Corrigendum: Preparing correctional settings for the next pandemic: a modeling study of COVID-19 outbreaks in two high-income countries

Since the start of the coronavirus disease (COVID-19) pandemic in early 2020, correctional facilities around the world have experienced significant outbreaks of severe-acute-respiratorysyndrome-related coronavirus (SARS-CoV-2) (1)(2)(3)(4)(5). Such facilities (including gaols/jails, prisons, and other custodial settings), termed here "prisons", are vulnerable to outbreaks of SARS-CoV-2 and other highly transmissible respiratory infections due to their congregate nature with unavoidable close contact between people. People in prison are particularly vulnerable to severe COVID-19 given the higher prevalence of co-morbidities and poorer social determinants of health compared to the general population (2,6,7). Prisons' enclosed environments mean that SARS-CoV-2 can easily spread between people in prison, correctional and healthcare staff (for an Australian prison setting and this terminology will be used throughout the manuscript)/correctional employees (in a Canadian prison setting), and visitors.The transfer of people in prison between correctional facilities and into the community can also fuel outbreaks in other facilities and into surrounding communities (8). Significant outbreaks of COVID-19 occurred in high-income country prison settings. For example, there was a reported 50% prevalence of COVID-19 among inmates in the Federal Training Centre of Correctional Service of Canada (9) and in a residential treatment unit at the Cook County Jail, Chicago, Illinois, USA (10)during 2020. In another prison setting in the USA, the attack rate was estimated at 82% in one dormitory during April-May 2020 (11).Prisons are therefore high-priority settings for coordinated public health responses to the COVID-19 pandemic and future outbreaks of other respiratory infections (3,(12)(13)(14)(15). However, the response to COVID-19 in prisons has been hampered due to limited access and use of personal protective equipment (PPE) in resource-limited settings, poorer access and delay the vaccination and vaccine hesitancy, security and logistical constraints, frequent movement of people between correctional settings, and the continuous entry and exit of people into the prison (16)(17)(18)(19)(20). Correctional settings, therefore, require system-level and evidence-based responses (21,22).There have been several modelling studies evaluating the potential impact of prison-specific interventions to mitigate COVID-19 outbreaks in correctional settings. It was estimated that a large COVID-19 outbreak would be expected in prisons without both non-pharmaceutical interventions (NPIs) such as the use of PPE for people in prison and staff, decarceration of people in prison, quarantine at reception, isolation of people who are infected with COVID-19, and vaccination (23)(24)(25)(26)(27), particularly with delta and omicron variants (28)(29)(30), resulting in significant mortality (31). These models, however, did not consider the heterogeneous transmission network inside prisons, the characteristics of the population (among whom there is an increased risk of severe disease), and, for the most part, failed to use real-world data for calibration. These previous studies also neglected to focus on the combination of public health interventions that could potentially mitigate COVID-19 outbreaks. To our knowledge, this is the first study which has sought to model a combination of intervention strategies using models validated with 'real-world' data. In this study, we aimed to develop a COVID-19 model for two high-income prison settings in Australia and Canada and validate the model outputs against outbreak data from these two settings, and evaluate the potential impact of various intervention scenarios in averting cases and morbidity.We previously developed a COVID-19 Incarceration model by expanding on an existing spreadsheet model originally developed by Recidiviz (https://www.recidiviz.org) (32). The model aimed to capture additional complex features within prison environments, to reflect the mixing patterns between people in prison and correctional and healthcare staff, and to model a The Australian prison is a maximum-level quarantine prison with approximately 1,000 adult men (>18 years old). The Canadian prison is the largest provincial prison in Quebec, where it was the epicenter of the SARS-CoV-2 epidemic, with a capacity of 1,400 adult men (>18 years old) (34,35). Both prisons experienced SARS-CoV-2 outbreaks while multiple non pharmaceutical interventions (NPIs) were in place, but prior to a vaccine being available. A more detailed explanation of COVID-19 outbreaks in both prisons and interventions implemented will be explained later.The scenarios and structures of the model were informed by a reference group drawn from both healthcare and correctional organizations. A detailed explanation of the model structure is available elsewhere (36). Here we provide a summary and focus on the intervention scenarios investigated in the two settings. Briefly, the model is compartmental and implemented in Microsoft Excel (Redmond, WA). The model includes compartments representing the number of people in prison and staff who are susceptible, exposed, infectious, have mild illness, severe illness, are hospitalized, and recovered, with the number of deaths and new infections calculated daily (Figure 1). The model incorporates potential virus transmission between people in prison, correctional and healthcare staff, and visitors. It allows the designation of the prevalence of vulnerabilities in the population which could lead to severe COVID-19 disease, varied numbers of close contacts, and the daily intake and release of people in prison. 