Forecasting the COVID-19 Surge Peak in Whatcom County

By On

Without fast action and unified leadership to acquire necessary resources for Whatcom County’s fight against Covid-19, the disease will surge to a devastating peak by April 25th and overwhelm our local health care systems.

The preceding recommendation is based on a “surge peak” forecast of the likely number of confirmed COVID-19 cases for Whatcom County that I developed using data on daily “confirmed” cases from the COVID-19 website set up at Johns Hopkins University.

With these data, I constructed a simple, extrapolative geometric model that is used to forecast confirmed cumulative and new COVID-19 cases by day to April 25th. The forecast shows an expected peak with 6,151 total confirmed cases. Whatcom County’s population is approximately 225,000. If this forecast comes to pass, nearly 3 percent of Whatcom County’s population will have been confirmed as being infected by April 25th. The forecast goes out to April 25th because that is approximately when this initial surge will likely peak.

The forecast reflects the social distancing and other conditions that were in place during the period, March 24th (48 confirmed cases) to 29th (116 confirmed cases). Keep in mind that this period partially incorporates the outbreak at the Shuksan Healthcare Center, which was announced on March 22nd, but its effects have yet to fully play out.

The graph shows the forecast, which includes both the cumulative and daily number of new cases. The model used to create the forecast is based on the calculated rate of change found using the number of cases reported on March 24th (48) and 29th (116), which is 1.1584, where 1.1584 = (116/48) (1/6). The forecast is done by multiplying the most recent reported number of cases (116 on March 29th) by the rate of change (1.1584) taken to the power of the number of days since the most reported cases. For example, the forecast for April 1st is 180 = 116*[(1.1584)(3)] and the forecast for April 25th is 6,151 = 116*[(1.1584)(27)]. The daily number of cases is found by calculating the difference between the forecasted total cases each day. For example, the daily number of cases expected on April 25th is 841, which is found by subtracting the forecasted total on April 24th (5,310) from the forecasted total on April 25th (6,151).

Unless the containment measures put into effect by Governor Inslee and others are able to substantially reduce the expected number of COVID-19 cases, our forecast suggests that Whatcom County’s health care facilities will be in turmoil by April 25th if 841 new cases come in that day to add to the 5,310 already confirmed as infected.

Clearly, fast action and resources are needed to avoid the severe triaging taking place now in France and Italy whereby the infected elderly and others are denied treatment. Perhaps even tighter containment measures should be enacted aimed at “flattening the curve” shown in the graph. They may be needed to give our health care system the critical time it needs to prepare for this unprecedented possibility.

About David A. Swanson

Posting Citizen Journalist • Member since Mar 31, 2020

David A. Swanson is Edward A. Dickson Emeritus Professor of Sociology, University of California Riverside. He served as a member of the U. S. Census Bureau’s Scientific Advisory Committee [...]

Comments by Readers

Steve Abell

Mar 30, 2020

If the goal of the study is to predict the demand on Whatcom County health facilities it would be necessary to adjust the data in two (or maybe more) ways. First, not all new COVID cases will require hospitalization. The number I have seen mentioned most often is 80% of new cases will not. That will reduce the load on health facilties. Second, during the April time frame shown on the graph some patients will be admitted and then later released from medical facilities to finish recovery at home. This will free up beds. Also, some patients will, tragically, not survive. This will also free up beds. So the curve showing cumulative cases will greatly overstate the cumulative load on health facilities. 

Read More...

Wynne Lee

Mar 31, 2020

modeled predictions of covid-related outcomes are currently bound to be ill-informed. If you want to know why, please read fivethirtyeight explains poor predictability. As someone who spent many years studying a complex behavior with integrated experimental and modeling methods, I think their analysis is accurate.
It’s totally understandable, from emotional and practical perspectives, that we lust for an accurate crystal ball or two right now, but that just ain’t what we got right now. Best ideas? Stay as safe as we can, support medical and other providers as we can, ditto small businesses supplying us with what they can, be kind and supportive of our neighbors - widely defined - when we can.

Read More...

