Forecasting COVID-19 Hospital Resource Use and Deaths

A study by the Institute for Health Metrics and Evaluation was published in MedRxiv (pronounced “Med Archive”) and forecasts the extent and timing of deaths and excess demand for hospital services due to COVID-19 in the US. Its forecast model uses data from the World Health Organization and local and national governments to estimate deaths and hospital resource utilization over the next 4 months. Hospital resource utilization is stratified by all hospital beds, ICU beds, and invasive ventilators. At a national level, resource utilization and deaths are forecasted to peak on April 16, 2020.

The projections can also be filtered down to a state level where forecasted hospital resource utilization is additionally displayed against capacity. The state-specific forecasts are displayed alongside state-mandated social distancing measures including stay at home orders, closures of schools and non-essential services, and travel restrictions. The models are also updated as new information becomes available. A FAQ page provides many additional insights into the research. Hopefully hospitals and health systems will use these forecasts to close or mitigate gaps in hospital capacity since peak volumes are only about 2 weeks away. Are the hospitals in your state prepared?

Feverish Illness - A Barometer of Social Distancing Effectiveness?

Kinsa is a company that produces internet-connected thermometers, and they have created a map of the United States that overlays 3 data sets (one of which is pictured below):

• Cumulative Atypical Illness: displays how much influenza-like illness above the normal expected levels have been detected
• 7-Day Trend in Illness: 7-day rolling average that shows whether illness has been increasing or decreasing
• Observed Illness: an index of how severely a population is being affected by influenza-like illness

As of March 24, 2020, they report that feverish illnesses were declining in many regions but were also clear in stating that this does not mean COVID-19 cases are declining. Obviously, actual and confirmed cases are expected to lag behind interventions such as social distancing. A related article provides more information about trends in the data. Positive news perhaps? Hopefully so. In any case, social distancing requires adherence and persistence to work so let’s keep it up America!

CDC Coronavirus Self-Checker

I previously wrote about Apple’s COVID-19 screening tool which was developed in partnership with The Centers for Disease Control and Prevention (CDC). Around the same time, the CDC released its own Coronavirus Self-Checker which serves a purpose that is similar to Apple’s screening tool. Similar to Apple’s screening tool, the CDC Coronavirus Self-Checker asks questions about symptoms and sick contacts. However, there are a few differences.

First, the CDC tool is only available via the web, while the Apple screening tool is available via web and an iOS app. The CDC does offer a CDC iOS app, but the Coronavirus Self-Checker is not available in the CDC app (at least as of today).

Second, there are usability issues in the CDC self-checker. For example, one question that is asked is, “Do you/they have any of the following? (check any)” where there are response options for fever, shortness of breath, cough, and other. However, all of those are affirmative responses. There is no option for “none of the above” or a negative response, nor does the self-checker allow you to continue to the next question unless you select one of the affirmative responses.

Third, the CDC app is ostensibly run by a bot named Clara, while the Apple screening tool is less pretentious and gets straight to the point of asking questions and providing recommendations. Sorry Clara, it’s nothing against you personally, but your “presence” contributes nothing to my user experience and feels unnecessary.

So in summary, I think Apple’s implementation of the COVID-19 screening tool is better in many ways than the CDC coronavirus self-checker, but they are both free, and it’s a great public service that they developed and made available for general consumption. Regardless of which one people use, I think the main point is that they are used and can help improve the triage of patients and potentially enable healthcare professionals to focus their time and efforts on the sickest patients who are in greatest need of help.

COVID-19 vs. Humans: Who is Winning?

An interesting data visualization by Aatish Bhatia plots the number of new confirmed COVID-19 cases in the past week vs. total confirmed COVID-19 cases, both on a logarithmic scale (which can also be toggled to a linear scale). While there is no time axis, the chart can be animated over time. The visualization is based on publicly available data from Johns Hopkins University which is the same data set that is used for a dashboard that I recently wrote about.

