COVID-19 impact on opioid overdose after jail release in Massachusetts

Highlights

• •Incarcerated persons were released during the pandemic to mitigate SARS-CoV-2 spread in jails and prisons.
• •In Massachusetts, release of jailed persons with opioid use disorder during the COVID-19 pandemic was associated with a higher rate of fatal overdose compared to the pre-pandemic period, but the number of associated deaths was small.
• •Compassionate jail releases were unlikely to explain much, if any, of the observed increase in community overdose during the pandemic.

Abstract

Introduction

Release from incarceration is a high-risk period for opioid overdose. Concern about COVID-19 spread in jails led to early releases; it is unknown whether pandemic era releases of persons with opioid use disorder (OUD) contributed to increases in community overdose rates.

Methods

Observational data compared overdose rates three months after release among jailed persons with OUD released before (9/1/2019–3/9/2020) and during the pandemic (3/10/2020–8/10/2020) from seven jails in Massachusetts. Data on overdoses come from the Massachusetts Ambulance Trip Record Information System and Registry of Vital Records Death Certificate file. Other information came from jail administrative data. Logistic models regressed overdose on release period, controlling for MOUD received, county of release, race/ethnicity, sex, age, and prior overdose.

Results

Pandemic releases with OUD had a higher risk of fatal overdose (adjusted odds ratio [aOR] 3.06; 95% CI, 1.49 to 6.26); 20 persons released with OUD (1.3%) experienced a fatal overdose within three months of release, versus 14 (0.5%) pre-pandemic. MOUD had no detectable relationship with overdose mortality. Pandemic release did not impact non-fatal overdose rates (aOR 0.84; 95% CI 0.60 to 1.18), though in-jail methadone treatment was protective (aOR 0.34; 95% CI 0.18 to 0.67).

Conclusions

Persons with OUD released from jail during the pandemic experienced higher overdose mortality compared to pre-pandemic, but the number of deaths was small. They did not experience significantly different rates of non-fatal overdose. Early jail releases during the pandemic were unlikely to explain much, if any, of the observed increase in community overdoses in Massachusetts.

1. Introduction

The time after release from incarceration is a high-risk period for opioid overdose (Binswanger et al., 2009; Lim et al., 2012; Merrall et al., 2010). The COVID-19 pandemic exacerbated both nonfatal and fatal overdoses (Imtiaz et al., 2021; Khare et al., 2022; Soares et al., 2022); indeed, the rate of opioid-related overdose deaths increased 5% from 2019 to 2020 and 9% from 2020 to 2021 in Massachusetts’ general population (Massachusetts Department of Public Health, 2022).

Concern about the spread of SARS-COV-2 among incarcerated persons led to compassionate jail and prison releases early in the pandemic (Aslim and Mungan, 2020; Marcum, 2020). On March 10, 2020, Massachusetts declared a state of emergency, and its court system soon instituted a process to release non-violent persons from jails and prisons. Anecdotally, many persons were released on short notice, making it difficult to arrange for care-continuity planning, including community-based treatment for substance use disorder and prevalent conditions among the non-violent incarcerated population. These challenges raise the possibility that the pandemic-related releases from jails and prisons contributed to the observed higher rates of overdose during the pandemic.

This study examines whether jail releases of persons with opioid use disorder (OUD) during the pandemic were associated with higher rates of overdose compared to releases prior to the pandemic. We further hypothesized that receipt of medications for opioid use disorder (MOUD) by persons with OUD prior to jail release would reduce any association of pandemic period release with opioid overdose.

2. Methods

2.1. Study design

These observational data compare rates of overdose among persons released immediately before (September 1, 2019 to March 9, 2020) and during the pandemic (March 10, 2020 to August 10, 2020). Data come from the Massachusetts JCOIN (MassJCOIN) research hub which is assessing the outcomes of the Massachusetts legislative mandate (MA Chapter 208 of the Acts of 2018) to offer medications for opioid use disorder (MOUD) in seven jails that started on September 1, 2019 (Evans et al., 2021).

