covid-19 current state analysis and forecasting for the ... · 13 hospital beds occupied by...

COVID-19 Current State Analysis and Forecasting for the DFW Region

Department of Internal MedicineLyda Hill Department of Bioinformatics Department of Emergency MedicineUTSW Health System Information ResourcesData as of July 20

UPDATED JULY 21, 2020

About the Model

2

The following slides illustrate a model of how COVID-19 is spreading across the DFW region based on real patient data we have received from Collin, Dallas, Denton and Tarrant counties. The slides capture a snapshot based on data available as of July 20. Every time we receive new data, we re-run the model and refine the graphs.

In the following slides we examine how well preventive measures including masking, staying at home, social distancing, hand hygiene and others have limited the spread of infection, and what might happen looking forward.

Model-building is an iterative process with inherent uncertainty in its predictions. It is intended to provide help in formulating plans but should not be the sole basis for policies or management decisions addressing the COVID-19 challenge. Please do check back every few weeks to see how well the DFW region is doing to keep COVID-19 at bay.

We would like to thank the local health departments, hospitals, and health systems that have contributed data to help us build this model.

Commentary

3

In both Dallas and Tarrant Counties, hospitalizations remain high but have flattened this week. However, sustained high hospitalization levels are straining regional healthcare resources, especially the workforce. Though the forecast for Dallas and Tarrant County is slightly lower as hospitalizations have decelerated over the last week, the percent of COVID-19 tests coming back positive is still high, suggesting more infection is present in the community than can been seen through testing data alone. In the DFW area, hospitalizations due to COVID-19 have increased by 2% compared to one week ago and by 107% compared to one month ago.

While mobility data through July 13 shows overall declines from the highs of mid-June, activity has increased somewhat over the last week. The mobility activity related to the July 4th holiday weekend appears to be muted and helped mitigate the disease’s impact on hospitalizations. To ensure healthcare capacity remains available, the community’s continued compliance with physical distancing, masking, hand hygiene, and crowd management policies are needed.

It is important to remember that people arriving at the hospital today were likely infected ~2 weeks ago. Our collective actions now are critical to changing the course of the outbreak in North Texas.


Cases of COVID-19 and those that require hospitalization are flattening in North Texas

4

Hospitalizations due to COVID-19 have increased by 2%compared to one week ago and by 107% compared to one

month ago.

Source (left): Dallas County HHS 7/17 report, data as of 7/11Source (right): NCTTRAC EMResource Data, Accessed 7/20/2020“North Texas” is defined as Trauma Service Area E

Confirmed COVID-19 Patients in North Texas Hospitals

27% of COVID-19 tests performed in Dallas County hospitals were positive in

the week ending in July 11.


-

200

400

600

800

1,000

1,200

1,400

1,600

1,800

2,000

26-Apr 10-May 24-May 7-Jun 21-Jun 5-Jul0%

5%

10%

15%

20%

25%

30%

35%

4-Apr 18-Apr 2-May 16-May 30-May 13-Jun 27-Jun 11-Jul

Percent positive COVID-19 tests in Dallas County Hospitals

0

100

200

300

400

500

600

700

800

900

20-Jun 27-Jun 4-Jul 11-Jul 18-Jul

How many people with COVID-19 have been hospitalized in Dallas and Tarrant Counties in the past month?

Source: NCTTRAC COVID EMResource Data Set, Accessed 7/20/20205

We closely track daily COVID -19 hospitalizations, as this indicator reliably monitors the severity of the outbreak in DFW. Hospitalizations trail new infections by 1-2 weeks but are not influenced by testing capacity or test reporting delays, thus giving us a clear picture of severe cases in the community.

Total Hospitalized COVID-19+ Patients in Dallas County Total Hospitalized COVID-19+ Patients in Tarrant County

ICU Non-ICU

In Dallas County, hospitalizations have increased by 88% compared to one month ago and 6% compared

to one week ago.

In Tarrant County, hospitalizations have increased 144% compared to one month ago and

decreased 2% compared to one week ago.


+37% +15% +6 % +6 %

% Week-over-week change

+66% +19% +18% -2%

How quickly is COVID-19 spreading in North Texas?

The Number of Days It Has Taken for COVID-19 Cases to Double

6

As the spread of disease slows, it takes longer for COVID-19 cases to double. The doubling time has decreased (i.e. faster spread) in all 4 counties and remained stable for 4 weeks. This drop corresponds with Memorial Day weekend and may also reflect the increased contact individuals had with each other.

Slower COVID-19 spread

Faster COVID-19 spread

Source: JHU CSSE COVID-19 US Confirmed Case Data, Accessed July 21, data through July 20Methods: Trailing 7-day log-linear regression


Rt Represents Contagiousness

§ Rt helps us measure how effective social distancing measures are after they are put into place.

§ If social distancing and measures like masking are effective, then the number of secondary infections is dramatically reduced.

§ In this scenario where social distancing measures were 50% effective, then only five people end up infected, rather than the original 31.

Source: Dallas and Tarrant County websites, Accessed July 21, data for positive tests with a specimen collection date of 7/7 or earlier1) Cori, A. et al. A new framework and software to estimate time-varying reproduction numbers during epidemics (AJE 2013).2) Assumes serial interval follows gamma distribution as calculated in Nishiura, et al . "Serial interval of novel coronavirus (COVID-19) infections." Int J Infect Dis. 2020 Mar 4;93:284-286. doi: 10.1016/j.ijid.2020.02.060.

8

Epidemic grows/ persists above this line

Epidemic declines below this line

R t

Each line represents how contagious COVID-19 has been in Dallas and Tarrant Counties over the last few weeks. Contagiousness depends on how well we social distance, wear masks, limit travel, clean high touch surfaces, etc.