The somewhat unique aspect of plotting new confirmed cases vs. total confirmed cases is that it provides a visual indicator of whether we are continuing to experience exponential growth of COVID-19 cases or if we are starting to win the war against coronavirus. The following video provide an excellent and understandable explanation of the math behind this thinking:

Kudos to Johns Hopkins University not only for hosting their dashboard for public use but also for making their data publicly available so that it can be used for other purposes like this. And of course, kudos to Aatish Bhatia for helping us see the same data in a different light to help us understand if we are winning the war against coronavirus. SPOILER ALERT: currently we are not.

Crowdsourcing Data for COVID-19 Research Studies

If you’re interested in contributing to our world’s understanding of COVID-19, consider downloading the COVID Symptom Tracker which is available on iOS and Android platforms.

Note that this app is NOT designed to give you advice related to COVID-19. It is a data collection research tool that was created by the Harvard T.H. Chan School of Public Health, Massachusetts General Hospital, King's College London and Stanford University School of Medicine, working in partnership with ZOE Global Ltd. It is intended to support studies like the Nurses’ Health Study that was established in 1976 and is longitudinally tracking health and lifestyle factors on outcomes in women. The app can also be used to support other studies across the United States. A FAQ provides more general information about the app and its purpose.

After downloading and answering a few questions related to your current symptoms (if any), you’ll be asked to try to self-report on a daily basis. This is a great example of how technology solutions can be used to scale research efforts. On a related topic, consider reading this viewpoint on how to preserve clinical trial integrity for COVID-19 research.

A Statistical Model for Flattening the Curve

The folks at SURF Stanford Medicine have made available a COVID-19 calculator that allows healthcare providers and policy makers to estimate ICU and acute care bed demand for COVID-19 patients. Why is this important? I’ve previously addressed social distancing which is one of our most effective methods to flatten the curve at present, given that a vaccine is still under development and we still have a huge shortfall in diagnostic testing capacity. Flattening the curve allows our health systems to “buy time” and avoid over-saturating our ability to deliver care to patients who need it. Exceeding our health system capacity will mean that doctors will be forced to ration care (i.e., treat the ones who are most likely to survive and let others potentially die).

The calculator includes a variety of inputs that enable users to more accurate forecast the impact to their local area hospitals. For example, one can input the cumulative confirmed cases as of today, and this data is available from a variety of public sources. Another input is the doubling time which is the number of days it takes, at the current trajectory, for the number of cases (or hospitalizations, if you choose to use that as the basis for your model) to double. There are also variables for symptomatic cases per confirmed case and number of days to model ahead. These all impact the steepness of the curve.

I think that the most interesting aspect of this calculator is the ability to simulate the impact of social distancing interventions. The calculator provides the ability to enter new doubling times on 3 different days, and each new doubling time is modeled into the shape of the curve.

This model reflects a population that does a poor job of social distancing.

This model reflects a population that practices effective social distancing.

While this calculator is intended for use by healthcare providers and policy makers, I think it would also be useful for the general public to use this tool to visualize the potential impact of social distancing on our healthcare system. If we can all shelter at home, cover our coughs and sneezes, disinfect surfaces as needed, and take all the precautions, we can reduce transmission of COVID-19. Reduction of transmission will result in higher doubling times, and the general public can see the impact of social distancing on flattening these curves.

Technology Consequences of the COVID-19 Pandemic

Given that the world is now sheltering at home to reduce transmission of COVID-19, we have become more reliant on technology to get us through our daily rituals.

Video Streaming

Given that schools around the nation have temporarily closed, video streaming has become more popular than ever before. Many working parents like me have probably thrown screen time recommendations out the window to let our mobile devices babysit our children while we work. Therefore, it was not surprising that when I tried to launch a movie on Disney+ on my Apple TV the other day, I was greeted with the following message:

It is well known that Disney+ ran into bandwidth issues with its November 2019 launch. This week, Disney+ reduced bandwidth for its Europe launch in an effort to reduce strain on infrastructure. I suspect that similar supply-demand issues are causing connection problems nowadays. Did you know that you can check Downdetector to see if others are having the same problem?