2.2. Study population

In Massachusetts, county sheriff's departments administer the jails, which detain persons who are awaiting trial, and houses of correction, which house convicted individuals with short sentences (generally 36 months or less). In this study we use these terms synonymously because each county sheriff's department has both a jail and a house of correction often in the same building, with high turnover of clients, fluidity and mixing of the pre-trial and sentenced populations. A separate Massachusetts state agency, the department of corrections, administers prisons, which house convicted persons with longer sentences and less population churn.

We examined overdose outcomes of all incarcerated adults with OUD who were released from seven Massachusetts county jails before and during the COVID-19 pandemic. For individuals with multiple releases from jail during the study period, only the first release to the community was included. Transfers to other correctional facilities were excluded. We defined the pre-pandemic release period as September 1, 2019 (the date on which participating jails were required to begin offering medication for opioid use disorder [MOUD] to inmates) until March 9, 2020. We defined the pandemic release period as March 10, 2020 (the date on which the Commonwealth of Massachusetts declared a state of emergency in response to the COVID-19 pandemic) until August 10, 2020.

Identification of incarcerated individuals with OUD varied by jail and included self-report and urine test results for opioid use or MOUD, staff verification of MOUD prescription information in the statewide Prescription Drug Monitoring Program database, and OUD screening/assessment conducted by clinical staff. OUD diagnosis information was data entered into the jail's electronic medical record along with whether individuals received MOUD during their incarceration, and if treated, which MOUD type (buprenorphine, methadone, naltrexone, none) was received.

2.3. Key variables

The primary outcomes of interest were non-fatal and fatal opioid overdoses within the first 3 months after jail release.

2.4. Data sources

The Massachusetts Ambulance Trip Record Information System (MATRIS) database provided information on all acute opioid-related overdoses treated by Emergency Medical Services (EMS) from an algorithm developed by the Massachusetts Department of Public Health (Bettano et al., 2021). Individuals could have both a non-fatal and a fatal overdose. To identify fatal opioid overdoses, the following International Classification of Disease, 10th Revision (ICD-10) codes for mortality were selected from the underlying cause of death field in the Registry of Vital Records and Statistics Death Certificate file to identify poisonings/overdoses: X40-X44, X60-X64, X85, and Y10-Y14. All multiple cause of death fields were then used to identify an opioid related overdose death:  T40.0, T40.1, T40.2, T40.3, T40.4, and T40.6. Deaths that did not yet have a valid ICD-10 code, a literal search of written cause of death from the medical examiner's office was performed.

Demographic and other baseline participant information, including MOUD type received, date of jail entry and exit, race/ethnicity, sex, and age, were extracted and compiled by JCOIN research assistants from jail administrative data sources including medical intake forms and Medication for Opioid Use Disorder Enrollment Assessments required by Massachusetts Bureau of Substance Addiction Services (BSAS). Enrollment information (e.g. demographics and time of treatment) were entered into the state-mandated Bureau of Substance Addition Services EIM/ESM system; the remainder into RedCap (Harris et al., 2019). If race/ethnicity information was not available from assessment data, it was obtained from the BSAS. Key identifying variables such as name, social security number, date of birth, and sex, were also compiled and utilized to link participant data to additional variables available in the statewide public health data warehouse (Bharel et al., 2020; Larochelle et al., 2018). These datasets were probabilistically linked by name, date of birth, sex, and Social Security number using Match*Pro software (Surveillance Research Program, n.d.).

2.5. Statistical methods

Descriptive statistics for participant demographics and other study characteristics were reported using means and standard deviations or frequencies and percentages. Small cell counts were suppressed per confidentiality restrictions.

We calculated relative risks and 95% confidence intervals comparing the proportion of jail releases with OUD who received different MOUD types before and during the COVID-19 pandemic. Simple associations between overdose and both COVID release timing and MOUD receipt were explored using chi-square or Fisher's exact tests.

