Temporal variation in transmission during the COVID-19 outbreak in Italy
This study has not yet been peer reviewed.
* This analysis is now archived, please visit the updated version.
updated: 2020-04-04
Note: this is preliminary analysis, has not yet been peer-reviewed and is updated daily as new data becomes available. This work is licensed under a Creative Commons Attribution 4.0 International License. A summary of this report can be downloaded here
Summary
Aim: To identify changes in the reproduction number, rate of spread, and doubling time during the course of the COVID-19 outbreak in Italy whilst accounting for potential biases due to delays in case reporting.
Latest estimates as of the 2020-03-19
Region map
Figure 1: Regional map of the expected change in daily cases based on data from the 2020-03-19.
Summary of latest reproduction number and case count estimates
Figure 2: Cases with date of onset on the day of report generation and the time-varying estimate of the effective reproduction number (bar = 95% credible interval) based on data from the 2020-03-19. Regions are ordered by the number of expected daily cases and shaded based on the expected change in daily cases. The dotted line indicates the target value of 1 for the effective reproduction no. required for control and a single case required fror elimination.
Reproduction numbers over time in the 5 regions with the most cases currently and nationally
Figure 3: Time-varying estimate of the effective reproduction number (light grey ribbon = 95% credible interval; dark grey ribbon = the interquartile range) based on data from the 2020-03-19 in the regions expected to have the highest number of incident cases. Confidence in the estimated values is indicated by shading with reduced shading corresponding to reduced confidence. The dotted line indicates the target value of 1 for the effective reproduction no. required for control.
Latest estimates summary table
| Country/Region | Cases with date of onset on the day of report generation | Expected change in daily cases | Effective reproduction no. | Doubling time (days) |
|---|---|---|---|---|
| Lombardia | 971 – 3968 | Increasing | 1.1 – 1.7 | 6.1 – Cases decreasing |
| Emilia Romagna | 291 – 1206 | Increasing | 1.3 – 2.1 | 4.1 – Cases decreasing |
| Piemonte | 274 – 1094 | Increasing | 1.5 – 3.1 | 2.3 – 9.5 |
| Veneto | 123 – 481 | Increasing | 1.1 – 1.8 | 5.6 – Cases decreasing |
| Campania | 73 – 394 | Increasing | 1.5 – 2.9 | 2.8 – 27 |
| Liguria | 68 – 347 | Increasing | 1.2 – 2.3 | 4 – Cases decreasing |
| Marche | 73 – 316 | Increasing | 1.1 – 1.8 | 5.7 – Cases decreasing |
| Friuli Venezia Giulia | 50 – 286 | Increasing | 1.3 – 2.4 | 2.6 – Cases decreasing |
| Toscana | 60 – 284 | Increasing | 1.2 – 2.3 | 3.3 – Cases decreasing |
| Trentino-Alto Adige | 47 – 271 | Increasing | 1.1 – 2.1 | 2.4 – Cases decreasing |
| Abruzzo | 47 – 253 | Increasing | 1.6 – 3.5 | 2.2 – 7.2 |
| Puglia | 36 – 194 | Increasing | 1.4 – 2.8 | 1.9 – Cases decreasing |
| Lazio | 40 – 191 | Increasing | 1.2 – 2.2 | 3.4 – Cases decreasing |
| Umbria | 29 – 181 | Increasing | 1.5 – 3.3 | 2 – 76 |
| Sardegna | 25 – 158 | Increasing | 1.6 – 4 | 1.6 – Cases decreasing |
| P.A. Trento | 23 – 141 | Increasing | 1 – 2 | 2.6 – Cases decreasing |
| P.A. Bolzano | 18 – 134 | Increasing | 1.2 – 2.6 | 2.7 – Cases decreasing |
| Sicilia | 19 – 126 | Increasing | 1.3 – 2.4 | 3.2 – 27 |
| Valle d’Aosta | 16 – 109 | Increasing | 1.5 – 4 | 2.1 – Cases decreasing |
| Calabria | 12 – 92 | Increasing | 1.3 – 3 | 2.1 – 46 |
| Molise | 5 – 51 | Increasing | 1.9 – 7 | 0.19 – Cases decreasing |
| Basilicata | 2 – 34 | Increasing | 1.6 – 5.3 | 0.15 – Cases decreasing |
Table 1: Latest estimates of the number of cases by date of onset, the effective reproduction number, and the doubling time for the 2020-03-19 in each region included in the analysis. Based on the last 7 days of data. The 95% credible interval is shown for each numeric estimate.
Methods
Summary
- Case counts by date, stratified by region, were constructed from daily datasets made publically available by the Dipartimento della Protezione Civile [1,2].
- Case onset dates were estimated using case counts by date of report and a distribution of reporting delays fitted to a European line-list.