Video Conferencing

Many workplaces have transitioned to remote work where possible, and this has driven a tremendous increase in demand for video conferencing. Zoom has capitalized on this opportunity, but it has not yet overtaken some of the incumbents in the video conferencing space. We use both Microsoft Teams and Cisco Webex at my workplace, where Teams had traditionally been used more for internal collaboration, and Webex had been used for external calls. In retrospect, it is fortunate that we have 2 video conferencing services. The day after Microsoft aired a Super Bowl commercial for Teams, most of us were unable to connect to Teams meetings and switched on the fly to Webex. This past week, many of us were able to connect to Webex but unable to get the computer audio to work, so we switched on the fly to Teams. I don’t know the root cause of these issues, but it would not surprise me if their connections were overloaded due to a surge in demand.

Mobile Communication

Finally, with businesses transitioning to remote work, mobile device usage has been rising. Wireless service providers are recognizing the need for increased communication during the pandemic, and I was pleasantly surprised to receive the following notification from Verizon Wireless:

What other ways have you leveraged technology to get through your day, and what kinds of unintended consequences have you witnessed? Regardless, I hope you stay safe, productive, and connected (at a distance, of course).

COVID-19 Public Safety Alerts and #STFH

My iPhone was placed on my desk about 1 foot away from me when I was startled by a piercing alarm accompanied an emergency alert:

I received this message because I enabled notifications on my iPhone. There are actually 3 kinds of government alerts that you can elect to receive: AMBER alerts, emergency alerts, and public safety alerts. I guess the above alert falls into 2 of the 3 categories. More information about these alerts can be found on the Apple support page.

About 2 minutes later, I received a text message and email with the same message:

The text message and email were generated by NotifyLA, a service of the City of Los Angeles Emergency Management Department. I have received numerous text messages and emails from NotifyLA about wildfires and COVID-19. If you live in a major metropolitan area, chances are that there will be similar services available to you.

On the plus side, I think it’s terrific that we have so many ways of rapidly disseminating information to a large number of people. On the other hand, I continue to see social media posts that many people are not adhering to social distancing measures, and I have witnessed this first hand while running “essential” errands (e.g., getting food) around Los Angeles. So overall, it’s too bad that we have to use this technology. Let’s get it together folks and stay the f*ck home.

Should I See A Doctor About COVID-19?

Now that COVID-19 is gaining steam and the United States is now the country with the most confirmed coronavirus cases in the world, we are also starting to see the emergence of online screening tools.



One of these is the COVID-19 Screening Tool which was developed by Apple in partnership with The Centers for Disease Control and Prevention (CDC). A web version and iOS (mobile) version are available as of today. Users answer questions about symptoms, travel, and contact with others.

Low-risk responses result in a recommendation to practice social distancing. Other responses that reveal risk factors for developing COVID-19 result in a recommendation to talk to a medical professional about getting tested for COVID-19. Residents of long-term care facilities are instructed to reach out to the doctors in charge of their facilities. Finally, people who have worked in a hospital or other care facility in the last 14 days are instructed to contact the occupational health provider where they work.

Of note, the tool is intended for people who are at least 18 years old. Younger people are directed to the CDC web page for Children and COVID-19.

Tracking COVID-19 Data

Some of my recent blog posts have been about COVID-19 data visualization. This post is about the data itself, and I wanted to highlight work being done by The COVID Tracking Project which provides several interesting features.



The first is a page where you can access the most recent data about positive and negative tests, pending test results, and deaths. In addition to a summary statistic for the United States, you can view the numbers that are broken down by state, along with any caveats about how they are collecting and reporting data for each state.

The second is a live spreadsheet on Google Sheets which lets you see the raw numbers. If you’re a geek like me, you might download the data, play with it in Excel, and perhaps try to make some interesting charts (hey, I couldn’t resist).

Lastly, there is support for a basic API that currently supports a GET request that returns either data or a .csv file. Since I’m not a developer, I haven’t taken advantage of the API, but you can simply go to the page and download the latest CSV files if you want.

This is an important topic because one of the main gaps in our current knowledge of COVID-19 is the number of people who are infected. Their methodology and FAQs provide information about where the data come from, and it appears that they are doing their best to provide reliable data and to provide transparency about its limitations. Kudos to their entire team of contributors.

More COVID-19 Data Visualizations

If you’re trying to track the spread of COVID-19 in near-real time, perhaps you found my recent post about Dashboard Visualization of the COVID-19 Pandemic to be useful. Since then, I’ve found a few more visualizations that I’d also like to share.