1). Agespecific infection fatality rates were specified for each age group. As people in prison enter the prison, they are allocated to each age group based on the age distribution of the people currently incarcerated. The model is implemented in a difference equation framework with the number of people in each compartment updated daily over 120 days. The model tracks people in prison who enter the prison either through reception or via transfer from another correctional setting, the daily number of visitors, and correctional and healthcare staff working at the site. People in prison leave the model population to reflect the number that are released after the end of their sentence or released early as a public health mitigation measure. We assumed symptomatic people in prison were not released until recovered. Staff are assumed to attend the prison site every day. Individuals from the community can visit the site every day to represent family visitors, but they are assumed to only have a limited number of contacts each visit (with a family member in prison and correctional staff).The COVID-19 progression rates were based on published data. The transmission of COVID-19 from infected to susceptible people per close contact with an infectious person had a value of 0.05 for the alpha variant (37), a value based on epidemics in Wuhan, China, accounting for different contacts through school, home, work, and other contacts. While the distance used for a close contact varies internationally, we defined a close contact to be a distance of less than 1.5 meters for longer than 15 minutes. We assumed the transmission probability was 1.5 times higher for the delta variant (38), and two times higher for the omicron variant, compared to the alpha variant (29,30). This transmission probability was adjusted to reflect the variable of susceptibility by age, the use of PPE (including masks, hand washing, and personal hygiene measures) and disease stage (Table 1). Viral shedding during the course of infection was considered to be low during the exposed stage (39), and hospitalized patients (assumed to be isolated) and healthcare workers were assumed to always wear PPE (assumed to be 1 in the 'Infectious' stage and from 0 'Exposure' to 0.8 among healthcare staff, Table 1). The number of contacts is specified in the model for each population group, and we assumed homogeneous mixing within the modelled prison setting. The effect of vaccination on preventing transmission and reducing hospitalization among people in prison and staff receiving the first and second dose is detailed in Table 1.The effects of five intervention strategies to reduce the risk of COVID-19 transmission were incorporated into the model (Figure 2). All NPIs are delineated in light pink and vaccination in dark orange. These include: 1. Deferred incarceration or early release of people in prison (decarceration), 2. Use of PPE by staff or people in prison, including gloves and masks, 3.Quarantine of new people in prison at reception (assumed quarantine for 14 days for all newly admitted people in prison in single cells (preferred) or in groups (if quarantine capacity is limited), 4. Isolation of people in prison with suspected or proven infection (assumed isolation group). As this information was not available for the Canadian prison, we used similar intermingling and number of contacts per inmate and staff as the Australian prison as both facilities have similar structures and resources. We gathered all the detailed data explained above through consultation with the reference group which was then incorporated into the parameters.The Australian prison experienced an outbreak with the delta variant from 11 August 2021 following multiple entries of infected inmates and staff. There were multiple NPIs in place at the time of the outbreak including: decarceration of people in prison, reduction in contacts, quarantine for 14 days at reception (entry), isolation of people in prison with suspected or proven infection, PPE for people in prison and staff, and thermal screening of non-essential staff and family visitors. Note that decarceration of people in prison in the Australian prison strategy was existed but the population size was not changed during the outbreak of COVID-19.The Canadian prison experienced an outbreak during the early stages of the pandemic from 15April 2020 when staff infected with the alpha variant entered the prison. Prior to this outbreak, there were several NPIs already in place aimed at controlling the number of close contacts each day, including: isolation among people in prison with suspected or proven infection, cessation of all visitors, 14-day quarantine of newly incarcerated people, and the distribution of PPE for all staff. Distribution of PPE to all people in prison was introduced in this prison during the outbreak from 2 June 2020 onwards.To correspond to what is believed to have occurred and set a 'baseline scenario' for both the Australian and Canadian prison models (Figure 3), we used the prison-specific demographic data as well as the interventions in place at the time of each prison's first outbreak. No vaccines were available in either prison at the time of the outbreak, however, vaccination began in the Australian prison among people in prison and staff during the outbreak and it likely contributed to mitigating the outbreak. In the Canadian prison, the first vaccine was administered on April 30, 2021 (personal communication on 19 January 2022, CIUSSS du Nord-de-l'Île-de-Montréal). A counterfactual 'no-response' scenario was run to see how large the COVID-19 outbreak could have been with no interventions in place (Figure 3).We simulated each model intervention separately (using scenarios 1 to 5) and in a combination scenario for 120 days to project the potential epidemic of COVID-19 within people in prison and staff. The number of hospital and intensive care unit (ICU) beds required are estimated from the model.For the vaccination scenarios, based on the advice of the reference group for the Australian prison, we assumed 50% of people in prison and 100% of staff were vaccinated (an estimate of the likely achievable coverage as vaccination of staff was mandated in the prison system).We assumed the same vaccination coverage among people in prison and staff in the Canadian prison as this information was not available. For intervention scenarios, we used the delta variant for both Australian