Artificial Neural Network Prediction of COVID-19 Daily Infection Count

Abstract

This study addresses COVID-19 testing as a nonlinear sampling problem, aiming to uncover the dependence of the true infection count in the population on COVID-19 testing metrics such as testing volume and positivity rates. Employing an artificial neural network, we explore the relationship among daily confirmed case counts, testing data, population statistics, and the actual daily case count. The trained artificial neural network undergoes testing in in-sample, out-of-sample, and several hypothetical scenarios. A substantial focus of this paper lies in the estimation of the daily true case count, which serves as the output set of our training process. To achieve this, we implement a regularized backcasting technique that utilize death counts and the infection fatality ratio (IFR), as the death statistics and serological surveys (providing the IFR) as more reliable COVID-19 data sources. Addressing the impact of factors such as age distribution, vaccination, and emerging variants on the IFR time series is a pivotal aspect of our analysis. We expect our study to enhance our understanding of the genuine implications of the COVID-19 pandemic, subsequently benefiting mitigation strategies.

Appendix A: Additional Data About COVID-19

In this section we present many figures that demonstrate raw data, processed data, and intermediate results used to generate the training set. Some data for selected states have been already demonstrated in the main text. This includes

• 1 Time series of IFR for all 50 states plus Washington DC

• 2 Time series of recovered true cases and undercounting factor for all 50 states plus Washington DC

• 3 Raw and smoothed confirmed daily case count and daily death count for all 50 states plus Washington DC

• 4 Time series of case rate per age group at all regions of the United States

• 5 Time series of vaccination rate of all age group for all 50 states plus Washington DC

• 6 Incident rate ratio of COVID-19 case and death for vaccinated and unvaccinated groups.

• 7 Time series of testing volume for all 50 states plus Washington DC

1.1 A.1 Time Series of State IFR

The time series of IFR for 10 selected states are presented in the main text. Below we demonstrate the time series of IFR for all 50 states plus Washington DC after considering age group case rate, vaccination, variant in Figs. 13 and 14.

Fig. 13

Time series of state IFR for 24 states plus Washington DC (Color figure online)

Fig. 14

Time series of state IFR for 26 states (Color figure online)

1.2 A.2 Time Series of State Recovered True Case

The time series of recovered true case and under counting factor for 10 selected states are demonstrated in the main text. Here we show these data for all 50 states plus Washington DC in Figs. 15 and 16.

Fig. 15

Time series of recovered true case count for 24 states plus Washington DC (Color figure online)

Fig. 16

Time series of recovered true case count for 26 states and Washington DC (Color figure online)

1.3 A.3 State Confirmed Case and Death

Figures 17 and 18 show the daily case count and $100 \times$ daily death count of all 50 states plus Washington DC. The data comes from the JHU COVID-19 database (Center 2023). Figure 19 and 20 are the processed daily case count and daily death count after addressing data dump and holiday issues.

Fig. 17

Daily confirmed case count and $100 \times$ daily death count for 24 states plus Washington DC. Raw data before processing (Color figure online)

Fig. 18

Daily confirmed case count and $100 \times$ daily death count for 26 states. Raw data before processing (Color figure online)

Fig. 19

Daily confirmed case count and $100 \times$ daily death count for 24 states plus Washington DC. Processed data after addressing weekday issue, holiday issue, and artificial data dump from backlogs (Color figure online)

Fig. 20

Daily confirmed case count and $100 \times$ daily death count for 26 states. Processed data after addressing weekday issue, holiday issue, and artificial data dump from backlogs (Color figure online)

1.4 A.4 Case Rate Per Age Group

Figure 21 shows the time series of case rate of each age group from all 10 regions provided by CDC (Disease Control 2023a). The HHS regions used by CDC is described in the following Table 1.

Table 1 List of states and districts in each CDC region

Fig. 21

Case rate of each age group in all 10 regions (Color figure online)

1.5 A.5 State Vaccination Rate

Figures 22 and 23 gives the time series of vaccinate rate for each age group older than 18 years old in all 50 states plus Washington DC. This data is obtained from CDC (Disease Control 2023b).

Fig. 22

Time series of vaccination rate of each age group for 24 states plus Washington DC (Color figure online)

Fig. 23

Time series of vaccination rate of each age group for 26 states (Color figure online)

1.6 A.6 Incident Rate Ratio (IRR) of Vaccinated and Unvaccinated Groups

The incident ratio of COVID-19 infection and death for each group is given in Fig. 24. This data is obtained from CDC website (Disease Control 2023d). Note that death data of younger age group is not included because there are too few, sometimes zero, death count from vaccinated young group in many weeks. The ratio of IFR of unvaccinated group to vaccinated group of three older age groups are shown in Fig. 24 Right.

Fig. 24

Left and Middle: Incident rate ratio (IRR) of COVID-19 infection and death for each age group. Right: Ratio of IFR of unvaccinated group to vaccinated group (Color figure online)

1.7 A.7 State Testing Volume

Figures 25 and 26 gives the time series of smoothed COVID-19 test volume in all 50 states plus Washington DC. This data comes from the Coronavirus Resource Center of Johns Hopskins University (Center 2023).

Fig. 25

Time series of COVID-19 testing volume for 24 states plus Washington DC. The blue and red lines are raw data and smoothed data respectively (Color figure online)

Fig. 26

Time series of COVID-19 testing volume for 26 states. The blue and red lines are raw data and smoothed data respectively (Color figure online)

1.8 A.8 Training Set Data Distribution

Figure 27 displays the distribution of the data in the training data set.

Fig. 27

Distribution of normalized data (Color figure online)