The first is a U.S. Coronavirus Tracking Map that shows confirmed COVID-10 cases and deaths, based on data from federal, state and local officials.



The second is a collection of statistics and visualizations from Worldometer which is run by an international team of developers, researchers, and volunteers with the goal of making world statistics available in a thought-provoking and time relevant format to a wide audience around the world.



Finally, Our World in Data provides COVID-19 statistics based on data from the World Health Organization’s situation reports. One interesting chart allows you to see the number of cases over time and compare them across countries. It also lists doubling times and other interesting information.



Remember that most data and visualizations related to new cases reflect the time of detection rather time of actual infection. This is due to possible delays in testing which has occurred in the U.S. With that limitation in mind, I hope you’ll find these maps and statistics to be useful.  Above all, stay safe and remain calm. We’ll eventually get through this!

Mapping School Closures Due to COVID-19

If you’re a parent of a school-aged child like me, you might be impacted by school closures due to COVID-19. The folks at Education Week are collecting information about school closures in the United States and visualizing the data on a map.



If you’re a geek like me, you’ll be happy to find that the data are also presented in tabular format and made available for download. For a full-page version of the map, click here.

COVID-19 and Social Distancing Analogies

Normally I write about geeky technology topics, but since we are in the midst of a pandemic, I thought I’d deviate from the theme of this blog to share a few analogies related to social distancing which is one of our most effective ways of limiting the spread of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the virus that causes COVID-19.

The CDC defines social distancing as “remaining out of congregate settings, avoiding mass gatherings, and maintaining distance (approximately 6 feet or 2 meters) from others when possible.” Given that a vaccine is still under development and possibly still a year away, social distancing is one of our best bets for limiting viral transmissions (in addition to personal hygiene, isolation, quarantine, and other measures). This article and this article address the concept of “flattening the curve” of the number of cases over time, based on how the cities of Philadelphia, PA and St. Louis, MO responded to the 1918 flu pandemic. While there are several excellent publications on the topic, I’d also recommend reading this article, and this article too.

I’ve shared these publications with many friends and family, and while many have embraced the concept of social distancing and are making significant changes in their daily routines, I feel that some people still do not completely understand its importance or have not yet translated their understanding into making immediate lifestyle changes. Therefore, I’d like to offer 3 analogies in hopes that 1 or more of them help to drive home the point.

Analogy #1: Herd Immunity

Immunizations are given to prevent the spread of infectious diseases for both individuals and populations. While most people understand the benefit of immunizations for themselves (e.g., prevention of influenza, measles, varicella, polio, and many other infectious diseases), it is often necessary to remind people that if a sufficient portion of the population is immunized, it confers protection to those who those who are unable to receive immunizations, often due to a compromised immune system. This phenomenon is known as herd immunity.

When an individual decides not to immunize, it weakens the effect of herd immunity. Therefore, I view it as a social and moral responsibility to get immunizations (and when a SARS-CoV-2 becomes available, we should all get that one too, provided that we don’t have any contraindications). Similarly, I feel that everyone has a social and moral responsibility to practice social distancing. Yet, I continue to hear about people moving forward with previously planned non-essential social gatherings. This article provides one such example, but there are less dramatic examples such as family gatherings with elderly individuals or people with high-risk conditions. The prioritization of one’s individual liberties and personal freedom over that of the public good mitigates the effectiveness of social distancing and weakens our “herd immunity” to COVID-19 which means that more people will get infected, and more people will die.

Analogy #2: The Lottery

If you take the perspective of SARS-CoV-2, the propagation of your genetic heritage is dependent on having as many opportunities as possible to spread to other hosts. In epidemiology, there is a statistic called the “reproduction number” or R0 (pronounced “R naught”) which represents the number of new cases that are infected by one infected individual. Earlier this year, the R0 was estimated in this article, this article, and this article to be somewhere between 2 and 3 (plus or minus), which means that for each person with COVID-19, 2 to 3 additional people had been infected on average. This is a recipe for sustained community transmission.