and Canadian prisons to determine the impact of intervention strategies in both prisons. We further simulated a vaccination scenario using the omicron variant in both prisons to assess the possible impact of vaccination status in reducing COVID-Our model matched both the Australian and Canadian COVID-19 outbreaks well (Figure 3).In the Australian prison, where all NPIs were in place before the outbreak, the infections peaked on day 23 (Figure 3) with the first death from COVID-19 on day 26. The model estimated that there would have been 850 cumulative infections over 120 days with 1.7% of cases hospitalized at the peak of the infection (Table 2). In the Canadian prison, the infections peaked on day 28, with the first death from COVID-19 on day 33. The model estimated that there would have been 910 cumulative infections over 120 days with 80 people hospitalized at the peak of the infection (Table 2). Although the modelled estimates were higher than the number of people who were diagnosed with COVID-19 in the Canadian prison, we believe that there were undiagnosed cases in the prison (personal communication on 28th July 2020, CIUSSS du Nordde-l'Île-de-Montréal). Therefore, the estimated modelling outbreak (blue line) was used as the baseline to assess the impact of the interventions compared to the baseline scenario.Our model showed that, in the absence of any interventions (no response scenario, assuming admissions and releases continue), almost 100% of people in prison would become infected within 21 days of the outbreak, with 190 infections a day at the peak of the outbreak (day 14; Table 2 and Figure 3). Within 120 days of the outbreak, a total of 180 deaths due to COVID-19 were estimated (with deaths yet to plateau by 120 days; Table 2). At the peak of prevalence, approximately 84% of correctional and 92% of healthcare staff would also be infected and unable to attend work at the prison (Appendix, Figure A.1). Furthermore, our model estimated that if no response were in place, 470 hospital including 70 ICU beds would be needed at the peak of the outbreak in the local hospital facility (Table 2).For the following intervention scenarios in the Canadian prison, the delta variant was used to ensure consistency with the Australian prison. The model showed that, in the absence of a public health response (no response scenario), there would have been a large spike of COVID-19 cases (assuming admissions and releases continue) with almost 100% of people in prison becoming infected within 20 days (Figure 3). In this scenario, a total of people in prison and staff would be infected during 120 days of an outbreak, with deaths (Table 2). At the peak of the prevalence among people in prison, of correctional and of healthcare staff would also be infected and unable to attend work at the prison A.1). model estimated that if no response were in place, hospital including ICU beds would be needed at the peak of the outbreak (Table model that reducing the prison population size had the impact in reducing infections (among both people in prison and in the Australian prison reduction in cumulative by isolation of people in prison PPE and quarantine at reception reduction in reception (Figure Table 2). The model also showed that each intervention would reduce the number of hospital and ICU beds (Table 2). In the interventions to a outbreak with infections during the outbreak compared to the scenario (Figure Table the Canadian prison, decarceration also had the impact in reducing cumulative infection over 120 days by PPE quarantine at reception and isolation of people in prison (Figure Table 2). The impact of isolation was in the Canadian prison than in the Australian prison due to the capacity for isolation and quarantine of people in prison of inmates for isolation and for quarantine in the Canadian prison compared to a of for isolation and quarantine in the Australian In the baseline scenario, of COVID-19 infections were compared to the scenario over 120 days (Figure Table model of people in prison and staff would have a impact on COVID-19 outbreaks. at 50% of people in prison are vaccinated with 100% of staff vaccinated a which occurred in Australia but not in from day 0 in both prison settings would an outbreak from if other NPIs in place during the outbreak (Figure Table model that 50% vaccination coverage among people in prison is not to COVID-19 outbreaks when transmission probability is such as for the omicron variant, which has the transmission probability of the alpha variant Figure 100% of correctional and healthcare staff are vaccinated as a of two of any would mitigate transmission among staff without other NPIs there would be impact on the (with delta variant, Figure In this scenario, almost of the outbreaks among staff would be but would among people in prison, with a rate and a peak in daily if the outbreak was by a staff developed a model the of prison settings, and disease The model was used to assess of public health strategies to epidemic patterns and the effect of or mitigation The model can be for to different prison settings and to other respiratory with similar transmission and could be used for general pandemic in prison settings in the We the model to two prisons in two high-income and including the characteristics of the prisons where real-world COVID-19 outbreaks model showed that modelling outputs the COVID-19 and the that there is a risk of a COVID-19 outbreak within prisons if an infected in the absence of our model that with vaccination an outbreak from was the most by decarceration reduction in cumulative over 120 in reducing COVID-19 outbreaks in prison model showed that and in can reduce the size of a COVID-19 outbreak within prisons and reduce and the prison population size (decarceration), quarantine of people in prison at reception, and isolation of symptomatic people in prison are to reduce close contacts between infected and susceptible reduce the susceptible population within a prison, while the use of PPE and vaccination the risk of transmission during close While these interventions are our modelling showed that an