Question: What projections do we have for covid/Delta variant surge for the DC area? How far ahead can we usefully predict covid activity for any particular area?
Answer: We have forecasts and then we have planning scenarios. I’ll briefly describe both.
Forecasts are like your 14-day weather report; likely pretty good estimates of what’s to come within a short window of time. For COVID, our forecasts generally project no more than 4 weeks into the future. CDC updates its forecasts on a weekly basis, creating an ensemble forecast built on forecasts of 26 different modeling groups. CDC provides these forecasts on cases, hospitalizations, and deaths, at the national level, as well as state and county levels. For DC, the most recent forecast of new weekly cases (as of July 19, Figure 1) shows a range of possibilities across the 17 modeling groups included (left panel), with the ensemble forecast estimating that weekly cases will plateau around 200/week for the next several weeks before slightly declining in mid-August (right panel, red dotted line). As you can see in the right panel, the prediction intervals (similar to confidence intervals) are quite wide (darker red highlight= 50% prediction interval; pink highlight= 90% prediction interval) and the farther out in time we go, the wider the prediction intervals become. As these prediction intervals indicate, while our current best guess is that cases in DC plateau and then slightly fall, it’s also within the realm of possibilities that they continue to substantially rise. Additional forecast data are available at the COVID-19 Forecast Hub (which CDC uses), including more recent state-level forecasts (as of July 27). As of July 27th, the ensemble forecast for DC “predicts that weekly totals of observed cases in each of the next four weeks will be between 280 and 310 cases each week. However for the week ending August 21, the ensemble shows substantial uncertainty, with observed cases between 100 and 900 deemed possible (95% prediction interval: 65 – 940).”
Planning scenarios are longer-term (6+ months) estimates based on different potential scenarios. I know I’m stating the obvious, but planning scenarios are generally used to help policy and program managers plan for the future. To develop “worst-case”, “best-case” and “middle-of-the-road” scenarios (among others), modelers generally make multiple different scenarios based on multiple different assumptions. For COVID, modelers would generally make scenarios based on vaccine coverage, vaccine efficacy, virulence of new variants, and scope of use of non-pharmaceutical interventions (mask wearing, social distancing, etc.) (Table 1). The COVID Scenario Modeling Hub takes an ensemble approach to modeling such scenarios, making estimates based on scenarios modeled by multiple different groups. For DC, the ensemble scenario estimates that weekly cases will remain low for all four scenarios (Figure 2), but the prediction interval is quite wide, and is even wider under high variant transmissibility scenarios (right panel), which seem like the likely scenarios given Delta’s spread.
Finally, a periodic reminder from me that models– for forecasts and planning scenarios– are only as good as their underlying data and the assumptions the modelers make, and models are going to vary even further based on the analytic methods the modelers employ. Please remember, models are not crystal balls, they are tools designed to help us to understand the range of possibilities and plan accordingly!
Figure 1. Forecast of Weekly Cases in DC (mid-July through mid-August) (from CDC, July 19 update)
Table 1. Four Scenarios used by the COVID Scenario Modeling Hub for the July/Round 7 Projections (from COVID Scenario Modeling Hub)
Figure 2. Ensemble Scenario Projections of new weekly cases for DC (July through December 2021), based on 4 scenarios (from COVID Scenario Modeling Hub)