Logistic regression models were fit to examine COVID release timing as a correlate of overdose, with baseline characteristics included as covariates. We generated odds ratios and 95% confidence intervals for each covariate in the model. Each model included the primary explanatory variable, COVID-19 release timing (before or during the pandemic), as well as MOUD received, county of release, race/ethnicity, sex, age, and whether they had experienced overdoses prior to their index jail stay. All statistical analyses were performed in SAS Studio (Version 3.8).

3. Results

During the pre-pandemic period (September 1, 2019 to March 9, 2020), 2950 persons with OUD were released from these jails, with an average of 15.45 releases per day compared to 1522 with an average of 9.88 releases per day during the pandemic period (March 10, 2020 to August 10, 2020) (Table 1).

Table 1.

Baseline comparison of jailed persons with opioid use disorder released before and during the COVID-19 pandemic.

	Release before pandemic (9/1/19 – 3/9/20) N = 2950	Release during pandemic (3/10/20 – 8/10/20) N = 1522
Age, mean (SD)	36.1 (9.2)	36.7 (9.4)
Sex, N (%) Female Male Transgender Unknown	752 (25.5%) 2194 (74.4%) …* 0	319 (21.0%) 1201 (78.9%) …* …*
Race/Ethnicity, N (%) White, non-Hispanic Black, non-Hispanic Hispanic Multiracial/Other Missing	2028 (68.8%) 182 (6.2%) 650 (22.0%) 83 (2.8%) †	984 (64.7%) 125 (8.2%) 368 (24.2%) 41 (2.7%) …*
Overdose prior to incarceration, N (%)	743 (25.2%)	375 (24.6%)
County Jail of Release, N (%) A B C D E F G	532 (18.0%) 224 (7.6%) 969 (32.9%) 188 (6.4%) 324 (11.0%) 315 (10.7%) 398 (13.5%)	212 (13.9%) 85 (5.6%) 428 (28.1%) 85 (5.6%) 189 (12.4%) 143 (9.4%) 380 (25.0%)

^⁎Cells with values 1–4 suppressed to meet confidentiality requirements.

^†Complimentary cell suppression applied to meet confidentiality requirements.

The pandemic period saw a slight increase in the proportion of released persons with OUD who were male, with 74.4% male in the pre-pandemic period versus 78.9% male in the pandemic period. At the same time, the proportion of releases with OUD who were White, non-Hispanic inmates decreased slightly, from 68.8% White, non-Hispanic in the pre-pandemic period to 64.7% White, non-Hispanic in the pandemic period. The proportion of releases who reported an overdose prior to incarceration did not change appreciably.

The distribution of inmate releases shifted slightly among the counties during the pandemic, with jail G experiencing an increase in the proportion of released individuals with opioid use disorder relative to the other jails (Table 1). Its population comprised 13.5% of the entire population in the pre-pandemic period and 25.0% of the entire population in pandemic period.

The proportion of released persons with OUD who had received MOUD in jail increased significantly during the pandemic period (55.5%) compared to pre-pandemic (51.4%) (relative risk for receipt of MOUD during the pandemic [RR]1.12, 95% CI 1.03, 1.21) (Table 2). Furthermore, released persons with OUD in the pandemic period had increased receipt of buprenorphine (RR 1.21, 95% CI 1.11, 1.32) and methadone (RR 1.14, 95% CI 1.00, 1.29) compared to the pre-pandemic period. However, naltrexone receipt decreased significantly in the pandemic period (RR 0.438, 95% CI 0.317, 0.604).

Table 2.

Proportion of jailed persons with opioid use who received medications for opioid use disorder (MOUD), by release period before and during the COVID-19 pandemic.

