JDweb - DCM-covid project

This is a temporary webpage that summarizes some recent modelling/data analysis work on the current COVID-19 pandemics.

In brief, we are developing a Dynamic Causal Model of the COVID pandemics (Friston et al. 2020a-b-c-d, Moran et al. 2020a-b, Daunizeau et al. 2020a-b). Our main contribution is to account for temporal changes in peoples' psychological attitude towards risks, which is influenced by broadcasted information regarding salient COVID outcomes (e.g., daily death counts), and eventually determines whether or not one will comply with social distanciation and other preventive health-related behaviours. This model however, ignores sporadic viral mutations and/or loss of acquired immunity.

You can find all related scientific papers here: https://www.fil.ion.ucl.ac.uk/spm/covid-19/

We use two data repositories:

- Our World In Data (see this ECDC github repository)

- Santé Publique France (see data.gouv.fr repository I and II)

... and the related data and VBA-compatible code can be downloaded below :)

Below is a graphical summary of analyses performed on French data (see Daunizeau et al 2020-b for more information).

Note: only hospital mortality is reported here (this does not account for, e.g., retirement homes).

One should question the reliability of the data we use in our analysis. Daily death counts, for example, are problematic for at least two reasons. First, different data repositories effectively give different numbers, e.g., people deceased in hospitals (as is the case for the data we present below), or in hospitals plus retirement homes (as for ECDC data). Second, they do not account for “normal” seasonal mortality. Testing procedures also have imperfect sensitivity and specificity, and ICU occupancy actually depends upon heterogeneous clinical criteria (e.g., respiratory support versus reanimation). All these limitations are difficult to account for, and impair the reliability of model-based predictions.

Figure 1: Prediction of main epidemiological outcomes at the national level.

Upper-left panel: daily death rates (y-axis) are plotted against time (x-axis, in weeks), starting on the 1st of January 2020 and ending on the 31st of December.

Black dots show the reported data (hospital data).

The red trace shows the model-predicted data .

Black dotted lines denotes the start and (intended) end of the governmental containment measures (17th of March and 11th of May, respectively).

The black plain line shows the date of data analysis (17th of June 2020).

Upper-right panel: positive test rates (medical laboratory data).

Lower-left panel: remission rates (hospital data).

Lower-right panel: ICU occupancy (hospital data).

Figure 2: Estimated impact of lockdown, per region.

Left panel: number of saved lives (up to now, per people).

Right panel: cumulative count of lost working days (up to now, per people).

Both counterfactual predictions are obtained by simulating the model, having switched off the component that controls peoples' context-dependent incentive to stay at home, aka "adaptive social distancing" (see Equation 1.1 in Friston et al., 2020). Social distancing clearly induces a cost in terms of lost working days during the epidemic outbreak.

Note:this estimate needs to be corrected for telecommuting home-workers.

References

Daunizeau J. , Moran R. J., Brochard J., Mattout J., Frackowiak R., Friston K.J. (2020), Modelling lockdown-induced 2nd COVID waves in France. MedRXiv, 2020.06.24.20139444 [DOI]

Moran, R.J., Billig, A.J., Cullen, M., Razi, A., Daunizeau, J., Leech, R., and Friston, K.J. (2020). Using the LIST model to Estimate the Effects of Contact Tracing on COVID-19 Endemic Equilibria in England and its Regions. MedRxiv 2020.06.11.20128611 [DOI]

Friston, K.J., Parr, T., Zeidman, P., Razi, A., Flandin, G., Daunizeau, J., Hulme, O.J., Billig, A.J., Litvak, V., Price, C.J., et al. (2020). Effective immunity and second waves: a dynamic causal modelling study. ArXiv:2006.09429 [DOI].

Friston, K.J., Parr, T., Zeidman, P., Razi, A., Flandin, G., Daunizeau, J., Hulme, O.J., Billig, A.J., Litvak, V., Price, C.J., et al. (2020). Tracking and tracing in the UK: a dynamic causal modelling study. ArXiv:2005.07994 [DOI]

Friston KJ, Parr T, Zeidman P, Razi A, Flandin G, Daunizeau J, Hulme, O.J., Billig, A.J., Litvak, V., Moran, R.J., et al. (2020). Second waves, social distancing, and the spread of COVID-19 across America. arXiv:200413017 [DOI]

Daunizeau J., Moran R.J., Mattout J., Friston K. (2020). On the reliability of model-based predictions in the context of the current COVID epidemic event: impact of outbreak peak phase and data paucity. MedRXiv (2020) [DOI]

Moran, R.J., Fagerholm, E.D., Daunizeau, J., Cullen, M., Richardson, M.P., Williams, S., Turkheimer, F., Leech, R., and Friston, K. (2020). Estimating required lockdown cycles before immunity to SARS-CoV-2: Model-based analyses of susceptible population sizes, S0, in seven European countries including the UK and Ireland. MedRxiv 2020.04.10.20060426. [DOI]

Friston, K.J., Parr, T., Zeidman, P., Razi, A., Flandin, G., Daunizeau, J., Hulme, O.J., Billig, A.J., Litvak, V., Moran, R.J., et al. (2020). Dynamic causal modelling of COVID-19. ArXiv200404463 Q-Bio. [DOI]