- Censoring of cases was adjusted for by assuming that the number of cases is drawn from a binomial distribution.
- Time-varying effective reproduction estimates were made with a 7-day sliding window using EpiEstim [3,4] adjusted for imported cases and assuming an uncertain serial interval with a mean of 4.7 days (95% CrI: 3.7, 6.0) and a standard deviation of 2.9 days (95% CrI: 1.9, 4.9) [5].
- Time-varying estimates of the doubling time were made with a 7-day sliding window by iteratively fitting an exponential regression model.
- The methods in this report are based on our previous study of the global temporal variation during the COVID-19 outbreak [6].
Limitations
- The estimated onset dates are based on current European data for the delay in reporting and are mostly from the beginning of the outbreak. This means that these data may not be representative of the underlying delay distribution.
- The estimate of not-yet-confirmed cases to scale up recent numbers is uncertain and relies on the observed delays to confirmation to remain constant over the course of the outbreak.
- All data used is at a national/regional level; diagnostic capabilities may vary in different parts of each region, adding uncertainty to the reported numbers. The true number of infections reflected in a given number of confirmed cases probably varies substantially geographically.
- Trends identified using our approach are robust to under-reporting assuming it is constant but absolute values may be biased by reporting rates. Pronouced changes in reporting rates may also impact the trends identified.
- Data on imported cases was not available (either international imports or between region imports).
- As our estimates are made at the date of symptom onset any changes in the time-varying parameters will be delayed by the incubation period.
Detail
Data
We used a European line-list that contained the date of symptom onset, date of confirmation and import status (imported or local) for each case [2,7] where available. Daily case counts by date of report and region were extracted from daily datasets made publically available by the Dipartimento della Protezione Civile [1,2].
Statistical analysis
We used the same approach as in our previous global study of the temporal variation in transmission during the COVID-19 outbreak [6]. However, due to a limited line-list of Italian cases we used a combined linelist of cases from Germany, France, Italy, Austria, the Netherlands, Belgium, and Spain to estimate the report delay. We could also not account for imported cases (either international or between region) due to a shortage of data. Code and results from this analysis can be found here and here.
Regional reports
Lombardia
Summary
Figure 4: A.) Cases by date of report (bars) and estimated cases by date of onset. B.) Time-varying estimate of the effective reproduction number. Light grey ribbon = 95% credible interval. Dark grey ribbon = the interquartile range. Based on data from the 2020-03-19. Confidence in the estimated values is indicated by shading with reduced shading corresponding to reduced confidence.
| Estimate | |
|---|---|
| Cases with date of onset on the day of report generation | 971 – 3968 |
| Expected change in daily cases | Increasing |
| Effective reproduction no. | 1.1 – 1.7 |
| Rate of spread | -0.026 – 0.11 |
| Doubling time (days) | 6.1 – Cases decreasing |
| Adjusted R-squared | -0.16 – 0.89 |
Table 3: Latest estimates of the number of cases by date of onset, the expected change in daily cases, the effective reproduction number, the rate of spread, the doubling time, and the adjusted R-squared of the exponential fit for the 2020-03-19. Based on the last 7 days of data. The 95% credible interval is shown for each numeric estimate.
Time-varying rate of spread and doubling time
Figure 5: A.) Time-varying estimate of the rate of spread, B.) Time-varying estimate of the doubling time in days (note that when the rate of spread is negative the doubling time is assumed to be infinite), C.) The adjusted R-squared estimates indicating the goodness of fit of the exponential regression model (with values closer to 1 indicating a better fit). Based on data from the 2020-03-19. Light grey ribbon = 95% credible interval; dark grey ribbon = the interquartile range. Confidence in the estimated values is indicated by shading with reduced shading corresponding to reduced confidence.
Implementation details
Emilia Romagna
Summary
Figure 7: A.) Cases by date of report (bars) and estimated cases by date of onset. B.) Time-varying estimate of the effective reproduction number. Light grey ribbon = 95% credible interval. Dark grey ribbon = the interquartile range. Based on data from the 2020-03-19. Confidence in the estimated values is indicated by shading with reduced shading corresponding to reduced confidence.
| Estimate | |
|---|---|
| Cases with date of onset on the day of report generation | 291 – 1206 |
| Expected change in daily cases | Increasing |
| Effective reproduction no. | 1.3 – 2.1 |
| Rate of spread | -0.012 – 0.17 |
| Doubling time (days) | 4.1 – Cases decreasing |
| Adjusted R-squared | -0.14 – 0.97 |
Table 4: Latest estimates of the number of cases by date of onset, the expected change in daily cases, the effective reproduction number, the rate of spread, the doubling time, and the adjusted R-squared of the exponential fit for the 2020-03-19. Based on the last 7 days of data. The 95% credible interval is shown for each numeric estimate.
