Testing lags and emerging COVID-19 outbreaks in federal penitentiaries in Canada

Objectives: To provide the first known comprehensive analysis of COVID-19 outcomes in a federal penitentiary system. We examined the following COVID-19 outcomes within federal penitentiaries and contrasted them with the overall population in the penitentiaries' respective provincial jurisdictions: testing, prevalence, the proportion recovered, and fatality. Methods: Data for prisons were obtained from the Correctional Service of Canada and, for the general population, from COVID-19 Esri Canadian Outbreak Tracking Hub. Data were retrieved between March 30 and April 21, 2020, and are accurate to this date. Penitentiary-, province- and sex-specific frequency statistics for each outcome were calculated. Results: Data on 50 of 51 penitentiaries (98%) were available. Of these, 72% of penitentiaries reported fewer tests per 1000 population than the Canadian general population average (16 tests/1000 population), and 24% of penitentiaries reported zero tests. Penitentiaries with high levels of testing were those that already had elevated COVID-19 prevalence. Five penitentiaries reported an outbreak (at least one case). Hardest hit penitentiaries were those in Quebec and British Columbia, with some prisons reporting COVID-19 prevalence of 30% to 40%. Of these, two were women's prisons. Female prisoners were over-represented among cases (31% of cases overall, despite representing 5% of the total prison population). Conclusion: Increased sentinel or universal testing may be appropriate given the confined nature of prison populations. This, along with rigorous infection prevention control practices and the potential release of prisoners, will be needed to curb current outbreaks and those likely to come.


49
As we are (re)learning from the evidence of COVID-19 outbreaks in long-term care facilities, 50 housing large groups of individuals in confined, institutionalized environments is a recipe for 51 large-scale, human-to-human transmission of infectious disease. Prisons have similar dynamics 52 of confinement, of residents who are dependent on the institutions' conditions and policies for 53 basic living and hygiene and, of clinical susceptibility due to underlying health conditions [1][2][3]. 54 Epidemiologic findings from past outbreaks of respiratory diseases such as influenza, 55 adenoviruses, and tuberculosis in high-income countries highlight the higher likelihood of 56 infection transmission and disease incidence within prisons compared to the general population. 57 [1][2][3]. Early reports suggest that several penitentiaries in the United States are experiencing 58 , though, as far as we know, no comprehensive analysis of theses 59 penitentiaries has yet been conducted, and very little is known about COVID-19 in prison 60 contexts in other high-income countries. In this study, we use Canadian data to provide the first 61 known examination of COVID-19 statistics for a federal penitentiary system. In doing so, we 62 also bring front-and-center a major health equity issue, because Indigenous and racialized 63 communities are over-represented within the Canadian prison system [5]. 64

65
In the context of the COVID-19 pandemic, there are specific factors that place prisoner 66 populations at particularly high risk of COVID-19 infection, and of a higher risk for severe 67 (compared to mild) COVID-19 outcomes. These include housing multiple prisoners to one cell 68 ("double-bunking") [6], the ageing population within prisons-particularly in federal prisons [7], 69 a high prevalence of chronic disease comorbidities and immuno-compromised health status 70 associated with substance use and blood-borne infections [8], and the regular entrance of 71 custodial and health care staff from communities with likely community-based transmission. 72 73 At the time of writing (late April 2020), Canada has been documenting the COVID-19 pandemic 74 on its soil for almost two months. In late March 2020, the Correctional Service of Canada (CSC), 75 Canada's federal body responsible for incarcerated individuals serving sentences of at least two 76 years, had indicated that they had taken steps to prevent and were prepared to respond to 77 COVID-19 outbreaks inside its federal penitentiaries [9]. With this study, we aim to offer a 78 Given that several federal penitentiaries are in fact multiplex sites-either with distinct buildings 109 or units operating under different security levels (e.g. minimum, medium or maximum) or 110 offering distinct services (e.g. treatment facilities)-and that sub-population capacities were not 111 always available for each separate unit, we chose to group several multi-complex facilities in our 112 analyses. This was the case of Quebec's Federal Training Center Multi-Level Unit and Minimum 113 security facility; British Columbia's Pacific Institution, Pacific Institution Regional Treatment 114 Center and Pacific Institution Reception Center; Ontario's Millhaven Institution, Millhaven 115 Regional Hospital and Millhaven Regional Treatment Center, as well as Collins Bay Minimum 116 Institution and Collins Bay Regional Treatment Center, and Joyceville Institution and Joyceville 117 Minimum Institution. With these groupings, we recorded 51 facilities. We were able to analyze 118 data from 50 of these facilities (98%) (all data in Supplementary File 1). 119 120 Measures 121 122