	Release period				Medication received during pandemic
	Before pandemic		During pandemic		Relative Risk (95% CI)	p-value
	N	%	N	%
Any MOUD Received†	1516	51.4%	845	55.5%	1.12 (1.03 – 1.21)	0.009
MOUD Type Received†
Buprenorphine	954	32.3%	606	39.8%	1.21 (1.11 – 1.32)	<0.0001
Methadone	358	12.1%	206	13.5%	1.14 (1.00 – 1.29)	0.0418
Naltrexone	202	6.9%	33	2.2%	0.438 (0.317 – 0.604)	<0.0001
No MOUD Received†	1434	48.6%	677	44.5%	referent†

*Cells with values 1–4 suppressed to meet state confidentiality requirements.

^†No MOUD receipt is referent for receipt of any MOUD, and for each MOUD type.

Among individuals with OUD released from jail during the period of this study, a small proportion, 4.7% overall, experienced a non-fatal or fatal opioid overdose within three months of release from jail, and that proportion remained stable across study periods (Table 3). The proportion of releases with OUD who experienced a non-fatal opioid overdose in the COVID period decreased slightly compared to the pre-COVID period (3.5% vs. 4.2% respectively). Conversely, the proportion of those with OUD who experienced a fatal opioid overdose in the COVID period increased compared to the pre-COVID period (0.5% vs. 1.3%).

Table 3.

Proportions of jailed persons with OUD with non-fatal, fatal or any overdose within 3 months after release from jail, by release period before or during the COVID-19 pandemic, and medication for opioid use disorder (MOUD) received in jail.

	Non-fatal overdose	p-value‡	Fatal overdose	p-value‡	Any Overdose*	p-value‡
Release period, N (%)
Before pandemic During pandemic	125 (4.2%) 53 (3.5%)	0.2210	14 (0.5%) 20 (1.3%)	0.0022	138 (4.7%) 71 (4.7%)	0.9844
MOUD received in jail, N (%)
Buprenorphine Methadone Naltrexone None	63 (4.0%) 10 (1.8%) 9 (3.8%) 96 (4.6%)	0.0294	13 (0.8%) …† …† 16 (0.8%)	0.8796§	74 (4.7%) 13 (2.3%) 11 (4.7%) 111 (5.3%)	0.0330

^⁎Individuals could have both a non-fatal and a fatal overdose.

^†Cells with values 1–4 were suppressed to meet state confidentiality requirements.

^‡P-values in this table from Chi-square tests, except where indicated.

^§P-value from Fisher exact test because 25% of cells have expected counts less than 5.

The proportions of released persons with OUD who experienced an opioid overdose varied slightly by MOUD type (Table 3). Of those who received buprenorphine, 4.7% experienced a non-fatal or fatal opioid overdose. Similarly, 4.7% of those who received naltrexone experienced a non-fatal or fatal opioid overdose within three months of release. Released persons with OUD who received methadone experienced a slightly smaller proportion of opioid overdoses (2.3%) while those who did not receive any MOUD experienced a slightly larger proportion of opioid overdoses (5.3%).

Released individuals with OUD who received buprenorphine, naltrexone, or no MOUD experienced similar proportions of non-fatal opioid overdoses (4.0%, 3.8% and 4.6% respectively) while those who received methadone experienced a slightly lower frequency of non-fatal opioid overdose (1.8%, n = 10). Only 0.8% of releases with OUD experienced a fatal opioid overdose, regardless of whether or not they received MOUD. To ensure data privacy, small cells for overdose fatality were suppressed among releases with OUD who received methadone or naltrexone (Table 3).

3.1. Non-fatal opioid overdose

Released persons with OUD saw a non-significant 16% decrease in odds of non-fatal overdose during the pandemic compared to the pre-pandemic period (Table 4). Released individuals with OUD who received buprenorphine, methadone, and naltrexone all saw a decrease in the odds of non-fatal opioid overdose, but only those who received methadone saw a significant decrease (adjusted odds ratio, 0.34). Age, sex and race/ethnicity appeared to have little to no relationship with the rate of non-fatal opioid overdose. Released individuals with OUD who had experienced at least one opioid overdose prior to incarceration exhibited a significant 98% increase in odds of non-fatal opioid overdose post-release. Lastly, the jail of release did not appear to have a significant association with the odds of non-fatal opioid overdose within 3 months of release.