Ryan Knowlton

Mar 31, 2020

Copying this from my Dr. Lin’s Letter Post/Comments:

“As a lifelong resident of Bellingham, I remember having TWO hospitals to choose from as well. I am going to save the finger pointing and focus on the problem at hand. Despite the large amount of growth in the area, we now only have one hospital which is listed as having only 241 total beds. To put some #‘s out there, New York is woefully under-prepared for this pandemic, having only 53,000 hospital beds for the 21 million plus people in the greater New York area. This is equal to coverage for only 1/4 of one percent of the population. Sadly, 1/4% of our county population would equal around 625 beds, and we don’t even have half of that. Lets say that we practice good social distancing and therefore limit the spread of the virus to 10% of the population, and 10% of those end up needing to be hospitalized (better numbers than we are seeing in areas where this is more advanced). That is still around 2,500 people which will require hospitalizaiton, and even if we were to spread that out over a 3 month period it still puts our hospital at over 3X capacity. It is time for Whatcom county leadership to prepare for this and have a contingency plan in place for having additional hospital beds ready to go. With schools closed, they could be used for additional hospital space, and even have the cafeterias to prepare and serve food to patients and staff that could end up quarantined there.

I would hope that we fare better and this doesn’t become necessary, but we are far better off preparing for the worst and not needing it, than the opposite.”

 

Added thoughts:

I think we are doing better than average since the stay at home order was put into effect, but unfortunately and sadly the virus is loose in many of the nursing homes. As a result we’re likely to see a higher hospitalization rate and death rate yet lower infection rate of around 3% of the county population. Take 3% of the county population and figure 1/3rd of them will require hospitalization and we’re back around the 2,500 number again.  

Stay home and stay safe. Utilize grocery pickup and delviery if possible instead of going into the stores. 

 

Read More...

David A. Swanson

Apr 01, 2020

Hi, Steve Abell. I disagree. Let me lay out why. Underpinning my argument is the baseline forecast you are commenting on, which I will be updating as more data come in. Because we should see the effects of containment measures, I expect that the updates will indicate a lower demand for beds in terms of the confirmed cases they predict, which suggests that your point may be correct. However, as you will see in my conclusion, there is a disconcerting gap of unknown width between the number of confirmed cases and the actual number of infections and it may be that both of us are under-estimating the number of beds (and staff and resources) that will be needed as of April 25th.

According to Wikipedia, Whatcom County has 253 hospital beds (https://en.wikipedia.org/wiki/List_of_hospitals_in_Washington_(state) This number matches a report from St. Joseph (https://www.peacehealth.org/st-joseph/community-health-and-wellness/news-items/Pages/fast-facts-about-stjosephs-hospital).

The average hospital stay for admissions for pneumonia with or without complications or comorbidities is 3.2 days. Patients admitted for respiratory conditions which require a ventilator are often much longer; patients who receive 96 hours of ventilator support or more are admitted for 22.6 days on average and those with less than 96 hours of ventilator support spend 5.8 days in the hospital on average. (https://www.healthsystemtracker.org/brief/potential-costs-of-coronavirus-treatment-for-people-with-employer-coverage/). Given this, a reasonable assumption is that once admitted a COVI-19 patient can be expected to stay about four days.

The baseline forecast shows 541 new cases on April 22nd, 627 new cases on April 23rd, 726 on April 24th and 841 on April 25th, for a total of 2,735 expected COVID-19 patients over the three day period ending on April 25th

Here are some estimates of hospitalization rates of COVID-19 cases.  “…While 11.8% of infected people in their 60s were estimated to have symptoms severe enough to need to be hospitalized, 16.6% of those in their 70s did. By age 80, 18.4% needed to be hospitalized.Least likely to need hospitalization were 10-to 19-year-olds, with a 0.04% rate. Just 1% of people in their 20s and 3.4% who are 30-39 needed hospitalization. At ages 40-49, hospitalization rates were 4.3%, but nearly doubled, to 8.2%, for 50- to 59-year-olds. (https://www.webmd.com/lung/news/20200331/covid-19-death-rate-drops-still-deadly-to-seniors)

In terms of the age structure of the population of Whatcom (which includes WWU and other college students whose usual residence is elsewhere), the Washington Office of Financial Management estimates that the 2019 population of Whatcom County is 225,300, of which 24 percent are aged 0-19, 57% aged 20-64, and 19% aged 65 and over. Applying these proportions to the 2,735 cases, we have approximately 656 cases aged 0-19, 1,559, aged 20-64 and the remainder, 520, aged 65 years and over.  Approximately 0.4 percent of those aged 0-19 yields 3 hospitalizations; 17 percent of those aged 65 and over yields 88 hospitalizations; and 5 percent of those 20-64 yields 78  hospitalizations for a total of 13. Thus, the estimated total of hospitalizations from these 2,735 COVID-19 cases is 169

Finally, what percent of hospital beds in Whatcom County are being used in normal times? In Washington State, approximately 71 percent of hospital beds in urban areas are occupied during normal times (https://www.ahd.com/news/HFM_DataTrends_2018_July.pdf). The 253 beds in Whatcom County are at St. Joseph’s in Bellingham and, therefore considered urban.