Total tests and cases 123
The total number of tests performed was obtained directly from the CSC and the Canada 124 Outbreak Tracker Resource Hub websites. CSC reports on the number of "positive tests". We 125 considered all "positive tests" as confirmed cases. 126 127

Tests per 1000 population 128
Tests per 1000 population were estimated by dividing the total number of tests performed by the 129 total population in each facility, in the prisoner population of each province, and the general 130 population of each province, respectively, and multiplying the fraction by 1000. 131 132

Test-positive rate 133
We estimated the proportion of tests that were positive by dividing the total number of positive 134 tests (confirmed cases) by the total number of tests, in each prison, provincial prison population, 135 and the provincial general population. 136 137 Tested COVID-19 prevalence 138 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 8, 2020. . https://doi.org/10.1101 We estimated the prevalence of COVID-19 as identified through tests in the penitentiary and 139 provincial populations by dividing the total number of positive tests (confirmed cases) by the 140 population of each facility, the total prison population by province, and the total provincial 141 population, respectively. This indicator can also be considered the cumulative incidence 142 proportion. 143 144

Proportion recovered 145
To offer a crude estimate of the evolution of the epidemics inside and outside of prisons, we 146 estimated the proportion of cases that had recovered. This was done by dividing the number of 147 cases who had recovered by the total number of positive tests (total confirmed cases). 148 Operational definitions of recovered cases vary across jurisdictions in Canada, and CSC does not 149 define their measure in their reporting. Recovered cases are those who either received 150 confirmatory negative test results or for whom 10 to 14 days have elapsed since the start of their 151 symptoms and they are symptom-free for at least 2 to 3 days [14,15]. Therefore, we interpret 152 recovered case numbers with caution. 153 154

Population categories in prisons with outbreaks: Susceptible, Infected, Recovered and Died 155
An outbreak of COVID-19 within confined populations, such as long-term care homes, can be 156 declared after a single resident or staff member is confirmed positive [16]. Thus, we consider 157 prisons with one or more reported cases of COVID-19 among its prisoner population as those 158 experiencing outbreaks. To describe the evolution of testing and incidence in prisons with 159 outbreaks across our study period, we classified each prison's population into four categories, for 160 each calendar day of the study period. We estimated the number of prisoners who were 161 "susceptible" to infection by subtracting the total number of confirmed active, recovered, and 162 deceased cases from the maximum population capacity of each prison. Prisoners considered 163 "infected" were those with positive tests who had yet to recover or die. Prisoners declared to be 164 "recovered" were assumed to no longer present active infections or be susceptible to infection. 165 Lastly, the total number of prisoners who had "died" were those whose death was declared to 166 result from COVID-19 complications. 167 168

169
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted May 8, 2020.  hour drive away from Toronto) also experienced a smaller outbreak of 8 cases (6% prevalence). 200 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted May 8, 2020. In the five institutions with one or more cases, we found that testing was largely reactive, with 207 the onset of concentrated testing efforts occurring after COVID-19 outbreaks had already 208 established. An exception to this observation was Quebec's Federal Training Center, which was 209 the only facility recording negative tests before the observations of positive tests ( Figure 3). 210 Further, though institutions all recorded a gradual increase in the number of tests performed over 211 time, the high proportion of positive tests throughout the study period suggests that these prisons 212 likely did not implement more intensive, wide-spread testing throughout the prison populations 213 once a case was observed. An exception to this, however, was Ontario's Grand Valley Women's 214 Institution, which recorded a significant jump in testing once the first cases were confirmed, with 215 a majority of tests returning negative since April 7, 2020. 216 217

Proportion of cases recovered and fatality inside versus outside prisons 218 219
The proportion of cases that had recovered inside federal penitentiaries ranged from 0% to 33% 220 CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted May 8, 2020.

232
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted May 8, 2020.

