DataDrivenMD

The challenge to building a screening app is that there is no universal screening criteria (more on this in a bit). Moreover, the criteria has been changing by the day. For this MVP, we decided to stop chasing a moving target. We did the next best thing: link to the testing criteria that a location is using (to the extent that the information is publicly available).

Early in the process of sourcing data for this project, we realized that the CDC guidelines are used as a starting point, and that other factors come into play that are less transparent to the public. For example, there are still many states with extremely limited capacity. In those areas, public health departments have been forced to create lists of patients that would be tested immediately if it were possible. Lacking sufficient testing capacity, they prioritize the list of persons under investigation (PUI) using criteria that they develop internally through a collaborative effort with key stakeholders.

On our site, you will note that there are pockets of the country with lots of testing options. But that doesn't mean that testing is easy to come by. We learned that those areas are enriched for academic medical centers and large hospital systems that have been granted FDA approval for their own (custom) version of the COVID-19 test. Through strategic partnerships with community-based medical groups that predate the pandemic, patients in the broader community benefit from direct access to testing services. But the the problem has resurfaced: the demand for testing is already exceeding supply, and we witnessed the turnaround time for test results grow from 4 hours -> 12 hours -> 24 hours -> 4-5 days since we began this project exactly 7 days ago.

Make no mistake about it: the testing problem is far from resolved. Screening tools are a stopgap designed to lessen the mental anguish experienced by those who are tasked with rationing a (literally) vital resource. We will continue to explore ways to incorporate them into our project, but will approach the decision mindful of our commitment to empower everyone with the information they need to flatten the curve. To the extent that screening tools have thus far played a role in obfuscating where we are on the curve, you can rest assured that, if a screening tool finds its way onto our site, it will serve a greater purpose for a greater number of people.

Yes - all over the place. That was the biggest challenge: sourcing and validating the data. It's also why this is and will continue to be a work in progress: the situation on the ground is dynamic. We need help sourcing and validating data. If you're willing to volunteer, send a note to [email protected]

Short answer: they have all been bottlenecks thus far. In fact, that's a big reason why testing hasn't improved as rapidly as everyone hoped it would. The biggest bottleneck now is dwindling supplies of personal protective equipment (PPE) for those involved in obtaining and handling the test specimens.

Yes - Houston data will be coming soon. It took extra effort to figure out how testing was being coordinated among the TMC hospitals and their community-based partners. Stay tuned for updates - we are pushing data out regularly from this point forward.

Appreciate the slow clap - would love your help. See Andrew's note to learn about ways to help.

Also want to share that you're not imagining things: best we can tell, those living in Georgia have very few options beyond their health department. Even Emory is limited in its capacity - last I heard they were hoping to be able to get to 400 tests/day by this weekend. We'll add Emory to the site soon. Even so, I encourage those who live in Georgia to share their insight with us by adding testing locations so that we can help others in their community.

We will also continue to research and update the site as new options become available - including home-based testing. We've already heard from a couple of companies in this space, but we're waiting for them to provide us with proof that their test is FDA validated and approved. Once they do that, we'll add their information to the site as well.

We were careful not to assume that all public health departments would engage in testing. We manually verified the information, and for those public health agencies that explicitly indicated on their websites or through direct channels that they offered testing we marked them as "offers COVID-19 testing".

That being said, there are many areas of the country where the local, county, and state health departments effectively function as gatekeepers. In those communities, people have no choice but to call their health department and hope for the best. If they meet testing criteria, then they are referred to a testing location or fast-tracked through an existing healthcare provider’s queue.

If nothing else, this site crystallizes the fuzzy image that we all had in mind with respect to testing capacity. The situation is changing all the time, and we expect things to get better. You can help by raising awareness, and asking people you know in the Midwest and Southern states to visit our site and share their on-the-ground knowledge of the situation.

Well put. We’re working on the testing site locator right now. We’ve gathered contact info for 3500+ public health agencies, plus starting to add info for urgent care facilities, health systems, community hospitals, and academic health centers. We’ve got CIOs and CMIOs from across the country helping us gather the data we need. The goal is simply to make it easy for people to get tested.

Would be great to have your help. Send me an email if you’d like to learn more (see my profile)

Not being intimately familiar with the limitations of Kalman filters, how susceptible to bias is this approach when hidden states exist? In the case of COVID-19, the asymptomatic cases may be modeled as such, or not, depending on the approach. Also, the incubation and convalescent phase could also present a challenge given that patients can still transmit the virus.

It would be helpful to understand how Kalman filtering approach compares Markov chains, for example.

edit: Could Kalman filters be used to retroactively mark/flag/detect the introduction of a new strain of the same virus? If so, this could prove to be a novel way to quantify the clinically meaningful mutation rate - that is, the rate at which the virus mutates sufficiently to infect a new sub-population (or perhaps re-infect existing ones).

This reminds me of the time I spent at the NIH conducting research on cancer vaccines. After lots of trial and error, I had finally found something that seemed to prevent lung cancer in mice. I rushed to show my PI the results. He listened intently and asked thoughtful questions. Once I finished, I asked if the findings would be worthy of submitting to Nature.

He looked me straight in the eye and said, “I wouldn’t publish this anywhere, we still have work to do.”

“What do you mean?” I asked.

“If Nature published every article that purported to cure a human disease based on a mouse model, then I would be next in line for the Nobel Prize,” he replied.

Harsh but true. The challenge of curing human disease cannot be replicated in animal models. At least not yet.

> That seems like a harder problem because of patient privacy issues.

Agreed, but even a simple honor system would be better than the current situation. Perhaps verified via a simple browser-based cookie?

Separately: just dawned on me that HN has no obvious way for users to communicate directly. Any suggestions on how to share contact info without begging for spam/bots?

If we break down the challenge, we can start having an immediate impact. Here’s how:

- Now: there's an immediate need for a single source of truth for all the locations where COVID-19 testing is available. Trump recently said that Google was working on it, but it turns out not to be the case (they are weeks away from an MVP that would only cover the SF Bay Area). I propose we start by solving this problem.

- Next: What is the scope of COVID-19 infection? Today, the CDC doesn't have an accurate count of the # of COVID-19 tests that have been performed because they don't have a way to collect that info from the growing number of labs that are performing the tests. Moreover, there's no way for anybody to know how many people have actually been tested, because the same person can be tested multiple times. We could solve this problem by crowdsourcing that information directly from individuals.

- Next +1: We need to know where the sickest people are right now, and predict where they could be in the near future. This is crucial information for healthcare providers (like me) and public health officials alike. Without this information in hand, it's virtually impossible to know where to direct resources (like doctors, nurses, ventilators, medical supplies, testing equipment, etc.). We need to be able to track demand AND supply in real-time --> this is the point at which the project starts shaping itself into an Uber-like platform.

- Next +2: Enable individuals to post offers to help and for others to request assistance. This would enable individuals with medical training to help meet the need. It would also enable everyone to help in other ways, like delivering food for the elderly.