There are several important metrics to understand where a community, state or region is with the spread of the COVID-19:
- The true percentage of the population infected and those who are infectious: I(t).
- The rate of change of the infection as represented current reproduction factor: R(t).
- The percentage of the population who has been exposed to the virus and recovered who have antibodies: Recovered.
Tracking these parameters over time allow can shape intelligent policy decisions.
Deriving Key Parameters
These parameters can be derived from accurate models which have been calibrated to data that provides the strongest correlation: deaths, hospitalization rates and broad sample anti body testing. Consistent and rigorous “molecular” testing for an active infection of individuals who present with symptoms also will provide solid correlation.
A model calibrated to these factors will accurately predict infection rates and disease vectors. This will provide a sound basis for decision making and strategy management for the future.
There are two tools that can provide this insight:
Our COVID Decision Tool can accurately determine where we are and can be calibrated to against existing data for a wide variety of heterogeneous regions and populations. The CDM can be used to assess various recovery scenarios in detail and is an excellent planning tool. It can be used to determine the future velocity of the disease in response to various policy decisions, how things are likely to go and how fast things are likely to move.
Jesús Fernández-Villaverde and Charles I. Jones have constructed a modified SIRD model that accurately predicts these parameters and they have done so for many countries, states and cities: COVID R0(t) Dashboard
These are complimentary approaches. The closed form calibrated model of Fernandez-Villaverde and Jones can quickly predict where you are in time and can be easily updated. The CDM model can be used to model various recovery scenarios and strategies, but is more computationally intensive. Results between the two tools are consistent and both are based on calibration to real data.
Mainstream Metrics are Flawed: Testing and Hospital Capacity
The current CDC gating criteria for moving through the recovery phases is focused on “molecular” virus testing (testing for an active infection) to establish infection trends. The COVID Exit Strategy dashboard presents the commonly accepted scoring methodology and is largely aligned with White House and CDC gate criteria: COVID Exit Strategy Dashboard
Risk of proceeding to next recovery phase are based on key metrics largely derived from the expansion test programs. This is a flawed methodology to predict infection rates, reproduction rates and other general population metrics. A test program must be very disciplined and consistent (removal of bias and other factors) to provide meaningful metrics. That is very difficult to do.
Testing as a Metric is Not Scalable as the Infection Dies Out
Deriving infection parameters from testing alone is a fundamentally flawed approach. As the percentage of infectious population becomes small, the sample size for testing needed to track these low rates becomes very large and impractical to manage. For example, if the current infectious population is 0.1 % that means 1 person in 1000. For a population of 1 Million: we would have to test upwards of ten thousand individuals per week to get a stable look at the infection rate. If the rate dropped to 0.01 % that means 1 person in 10,000. That means 100 people in a population of 1 Million were we would have to test many tens of thousands per week to accurately resolve the infection rate. This is not practical.
Hospital and ICU Capacity
Hospital and ICU Capacity Metrics are tracked as scoring criteria. These gating metrics seem overly conservative, as we have not hit any hospital capacity issues in the US. Certainly, hospital and ICU capacity needs consideration, but would need to be assessed in terms of the actual infection trends and weighed against things like static or baseline hospital capacity metrics. These are secondary consideration and need to be included in the overall planning, but should not be used as primary assessment metrics.
General Population Anti Body Testing
Antibody serological testing is a compliment to this approach and allows for broader verification of model fidelity with respect tracking those who have recovered (the R in the SIRD model). This provides additional confidence in the model fidelity.
I(t) and R(t) calibrated to actual data from deaths, hospitalizations and rigorous symptomatic testing are far superior metrics and based on sound principals. The CDT and the Fernandez-Villaverde Jones model provide and excellent complimentary tool-set to strategic plan and manage a recovery using strong and verifiable metrics.
Derivative concerns, like hospital and ICU capacity can easily be managed and tracked in light of a clear understanding of actual disease core metrics.
Metrics based on broad test programs to track the infection are fundamentally flawed and break down as the infection rate in the general population grows small. These test programs are not well controlled and are not statistically rigorous.