The lottery analogy comes into play when you look at each opportunity for the virus to shed from one individual and infect another host. On a typical day, most people interact with many individuals at work, running errands, in social gatherings, etc. Most people also touch many things that other people touch such as doorknobs, coffee makers, elevator buttons, money, etc. Some people might also cough or sneeze into the air or perhaps be in close enough proximity to breathe directly into someone else’s personal space. All of these actions are opportunities for a virus to spread to another host. Any one particular action is a low-probability event for transmission, somewhat akin to the virus buying a lottery ticket. However, with enough opportunities, SARS-CoV-2 within a host will win the proverbial lottery 2 to 3 times and jump to additional hosts. Social distancing basically deprives SARS-CoV-2 of lottery tickets and helps drive the R0 to less than 1 so that viral transmissions eventually peter out.

Analogy #3: Game Theory

Game theory is the study of mathematical probabilities and decision-making between individuals or groups. It was initially applied to economic scenarios and extended to military decision-making and other fields. It is perhaps best illustrated through games like chicken and poker, and for a classic example of an application of game theory, read this description of the prisoner’s dilemma.

Let’s imagine a game theory scenario for social distancing. In this case, we have multiple participants consisting of everyone in a community, state, country, or even the entire world for that matter. The number of participants doesn’t really matter, but what does matter is when we look at 1 person who is faced with a choice to ignore or follow recommendations to practice social distancing. If that individual ignores (or is simply unaware of) recommendations to avoid crowds and maintain distance from others, then that individual may achieve some gain (e.g., connectedness with other humans, traveling somewhere new, enjoying a movie) at the expense of giving SARS-CoV-2 more opportunities to spread which translates to a public disadvantage.

Summary

I am certainly not suggesting that we cut off all ties to other human beings in the midst of the COVID-19 pandemic. There are also certain things that we must still do out of necessity such as going to work (if remote work is not an option), shopping for groceries, taking care of friends and loved ones who need assistance, etc. I am also not suggesting that we blame people who take less extensive precautions than we do. We are social animals and thrive on human interactions. We also have our own tolerances for change in different scenarios. Therefore, social distancing is not a binary measure. There are degrees to which we can practice social distancing, and the more strictly we adhere to it, the more we will limit the spread of COVID-19 and save lives. It all boils down to a numbers game. How effectively will we play the game? Time will tell. Be safe everyone!

Dashboard Visualization of the COVID-19 Pandemic

Coronavirus disease 2019 (COVID-19) is a respiratory illness that is caused by a virus named severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). A summary of the disease along with recommendations for diagnosis and prevention is provided by the Centers for Disease Control and Prevention (CDC) here.

The spread of COVID-19 has been progressing at an alarming rate. In January 2020, the World Health Organization (WHO) declared the outbreak a public health emergency of international concern (PHEIC) which means that it constitutes a public health risk through the international spread of disease, and it potentially requires a coordinated international response. Today the WHO declared COVID-19 a pandemic which the CDC and these authors define as an epidemic that spreads over several countries or continents. In short, COVID-19 is out of control, people are worried, and we need to act quickly to minimize further illness, death, and the downstream impact on our local communities, our country, and the rest of the world.

There are many aspects of COVID-19 diagnosis, treatment, and prevention that demand our attention and resources. One fundamental need during any infectious disease outbreak is to track cases over time. Peter Drucker is credited with saying, "you can't manage what you can't measure," and the principle definitely applies here.

Thanks to an engineering team at Johns Hopkins University, it is now possible for researchers, public health authorities, and the general public to access to an online interactive dashboard that can be used to visualize and track COVID-19 cases in near-real time. Here's a screenshot, and if you visit the dashboard, I think you'll find it fairly self-explanatory.

Initially the dashboard was manually updated twice a day, but the team has since transitioned to a semi-automated process that leverages a variety of data sources. A brief overview of methods are described in this article and this blog. The following figure illustrates that the Johns Hopkins dashboard statistics align well with figures provided by the Chinese CDC and WHO.



Hopefully we will start to see a decline in the number of new cases over time. To achieve that, we should all follow recommendations from the CDC including its Workplace, School and Home Guidance and its Framework for Mitigation. Be safe everyone!

April 24, 2020 Update: Here is another link to the Johns Hopkins University dashboard that contains some more features: https://coronavirus.jhu.edu/map.html