Table 4.

Logistic regression models for non-fatal and fatal overdose within 3 months of release among jailed persons with opioid use disorder released before or during the COVID-19 pandemic.

	Non-fatal overdose odds ratio (95% CI)	p-value	Fatal overdose odds ratio (95% CI)	p-value
Release period
Before pandemic During pandemic	referent 0.84 (0.60 – 1.18)	0.3186	referent 3.06 (1.49 – 6.26)	0.0022
MOUD receipt prior to release
Buprenorphine	0.77 (0.54 -1.10)	0.4578	0.98 (0.43 – 2.19)	0.9111
Methadone	0.34 (0.18 – 0.67)	0.0081	0.59 (0.17 – 2.08)	0.3322
Naltrexone	0.83 (0.40 – 1.77)	0.5065	1.38 (0.28 – 6.77)	0.5295
None	referent		referent
Age	1.00 (0.98 – 1.01)	0.6861	1.04 (1.00 – 1.08)	0.0415
Female Sex	0.98 (0.66 – 1.48)	0.9389	2.63 (1.09 – 6.34)	0.0309
Race/Ethnicity
White, non-Hispanic	referent		referent
Black, non-Hispanic	0.69 (0.34 – 1.39)	0.3205	0.88 (0.20 – 3.91)	0.6827
Hispanic	0.79 (0.52 – 1.19)	0.4532	0.58 (0.19 – 1.76)	0.1487
Multi-racial/other	1.26 (0.57 - 2.77)	0.2970	3.09 (0.90 – 10.66)	0.0456
Overdose prior to incarceration	1.98 (1.44 – 2.71)	<0.0001	2.24 (1.10 – 4.56)	0.0269
County Jail of Release
A	0.78 (0.46 – 1.32)	0.4438	1.24 (0.39 – 3.91)	0.5048
B	0.45 (0.19 – 1.07)	0.0540	0.41 (0.05 – 3.28)	0.3612
C	referent		referent
D	1.16 (0.62 – 2.16)	0.3523	1.69 (0.43 – 6.67)	0.2846
E	1.15 (0.69 – 2.16)	0.2420	0.85 (0.21 – 3.56)	0.8908
F	1.01 (0.57 – 1.77)	0.6531	0.98 (0.23 – 4.13)	0.9132
G	1.10 (0.68 – 1.79)	0.3385	0.79 (0.29 – 2.21)	0.7320

3.2. Fatal opioid overdose

Persons with OUD released during the pandemic period experienced a significant increase in the risk of fatal opioid overdose (adjusted odds ratio, 3.06) compared to those released during the pre-pandemic period (Table 4). Released individuals with OUD who received buprenorphine and methadone showed a non-significant decrease in odds of fatal overdose compared to no MOUD, while those who received naltrexone exhibited higher odds. Older releases with OUD showed a significantly higher odds ratio for fatal opioid overdose, with a 4% increase in odds per year (p = 0.0415). Released women with OUD were also at significantly higher risk of fatal opioid overdose compared to men (adjusted odds ratio, 2.63; p = 0.0309). Releases with OUD who experienced at least one opioid overdose prior to incarceration continued to experience a higher rate of fatal opioid overdose (adjusted odds ratio, 2.24; p = 0.0269), while the particular jail from which they were released continued to show no significant association with the odds of fatal opioid overdose (Table 4). Race was not related to the risk of fatal opioid overdose among Black and Hispanic releases with OUD, but those who identified as other races or multi-racial saw approximately three times the odds of fatal opioid overdose (p = 0.0456) compared to those who identified as White, non-Hispanic. However, the frequency of released individuals with OUD who identified in the multi-racial/other category and experienced a fatal opioid overdose was small enough to warrant suppression for data privacy reasons.

