy sacrifice that may be needed (data sharing across travel, and health databases for instance). Maybe we can design privacy-sensitive ways for doing this and blockchain-based ideas could work here. It will probably require annual funding that is significant in the health area – maybe $10 billion or more to do this very well. But chances are it can generate economic activity proportionally, both in doing this, as well as keeping people healthier to work.

What is the cost of not doing this? 

In addition to lives, fear, and massive disruption, the COVID-19 crisis has lost about $6 trillion in wealth so far in the US from the stock market drop. While some of this will come back as the market recovers in the future (we will conquer this), nobody wants to go through this again. As the financial quarter winds down we may get more information about how many lives were adversely affected economically – maybe over 50 million in the US. Those whose daily lives depend on foot-traffic or the gig economy are particularly severely affected right now.

What next?

Our leaders have addressed us recently - telling us about what they are doing economically, what they are doing to make tests available, and what they are doing to contain this spread – including seemingly draconian shutdowns to “flatten the curve”, which is the correct thing and best we can do at this point. Schools are being shutdown across the country for weeks. Imagine how sad students will be, not being in class learning geometry proofs. More seriously, millions of young students will not get their daily meals since they depend on school for food. The US National School Lunch Program feeds about 31 million students enrolled in our schools. Many of these students will likely eat much less given the loss of wages to their parents in this crisis. As awful as this is, a lot of this is needed now to keep the curve flat and outlast this crisis.

As politicians like to say, never let a crisis get to waste. If there is a way to use the crisis to mobilize effort around a powerful cause then so be it. So let's use this to figure out what infrastructure we can build to prevent this from happening again. Our leaders should also now present such a framework for the future, possibly similar to what was done in Taiwan, to help prevent these types of situations in the future. The markets and the citizens will react much better if they know a plan for the future exists, where they can have confidence that things like this will happen only once, since once is enough.

This will require significant new funds, and a “Niche Problem Big Data Big Tent” approach. Much like the one we just saw in Taiwan, that could have now had 45000 COVID-19 cases, but has only 45. Taiwan's economic impact is likely to be much less as well, and the savings with this incident alone could have paid for a good chunk of how much they have invested in building this system over the years.

Reference (Must-Read)

Wang CJ, Ng CY, Brook RH. Response to COVID-19 in Taiwan: Big Data Analytics, New Technology, and Proactive Testing. JAMA. Published online March 03, 2020. doi:10.1001/jama.2020.3151https://jamanetwork.com/journals/jama/fullarticle/2762689