COVID-19 dynamics in an Ohio prison

Case data

Mass RT-PCR testing at MCI of all inmates and partial testing of staff was conducted on April 16, 2020. Total number of inmates and number of inmates and staff at MCI testing positive for SARS-CoV-2 over time were obtained from public ODRC reports [2]. To account for lags in data entry, we accumulate the cases reported at MCI over April 16-23, 2020 as a single mass testing data point.

Mathematical model

We use a susceptible-exposed-infectious-recovered (SEIR) framework to model SARS-CoV-2 dynamics in MCI. Under the standard SEIR model, the proportions of individuals in the susceptible (S_t), exposed (E_t), infectious (I_t), and recovered (R_t) compartments (at time t) satisfy the following system of differential equations: where the positive parameters β, α, and γ denote the infection rate, the incubation rate, and the recovery rate.

Statistical analysis

We derive a likelihood function for observing n positives out of N incarcerated individuals on day u as follows. Using the Dynamic Survival Analysis approach of [9], we treat S_t as an improper survival function. Given observation up to time T > 0, the time T_E that an individual becomes infected and enters the E compartment follows the conditional probability density function where τ_T = 1 − S_T. The time T_I to becoming infectious has the conditional density and the recovery time T_R has the conditional density Note that the random variables T_E, T_I − T_E, and T_R − T_E are mutually independent, and that T_I − T_E and T_R − T_E are exponentially distributed with rates α and γ.

Mass testing yields the number of individuals who test positive and the total number of tests administered on the day of mass testing. To use these data, let T_N denote the time when virus first becomes undetectable in an individual. We then describe the epidemic process by the pair of random variables (T_E, T_N). Let ε = T_N − T_E. Then, the conditional probability of an individual testing positive on the day of the mass testing (at time u) is given by We fix ε = 21 days, corresponding to detectable virus for three weeks following an individual becoming infectious [10, 11]. We set 1/α = 5.1/ log(2) days (corresponding to a median incubation period of 5.1 days [12]), and assume a mean infectious period 1/γ of 5.6 days [13].

If n out of N individuals test positive on the day of the mass testing u, the likelihood function is given by with the probability of testing positive p_u as described in (2.5).

Reported outbreak time course

According to ODRC reports [2], the first identified COVID-19 case at MCI was an infected staff member on March 29. Following this initial primary case, precautions such as cohorting and modified movement were enacted in order to restrict mixing and reduce transmission.² The first COVID-19 case among inmates was identified on April 3. Mass RT-PCR testing of all inmates and partial testing of staff was conducted on April 16. By April 20, SARS-CoV-2 infection had been identified in 79% (1950/2453) of inmates and 35% (154/446) of staff.³ A time series of COVID-19 cases at MCI is shown in Figure 1.

• Download figure
• Open in new tab

Figure 1: Positive tests over time among (a) inmates and (b) staff at MCI, as reported in [2].

Basic reproduction number and initial exposure size

To examine what values for the basic reproduction number ℛ₀ are consistent with such rapid spread, we turn to the SEIR model (system (2.1)). In order for mass screening to identify 80% of the population as positive for COVID-19 on April 16, at least 80% of the population must have been infected by that date. Figure 2 shows the time needed to infect 80% of the population in the SEIR model as a function of the basic reproduction number (ℛ₀) and the initial number of exposed individuals (E₀). While ℛ₀ values of two or larger are able to eventually infect 80% or more of the population, for modest ℛ₀ values this can take on the order of months. Reproduction numbers larger than 14 are needed before outbreaks originating from a single exposed individual are able to generate infection levels in a three week span that are consistent with those reported for MCI.

• Download figure
• Open in new tab

Figure 2:

Time to infect 80% of population in SEIR model with median incubation period of 5.1 days [12] and mean infectious period of 5.6 days.

An alternative explanation is that the out-break involved more than one initially infected prisoner. From Figure 2, we see that for a fixed ℛ₀ value, increasing E₀ decreases the time needed to infect 80% of the population. However, an initial condition of E₀ > 471 is needed for an outbreak with ℛ₀ = 3.27 to infect 80% of the population within three weeks.

Time of initial outbreak circulation

A third possibility is that SARS-CoV-2 was circulating among prisoners prior to April 3. Figure 3 shows the log-likelihood (2.6) for observing the mass testing results in MCI according to the SEIR model (2.1) as a function of ℛ₀ and the outbreak onset date, with E₀ fixed at one. The outbreak onset date and ℛ₀ are indistinguishable from the mass testing data alone, with the yellow “wishbone” in Figure 3 corresponding to outbreak onset and ℛ₀ pairs that are equally likely given the data. Outbreak onsets in late March or later correspond to ℛ₀ values larger than ten, while earlier outbreak onsets correspond to smaller ℛ₀ values. Note that onset dates prior to March are required to give ℛ₀ values of less than five.

• Download figure
• Open in new tab

Figure 3:

Left. Log-likelihood in the ℛ₀-outbreak onset date plane, for mass testing data under a SEIR model with three week test-positive window following infectious onset. Right. ℛ₀ values that maximize the likelihood as a function of outbreak onset date. This curve corresponds to following the right-most branch of the ‘wishbone’ in the likelihood plot.