235
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted May 8, 2020. . https://doi.org/10. 1101 In contrast to cases in the general population, a smaller proportion of cases in federal 236 penitentiaries in Quebec, Ontario, and British Columbia had recovered ( Figure 2). In both 237 Ontario and British Columbia, the proportion of individuals who had recovered within prisons 238 was half that of the proportion of cases recovered in the community. With the available evidence, 239 it is difficult to know precisely what may be driving these differences in percent recovered inside 240 versus outside of prisons. One explanation is that outbreaks inside prisons may have started more 241 recently than those within the general public. Another possibility is that health and sanitary 242 conditions within prisons may lead to lags in recovery. Further, this cross-sectional summary of 243 differences must be interpreted with caution, since the proportion of cases that are recovered will

262
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted May 8, 2020. . https://doi.org/10. 1101 At the time of data collection, there were 59 cases of COVID-19 in women's penitentiaries. 263 These represented 31% of all cases in federal penitentiaries (N=189) (Figure 5), suggesting that 264 women, and women's penitentiaries, are over-represented among COVID-19 cases inside federal 265 prisons. COVID-19 prevalence was 8 times higher among women's prisons (8% prevalence) 266 than prisons for men (1% prevalence) and 80 times higher than in the general Canadian 267 population (0.1%). 268 The data suggest that at least five federal penitentiaries, one in British Columbia, three in Quebec 274 and one in Ontario, are experiencing COVID-19 outbreaks. Each of these penitentiaries is 275 reporting high levels of testing, compared to provincial and national rates, and this testing 276 appears to be a reaction to the emergence of outbreaks that have already begun, rather than a 277 proactive prevention effort. Though at least thirteen (26%) of facilities appear to be experiencing 278 very low if any COVID-19 cases, these are also facilities that have reported low levels of testing 279 per 1000 population, so it is unclear whether the low case counts reflect true case prevalence. We 280 observed lower levels of testing than the Canadian general population average in 26 281 penitentiaries (72%) and a total absence of testing in 12 penitentiaries (24%). Our study also 282 suggests a smaller proportion of cases inside prisons have recovered compared to cases in the 283 general population. This may indicate that the spread of COVID-19 is lagging in penitentiaries, 284 and the worst is yet to come. It may also indicate that prisoners face suboptimal conditions for 285 recovery. We also found that at least three of the current penitentiary outbreaks are in proximity 286 to city centers (e.g. Montreal and Vancouver), and higher COVID-19 prevalence has been 287 observed in women's (8% prevalence) compared to men's prisons (1% prevalence), overall. 288 289

Strengths and limitations 290 291
The primary strength of this study is its use of the best available data on COVID-19 testing and 292 case incidence in Canadian Federal penitentiaries and the general Canadian population to give an 293 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted May 8, 2020. . https://doi.org/10.1101/2020.05.02.20086314 doi: medRxiv preprint early situational report on testing and infection-related outcomes within these facilities. This 294 study is a baseline assessment from which future analyses using updated data can be based. A 295 limitation of this study is our use of the maximum potential capacity of each penitentiary as the 296 population denominator for rate calculations, rather than the 'true' population at the time of data 297 collection. We expect that these denominators are likely too large, given that prisons may not be 298 at their full capacity and pressures to release non-violent offenders to reduce the number of 299 susceptible individuals within prisons [17]. This may have led to the under-estimation of testing 300 and infection prevalence estimates. Further, this study was not able to also count the number of 301 tests and cases identified within custodial and health staff within these same penitentiaries, which 302 is an important contributing factor to disease transmission. While other deaths in federal 303 penitentiaries were recorded by CSC during the study period (e.g. [18]), only one case was 304 reported as COVID-19 related. It is unclear whether other deaths occurring within CSC facilities 305 have or will be tested for COVID-19 post-mortem. Lastly, these findings may not be 306 generalizable to provincial, remand, or immigration detention facilities, which may see far more 307 movement in and out given the shorter duration of sentences. 308 Our findings of sparse testing across several penitentiaries in Canada suggests that a more 321 proactive testing approach may be needed to help curb potential future outbreaks. This is 322 especially the case given studies suggesting that up to 60% of COVID-19 cases may be 323 asymptomatic [21][22][23]. Testing based on symptom-presence alone may thus need to be 324 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted May 8, 2020. . https://doi.org/10.1101 reconsidered within the confined spaces of prisons, as many cases may go undetected as long as 325 testing is not expanded. 326 327 Relevant testing strategies could include universal testing or, active, sentinel-based testing. 328 While universal testing [19] [24] is time and resource-intensive but is much more feasible within 329 small confined populations than in the general public. In the context of COVID-19, this strategy 330 has been used navy ships [21] and homeless shelters [25,26]  . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted May 8, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted May 8, 2020. . https://doi.org/10.1101  . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted May 8, 2020.  is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 8, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted May 8, 2020. . https://doi.org/10.1101 Data on facilities with one or more cases as of April 21, 2020 Data were available for March 30 and April 7-8, 10, 12, 13, 15, 17-18, 21 of 2020. Port Cartier Institution, Quebec (Maximum Population Estimate: 237 prisoners)