My conclusion: About 183 of Whatcom County’s hospital beds would be occupied in the absence of the COVID-19 pandemic, leaving 70 open for the 169 expected COVID-19 cases requiring hospitalization as of April 25th.    Even this conservative estimate (based on confirmed cases instead of actual infections) suggests that the capacity of Whatcom County hospital beds needs to be expanded by at least 100 beds along with the staffing and resources they require. If the confirmed cases are, for example, only 50 percent of the infected population, Whatcom County would need 200 additional beds under the baseline forecast. If it is 25 percent, then 400, and if only 10 percent of the infected population is confirmed, the county would need 1,000 additional beds.

Hi, Wynne Lee. Per the heavy assumptions and thin data used in my reply to Steve Abell, you would be correct in concluding that I agree with you that predictions of COVID-10 outcomes are currently bound to be ill-informed. However, to do nothing about trying to gain even a limited glimpse of the future is equivalent to assuming that it will be the same as it is now. With this pandemic, this assumption is the worst one we could make.  As such, I believe even ill-informed attempts are preferable as long as they are transparent and driven by avaialble data and reasonable assumptions. As George P. Box, the noted statistician, once observed,” all models are wrong, but some are useful.” In this case, I believe any reasonable model is more useful than the “model” that assumes the future will be the same as it is now.

Hi, Ryan Knowlton. As you can see from my reply to Steve Abell’s comments, I believe my (data thin and assumption heavy) estimate of the additional number of needed beds is likely to be low.  Given this, my wife and I are following your recommendation, namely, “Stay home and stay safe.”

Read More...

Dick Conoboy

Apr 01, 2020

 David,

Thanks for your article an excellent explanations.  Not sure if it is relevant but the student population in the city ofBellingham is indeterminantly  reduced at the moment.  The 4000 dorm rooms are not totally empty as some students have been allowed to stay but that practice is not encouraged and all courses are now given on line as of 6 April.  Unknown is how many of the 8-10 thousand student renters in town have remained through spring break or have returned to their rentals from their vacation destinations (possibly bringing back to town more infections).

Read More...

David A. Swanson

Apr 01, 2020

hi Dick,

Thanks for the note. The remaining students who otherwise would be elsewhere are relevant. Another component of the “gap” between what is happening and what we know now,  not to say what we know about the future.

One of my colleagues observed that forecasting is like driving a car by looking in the rear view mirror because the windshield is caked with mud. Trying to forecast any aspect of this pandemic is like driving not only with a mud-caked windshield, but also with mud splattered  on the review mirror.

 

Read More...

Dick Conoboy

Apr 01, 2020

 David,

We have stumbled upon statistical dark matter. :-)

 

Read More...

Steve Abell

Apr 01, 2020

David, when I started to read your reply to me I was unsure of what you were disagreeing with. My point, which I was making in a qualitative sense, was that a graph of daily and cumulative cases would grossly overstate a graph of daily and cumulative hospitalizations. I don’t think you are disagreeing with that. I was not claiming, nor do I believe, that current Whatcom County health facilities can accommodate expected incoming hospitalizations during this crisis. I suggested that further adjustment of your data would be needed to assess to what degree our current facilities would fall short.

In your reply you provided a very nice quantitative way to adjust the data. You took the prediction of 2735 cases, distributed them across three age groups, calculated hospitalizations for each group, and came up with a total of 169. That confirms the point I was trying to make. We don’t disagree. It would be possible to take your adjustment scheme and apply it to all the data points on your original graph. A horizontal line on this graph could represent facility capacity and the shortfall would be clearly visible. It might be possible to take into account the freeing up of beds due to discharges and deaths but qualitatively that should improve availability of beds over time.

I also agree with your reply to Dick. There are a lot of moving parts here and this makes forecasting an imprecise exercise.

Read More...