Data for the models of COVID-19 infections, hospitalizations, and ICU admissions considered the average detention population from October 1, 2019 to March 2, 2020 in 111 facilities across 32 states. The number of detainees in each facility during this period ranged from 25 to 1600 (average 381, median 234 detainees), with a total of 42,435 individuals detained. Thirty-two percent (n = 35) of the included facilities hold only ICE detainees, and 68% (n = 76) are city or county correctional facilities that hold ICE detainees and other individuals who are incarcerated. Although sex data were not available at the individual level, we obtained facility-level sex data. There were 577 women held in two female-only facilities; 15,560 individuals were held in 43 male-only facilities; and 26,208 detainees were in facilities that held both females and males.
Based on 2019 data provided by Freedom for Immigrants, the majority of immigrants detained by ICE are between 26 and 35 years old, where the median age is 30 . Country of origin for detainees included in this analysis was not available; however, a 2019 report by the U.S. Government Accountability Office found that the majority of detentions from 2015 to 2018 were males from Mexico, Guatemala, El Salvador, and Honduras, without a previous arrest or conviction record .
Number of Infections over Time in Single-Facility Outbreaks
Table 1 shows the estimated number of COVID-19 cases across the three R0 scenarios over 30-, 60-, and 90-day time horizons, for ICE facilities holding 50, 100, 500, and 1000 detainees. In a facility holding 1000 people, under the most optimistic scenario (R0 = 2.5), a median of 62 individuals are expected to have been infected by day 30, 363 by day 60, and 722 by day 90. Under the least optimistic scenario of R0 = 7, the number of cumulative COVID-19 cases is modeled to increase from 732 by day 30, to 989 by day 60, to 993 by day 90. The day 90 estimates, corresponding to 76%, 80%, 77%, and 72% of all detainees in facilities of size 50, 100, 500, and 1000 detainees, give a sense of the total population of ICE detainees who are likely to be infected by the end of an ICE facility outbreak under the most optimistic scenario of R0 = 2.5 (Fig. 1). Regarding the timing and size of the peak of the epidemic, as one would predict, simulated epidemics within small facilities reach their peak earlier than those within large facilities (for example, on day 19 for facilities of 50 detainees vs day 69 for facilities of 1000 detainees with R0 = 3.5; see Supplemental Materials).
Healthcare Service Utilization during Facility Outbreaks
Under the most optimistic scenario (R0 = 2.5), and imagining synchronized epidemics across all facilities, a median of 917 detainees system-wide will become infected and require hospitalization within the first 30 days, growing to an aggregate of 3129 patients by day 60 and 4909 by day 90. Averaging across the total ICE system, this represents 11.6% of all detainees. The high hospitalization rate is driven by sustained transmission leading to very high fractions of detainees becoming infected, in the absence of robust and sustained countermeasures. The corresponding numbers for ICE facilities of different sizes at day 90 are given in Table 1. Under the pessimistic scenario of R0 = 7, and imagining synchronized epidemics across all facilities, a median of 5145 detainees system-wide will become infected and require hospitalization within the first 30 days, growing to an aggregate of 6391 patients by day 60 and 6408 patients by day 90. Averaging across the total ICE system, this represents 15.1% of all detainees. Concretely, in a facility housing 1000 detainees, the estimated median number of people infected and requiring hospitalization by day 30 ranges from 10 (R0 = 2.5) to 26 (R0 = 3.5) and 116 (R0 = 7) and by day 90 ranges from 114 (R0 = 2.5) to 148 (R0 = 3.5) and 157 (R0 = 7).
Focusing now on estimating ICU admissions, we estimate that with R0 = 2.5, and again imagining that all epidemics are synchronized, there will be a median of 139 detainees infected who will require ICU admission after 30 days, growing to an aggregate of 475 patients by day 60 and 745 patients by day 90 (approximately 1.8% of the total population). Under the pessimistic scenario of R0 = 7, the corresponding numbers of patients are 782, 971, and 973 on days 30, 60, and 90, respectively. This corresponds to approximately 2.3% of the total population.
Estimating Hospital Utilization during an Unmitigated Epidemic
There are 33 facilities without hospitals within 10 miles of the ICE facility while all facilities have a hospital within 50 miles. A COVID-19 outbreak at a number of detention facilities would potentially overwhelm local ICU capacity within a 10- and 50-mile radius of each facility (Table 2). Under the most optimistic scenario (R0 = 2.5), and imagining synchronized epidemics across all facilities, estimated median aggregate ICU admissions from 56 facilities by day 30, 62 facilities by day 60, and 65 facilities by day 90 would exceed ICU capacity within 10 miles of the facility. These numbers include those facilities with no ICU capacity within 10 miles. Under the least optimistic scenario (R0 = 7), the numbers of facilities where ICU beds within 10 miles would not be adequate increases to 66 facilities within 30 days of onset and this persists through 60 and 90 days. A small number of facilities are also estimated to generate more ICU needs than ICU beds within 50 miles during the first 30 days of onset (six such facilities for R0 = 2.5 or 3.5 and eight such facilities for R0 = 7). For R0 = 2.5 and R0 = 7, the aggregate number of patients requiring ICU admission exceeds available beds within 50 miles for 8 and 10 facilities at 90 days, respectively.