outbreak could are for resource-limited settings where access to vaccines in correctional settings not available. For example, in prison were to to reduce COVID-19 outbreaks as the prison was than times a to release people from prison who risk to public was to reduce the number of people or prisons to the of the COVID-19 outbreaks was also introduced in prisons in the early outbreaks vaccination of both people in prison and staff, with will mitigate all future outbreaks and be in where this has not occurred model also that at the peak of prevalence among people in prison, of correctional and of healthcare staff would be infected and unable to attend work at prison. The of correctional and healthcare staff in person due to COVID-19, not only the and of incarcerated but also a significant to the public health of the as the virus can easily spread the prison through staff who the infection of the the absence of correctional staff can lead to a of and security within the prison, it more to the and of people in prison. it is to the health and of correctional staff in to ensure the of both incarcerated and the general It is also to that a of is to ensure the of both people in prison and staff in correctional settings which will mitigate the risk of new outbreaks and in a residential treatment unit at the Cook County Jail, Chicago, Illinois, USA In the prison, USA, the attack rate was estimated at 82% in dormitory It be that these outbreaks occurred with the of intervention such as and among symptomatic Furthermore, these outbreaks could potentially be capacity was limited early in the outbreak and was a after the of the first COVID-19 in which likely in transmission before the Therefore, it is to estimate the potential of the outbreak in the absence of such In other prison study the baseline scenario in the model (with no mitigation in was to of the population becoming based on outbreaks that occurred in the community where the transmission probability is than in the prisons. studies have used transmission rates to reflect data on transmission within different of a prison cells For our model we on the transmission per contact of where that contact We used a transmission probability per contact that with other study estimates but into the increased risk due to the congregate nature and unavoidable close contacts that among in prison settings. The model we developed is also that it can be to different prison settings and to other respiratory will not the and of our study also that vaccination with is required to mitigate the outbreak risk in prisons, for variants with transmission modelling study in the the impact of with various in place to reduce close contacts including of non-essential throughout the country or in reducing the number of people hospitalized and deaths due to COVID-19 model that with low vaccination can number of people being The of this will likely prisons, on the available health and correctional and nature of within the modelling study also showed that the combination of NPIs and vaccination can deaths due to COVID-19, but required For example, more than of the vaccination rate was needed to the deaths within days Therefore, it is that NPIs are needed in prison if the vaccination rates are particularly with the of transmissible COVID-19 are to our It is to that as our model is it not capture all the within a prison, or the between This it the of an outbreak in a prison where the structure includes multiple and that can be from each other in the of an outbreak. The from a are in model is also which it not capture effects when the infection numbers are The model the movement between quarantine and isolation and the general prison population as an rate to the of the This that there can be a release of infected from in the model an outbreak than what be on the intervention these are as a with a However, people in prison in with staff, and exposed be released at the end of their that this spread of infection from is not model did not into the reduction in population size resulting from and during COVID-19 outbreaks. For our study, we to develop a more detailed model that the movement of inmates between prisons and The model also assumed that if staff infected with COVID-19, they would not to work for days following infection, in with the work by in This the number of staff infected and unable to attend work in the prison, as staff be infected with COVID-19 but However, we believe our of the for interventions in prisons to outbreaks in the response scenario' will and with the where the prisons were and inmates were in their cells due to a staff in 2022, in prisons, Australia our model not the impact of varied strategies for study has several While our model was to interventions for SARS-CoV-2 transmission in prisons, structure is to consider other respiratory infections in other population settings by the transmission It also to and assess different scenarios and a combination of public health strategies to epidemic patterns and the effect of or mitigation we on COVID-19 outbreaks in two 'real-world' prison settings with intervention strategies to mitigate future outbreaks. the Australian and Canadian prisons were both prisons, however, model were based on published data for both and Therefore, our are likely to our that the entry of one infected person into prison is to an outbreak, almost all people in prison within 120 days in the absence of an A vaccination in combination with other would the risk of an outbreak in a prison, but the of these interventions will on both the health and custodial of the from this study can be used to evaluate other respiratory in congregate settings in the have no of of and the of and in data of and in data of and the study, in the of and and in data and the the study, in the of and the study, in the of and The model is available an 3) via with a from a COVID-19 outbreak in two prisons the and intervention scenarios in prison and intervention is separately and in combination and run for 120 days. are for the population the prison site and are to the

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.