Time-varying rate of spread and doubling time
Figure 8: A.) Time-varying estimate of the rate of spread, B.) Time-varying estimate of the doubling time in days (note that when the rate of spread is negative the doubling time is assumed to be infinite), C.) The adjusted R-squared estimates indicating the goodness of fit of the exponential regression model (with values closer to 1 indicating a better fit). Based on data from the 2020-03-19. Light grey ribbon = 95% credible interval; dark grey ribbon = the interquartile range. Confidence in the estimated values is indicated by shading with reduced shading corresponding to reduced confidence.
Implementation details
Piemonte
Summary
Figure 10: A.) Cases by date of report (bars) and estimated cases by date of onset. B.) Time-varying estimate of the effective reproduction number. Light grey ribbon = 95% credible interval. Dark grey ribbon = the interquartile range. Based on data from the 2020-03-19. Confidence in the estimated values is indicated by shading with reduced shading corresponding to reduced confidence.
| Estimate | |
|---|---|
| Cases with date of onset on the day of report generation | 274 – 1094 |
| Expected change in daily cases | Increasing |
| Effective reproduction no. | 1.5 – 3.1 |
| Rate of spread | 0.073 – 0.31 |
| Doubling time (days) | 2.3 – 9.5 |
| Adjusted R-squared | 0.4 – 0.98 |
Table 5: Latest estimates of the number of cases by date of onset, the expected change in daily cases, the effective reproduction number, the rate of spread, the doubling time, and the adjusted R-squared of the exponential fit for the 2020-03-19. Based on the last 7 days of data. The 95% credible interval is shown for each numeric estimate.
Time-varying rate of spread and doubling time
Figure 11: A.) Time-varying estimate of the rate of spread, B.) Time-varying estimate of the doubling time in days (note that when the rate of spread is negative the doubling time is assumed to be infinite), C.) The adjusted R-squared estimates indicating the goodness of fit of the exponential regression model (with values closer to 1 indicating a better fit). Based on data from the 2020-03-19. Light grey ribbon = 95% credible interval; dark grey ribbon = the interquartile range. Confidence in the estimated values is indicated by shading with reduced shading corresponding to reduced confidence.
Implementation details
Veneto
Summary
Figure 13: A.) Cases by date of report (bars) and estimated cases by date of onset. B.) Time-varying estimate of the effective reproduction number. Light grey ribbon = 95% credible interval. Dark grey ribbon = the interquartile range. Based on data from the 2020-03-19. Confidence in the estimated values is indicated by shading with reduced shading corresponding to reduced confidence.
| Estimate | |
|---|---|
| Cases with date of onset on the day of report generation | 123 – 481 |
| Expected change in daily cases | Increasing |
| Effective reproduction no. | 1.1 – 1.8 |
| Rate of spread | -0.14 – 0.12 |
| Doubling time (days) | 5.6 – Cases decreasing |
| Adjusted R-squared | -0.24 – 0.6 |
Table 6: Latest estimates of the number of cases by date of onset, the expected change in daily cases, the effective reproduction number, the rate of spread, the doubling time, and the adjusted R-squared of the exponential fit for the 2020-03-19. Based on the last 7 days of data. The 95% credible interval is shown for each numeric estimate.
Time-varying rate of spread and doubling time
Figure 14: A.) Time-varying estimate of the rate of spread, B.) Time-varying estimate of the doubling time in days (note that when the rate of spread is negative the doubling time is assumed to be infinite), C.) The adjusted R-squared estimates indicating the goodness of fit of the exponential regression model (with values closer to 1 indicating a better fit). Based on data from the 2020-03-19. Light grey ribbon = 95% credible interval; dark grey ribbon = the interquartile range. Confidence in the estimated values is indicated by shading with reduced shading corresponding to reduced confidence.
Implementation details
Campania
Summary
Figure 16: A.) Cases by date of report (bars) and estimated cases by date of onset. B.) Time-varying estimate of the effective reproduction number. Light grey ribbon = 95% credible interval. Dark grey ribbon = the interquartile range. Based on data from the 2020-03-19. Confidence in the estimated values is indicated by shading with reduced shading corresponding to reduced confidence.
| Estimate | |
|---|---|
| Cases with date of onset on the day of report generation | 73 – 394 |
| Expected change in daily cases | Increasing |
| Effective reproduction no. | 1.5 – 2.9 |
| Rate of spread | 0.026 – 0.25 |
| Doubling time (days) | 2.8 – 27 |
| Adjusted R-squared | 0.04 – 0.94 |
Table 7: Latest estimates of the number of cases by date of onset, the expected change in daily cases, the effective reproduction number, the rate of spread, the doubling time, and the adjusted R-squared of the exponential fit for the 2020-03-19. Based on the last 7 days of data. The 95% credible interval is shown for each numeric estimate.