The Rt value appears to have declined to around 1 in both counties as of two weeks ago. We now calculate this value using the date positive tests were collected rather than the date they were reported, since the tests reported on a given day were collected over many different days. Historical data points are subject to revision.

Mask order put

into place

Memorial Day

Mask order put

into place

Memorial Day

Dallas Tarrant

How contagious is COVID-19 in DFW now?


Are North Texans on the move?

9*The encounter rate difference is the rate of unique human encounters per km2 relative to the national pre-COVID-19 baseline.Source: UnaCast Mobility Data, showing trailing 7-day averages, accessed July 21, data through July 17.

To measure if residents are compliant with physical distancing, we use proxy measures for mobility. Three measures we follow are shown above. The rate at which people’s phones are uniquely encountering each other has remained relatively flat since mid-May and began to decline slightly in late June. Visits to non-essential stores rose sharply throughout May but began to decline again in late June, coinciding with the closure of bars in the state. July 4 activity appears to have been relatively muted. However, non-essential retail visits and contact rates have crept up again in the last week, so we are watching carefully for any potential impact over the coming week.

Non-Essential Retail Visits Encounter Rate Difference* Difference in Distance Traveled

More movement

Less movement


Dallas County 14-day forecast starting from 7/6 compared to actual data

Dallas County – Past Model Accuracy and Future Forecasting

This model assumes the combined impact of nonpharmaceutical interventions (NPIs) is 63%.

§ Dallas County total COVID-19 hospitalizations (blue line) are predicted to increase to 690-1,080 concurrent hospitalized cases by August 3.

§ An increasing number of COVID-19 patients are predicted to be in the ICU (red line).

§ Roughly 1,240 new COVID-19 infections (black line) per day are expected by August 3.

Dallas County 14-day forecast starting from 7/20

§ Our current model predicted the past two weeks of COVID-19 hospitalizations very well.

§ Where the reported hospitalizations (black line) are inside the modeled range (shaded blue region), and the reported ICU cases (gray line) are inside the modeled range (shaded red region) the model accurately predicted hospitalizations and ICU cases, respectively.

Shaded regions represent 90% credible interval.

Expected Daily New Cases and COVID-19-Occupied Hospital Beds


Tarrant County – Past Model Accuracy and Future Forecasting

This model assumes the combined impact of nonpharmaceutical interventions (NPIs) is 63%.

Tarrant County 14-day forecast starting from 7/20

§ Our current model predicted the past two weeks of COVID-19 hospitalizations very well.

§ Where the reported hospitalizations (black line) are inside the modeled range (shaded blue region), and the reported ICU cases (gray line) are inside the modeled range (shaded red region) the model accurately predicted hospitalizations and ICU cases, respectively.

Tarrant County 14-day forecast starting from 7/6 compared to actual data

Shaded regions represent 90% credible interval.

§ Tarrant County total COVID-19 hospitalizations (blue line) are predicted to increase to 560-920 concurrent hospitalized cases by August 3.

§ An increasing number of COVID-19 patients are predicted to be in the ICU (red line).

§ Roughly 660 new COVID-19 infections (black line) per day are expected by August 3.


Expected Daily New Cases and COVID-19-Occupied Hospital Beds

What is the impact of preventative measures on the future spread in Dallas County?

Note: model assumes perfect isolation of hospitalized cases, which has a dampening effect on effective Rt

Actual cases

Actual and Predicted Daily New Infections in Dallas County

63% effective(illustration of current impact)Rt=1.05

68% effectiveRt=0.9

65% effectiveRt=1

Our model is optimized to predict hospitalizations, depicted on the next slide. It is difficult to know the true number of new cases each day until weeks later due to significant delays in the reporting of test results to the county.

There are limits to our ability to precisely determine compliance levels with prevention measures.

61% effectiveRt=1.1

Social distancing & other prevention measures are currently ~63% effective at curbing the spread of COVID-19. The resulting Rt decreased slightly from last week, and these small changes can still lead to significant pressure on cases and potentially healthcare resources.

§ If measures remain 63% effective, the orange line will happen.

§ The other lines illustrate what would happen if measures were 61% effective, 65% effective, or 68% effective.


We can flatten the curve again by returning to our “mid-May” behavior

13

HOSPITAL BEDS

Needed by COVID-19 Patients in

Dallas County

This would happen if we returned to our behaviors from early June.This is our current trajectory, taking all recent policies into account. This includes the Dallas County masking order, which has been in place since June 19.

If we return to our behaviors from mid-May, we can flatten the curve again, and hospitalizations will start to go back down.

Forecast starting from 7/21/2020Source for past occupied beds: NCTTRAC EMResource Data through 7/21/2020

Changes in policy and behavior will not be reflected in hospitalization numbers for 4-14 days given incubation period of COVID-19.


14

This is a snapshot of what our cases counts might have been on May 29th if Dallas County had waited even one week to take action. Every day counts. Our actions today will have a big impact on the coming weeks and months.

Source: JHU CSSE COVID-19 US Confirmed Case Data, Accessed May 29, 2020Methods: Hypothetical scenarios assume that Dallas County policies maintained pre-March 22 trajectory of doubling cases every 2.5 days for 1 week or 2 weeks respectively, then resumed its actual daily growth progression thereafter

2-week delay(217,000 Cases)

1-week delay(47,000 cases)

Actual(9,787 cases)

Preventative measures worked before and will work again

-

50,000

100,000

150,000

200,000

250,000

11-Mar 18-Mar 25-Mar 1-Apr 8-Apr 15-Apr 22-Apr 29-Apr 6-May 13-May 20-May 27-May

Where would we have been on May 29th if we had delayed initial social distancing measures by 1 or 2 weeks?

covid-19 current state analysis and forecasting for the ... · 13 hospital beds occupied by...

Documents