4. Discussion

This study of seven jails in Massachusetts found that jailed persons with OUD released during the COVID-19 pandemic experienced a three-fold increase in fatal overdose in the three months after release compared to the period just prior to the pandemic, but the number of deaths was small. The period of release did not have a detectable relationship with the rate of non-fatal overdose in the released population with OUD.

Overdose deaths in the U.S. increased by approximately 20% in the first year of the pandemic (Ahmad et al., 2022). In Massachusetts, opioid-related overdose deaths statewide increased 5% from 2005 in 2019 to 2103 in 2020 (Massachusetts Department of Public Health, 2022). In the seven counties associated with the jails in the current study, opioid-related overdose deaths increased 6% from 1185 to 1257 over the same time period. This pandemic-associated rise in opioid overdose mortality was likely multifactorial, and it is likely that factors unrelated to jail policies explain much of our detected increase in post-release overdose mortality. Stay-at-home orders and other pandemic mitigation strategies increased social isolation and the likelihood that people use drugs alone (Khare et al., 2022). COVID-19-related disruptions in the drug supply appear to be associated with riskier drug consumption patterns (Aponte-Melendez et al., 2021; Zolopa et al., 2021). At the same time, concerns about COVID-19 may have reduced treatment seeking, recovery group participation, willingness to call Emergency Medical Services, and hospital visits out from fear of COVID-19 exposure.

Overdose fatality after release was also associated with older age, female sex, multi-racial or other race, and prior overdose. Older age is a known risk factor for overdose mortality, likely because it is associated with greater medical comorbidity; however, it bears emphasis because some have reported that older PWUDs perceive themselves as low risk for overdose (Rowe et al., 2016). For sex, some studies suggest women are at higher risk of overdose (Degenhardt et al., 2011; Gjersing and Bretteville-Jensen, 2014), while other studies have not found an association of sex with overdose (Darke et al., 2011). Although Black and Latinx persons are disproportionately represented in correctional settings and have experienced disproportionately high rates of pandemic-related overdose (Friedman et al., 2021), we did not find a higher burden of pandemic-associated overdoses among jail releases from these minority groups. The “other race” category, though small, includes people from multiracial and indigenous backgrounds, the latter of whom have experienced large increase in overdose mortality (Joshi et al., 2019). Prior overdose is also a well-described risk factor for overdose mortality (Caudarella et al., 2016; Darke et al., 2011; Krawczyk et al., 2020).

These seven jails are participating in a pilot program in Massachusetts to offer all three Food and Drug Administration (FDA)-approved forms of medication for opioid use disorder (Evans et al., 2021); notably, only methadone treatment was associated with a lower rate of non-fatal overdose in the 3 months after release. No medication had a detectable association with overdose fatality. This preliminary result differs from findings in other high-risk populations in Massachusetts that found a protective effect of methadone and buprenorphine.(Larochelle et al., 2018). The limited impact of methadone and buprenorphine appear to be the result of small cells, especially for mortality, but might represent limited effectiveness in a jail population that is at high-risk for overdose mortality and requires strong linkage to ongoing MOUD treatment when transitioning back to the community. Ongoing work with this population will examine the effectiveness of MOUD in jail with a larger sample and longer follow-up.

Several limitations merit consideration. We are unable to ascertain whether the COVID-19 period releases were early releases to reduce the jail census in response to the pandemic, or regular releases. The pandemic was associated with decreased crime and intakes as law enforcement and the courts reduced arrests and detentions (Boman and Gallupe, 2020; Marcum, 2020). As a result, these jails’ censuses decreased substantially: 36% fewer persons with OUD were released than before the pandemic and fewer individuals received MOUD. Common sense suggests that the remaining jailed persons likely had more severe offenses, but the data limitations do not allow us to comment on any changes in their charges or other unmeasured characteristics. In addition, the populations released during the two time periods could differ in other characteristics, such as other substance use and mental health diagnoses, that were unavailable in this dataset. Indeed, modeling was limited to available covariates that might reasonably be associated with the outcomes and do not represent theory-driven models of all factors related to post-release overdose; hence, residual confounding remains a possibility. Furthermore, the jails varied in how individuals with OUD were identified and it is possible that as the MOUD pilot programs matured, more persons with OUD were identified. Although we have no evidence that OUD screening improved over time, it remains possible that greater detection of persons with less severe OUD in the pandemic period than in the pre-pandemic period would cause underestimation of the true impact of pandemic jail releases on overdose mortality. In addition, avoidance of emergency care during the COVID-19 pandemic period might have spuriously reduced non-fatal overdoses from the MATRIS ambulance data (Czeisler et al., 2020). Misclassification of cause of death is possible because death certificates could have differentially underreported overdose as a cause of death during the pandemic; these false negatives could have caused an underestimation of the overdose risk associated with the COVID-19 period releases (Becker et al., 2020; Merlin et al., 2022). Conversely, the ICD-10 codes used in Massachusetts’ death certificates are highly specific for overdose [Dana Bernson, personal communication] and prior work has validated that Massachusetts death certificates indicate a specific drug in 99.9998% of overdose fatalities (Slavova et al., 2015), making it less likely that overdose risk after jail release was overestimated during the pandemic. Finally, we cannot rule-out seasonal variation in overdose mortality as source of bias, although a prior examination of the Massachusetts opioid-related death dataset did not detect this phenomenon [Dana Bernson, personal communication].

Release from incarceration is a high-risk period for opioid overdose, and the COVID-19 pandemic accelerated the rate of overdose deaths in many communities in the U.S. At the same time, concern about COVID-19 dissemination in jails led to judicial mandates for early releases, as occurred in Massachusetts. Despite its limitations, this study does not support that early releases to mitigate the spread of SARS-COV-2 in jails explained much, if any, of the observed increase in community overdose deaths during the pandemic. Although every preventable death is a tragedy, the number of post-release overdose deaths was small and the modest increase detected could be explained by the 6% secular increase in overdose mortality from 2019 to 2020 in the counties associated with the jails in this study (Massachusetts Department of Public Health, 2022). During future public health emergencies, concerns about post-release overdose among high-risk persons should not deter appropriate early releases from jail to reduce contagion.

5. Contributors

PDF and EE conceptualized the study, obtained funding, and developed the study protocol. PDF, EE, TJS, RR, RH and DW supervised data collection. PDF, DD, DB and DW developed the analysis plan. DD and DB linked the datasets and executed the analysis. PF, DD and PM wrote the first draft of the manuscript. All authors read and revised several drafts. All authors and the Massachusetts Department of Public Health approved the final manuscript.

CRediT authorship contribution statement

Peter D. Friedmann: Conceptualization, Funding acquisition, Methodology, Project administration, Formal analysis, Writing – original draft, Writing – review & editing. Devon Dunn: Data curation, Formal analysis, Writing – original draft, Writing – review & editing. Pryce Michener: Writing – original draft, Writing – review & editing. Dana Bernson: Data curation, Formal analysis, Writing – review & editing. Thomas J. Stopka: Project administration, Writing – review & editing. Ekaterina Pivovarova: Writing – review & editing. Warren J. Ferguson: Writing – review & editing. Rebecca Rottapel: Project administration, Data curation, Writing – review & editing. Randall Hoskinson: Project administration, Data curation, Writing – review & editing. Donna Wilson: Project administration, Formal analysis, Writing – review & editing. Elizabeth A. Evans: Conceptualization, Funding acquisition, Methodology, Project administration, Writing – review & editing.

Declaration of Competing Interest

No conflict declared.