Abstract
In response to the COVID-19 pandemic, jails were advised to reduce facility census, particularly the growing population of those with medical/behavioral health vulnerabilities that increased susceptibility to adverse outcomes. Although jail census decreased across the nation in the initial days to months following pandemic declaration, there are minimal data regarding the health status of those who remained in jail. The current investigation aspired to describe jail census trends before/since the onset of COVID-19 and offer snapshots of temporal changes and context for prevalence estimates of medical/behavioral health conditions in jail detainees from 2019 to 2023. Using a serial cross-sectional design, prescription information for individuals residing in 18 jails across the United States on June 30 of each respective year was extracted and categorized using MediSpan’s ontological system to determine prevalence estimates of prescribed agents/products. Although data evidenced an initial 31% census reduction (followed by gradual return to prepandemic rates), prescribing patterns for all major therapeutic drug classes steadily increased, with 10% more individuals prescribed at least one agent in 2023 than 2019. The largest increases were observed for behavioral health agents (e.g., 32.4% of the sample was prescribed psychotropic agents in 2023 compared with 25.7% in 2019). We provide considerations for future investigations.
Introduction
After the World Health Organization (WHO) and U.S. Department of Health and Human Services pronounced the novel coronavirus disease (COVID-19) a global pandemic and public health crisis in March 2020 (DHHS, 2020; WHO, 2020a), there was substantial concern about spread of infectious diseases in congregate environments such as jails and prisons (Marcum, 2020).
Concerns about disease susceptibility in correctional populations were further amplified given the disproportionate—and growing (Harzke & Pruitt, 2018; Lamb & Weinberger, 1998; Maruschak & Berzofsky, 2016; National Commission on Correctional Health Care [NCCHC], 2002; Torrey et al., 2010)—rate that incarcerated individuals experience chronic medical conditions (e.g., asthma, arthritis, diabetes, hypertension), accelerated aging, 1 and comorbid behavioral health conditions (e.g., severe mental illness, chemical addiction; Binswanger et al., 2009; Fazel & Baillargeon, 2011; Steadman et al., 2009; Veysey & Bichler-Robertson, 2004), as these conditions increased risk for adverse COVID-related outcomes (Centers for Disease Control & Prevention, 2023).
In response to stay-at-home mandates and social distancing advisements, jail administrators were imminently forced to examine institutional conditions and procedures that exacerbated spread of the highly infectious virus, such as overcrowding, medical vulnerabilities, environmental sanitation, and detainee movement (Abraham et al., 2020; James & Foster, 2020). To reduce the overall population, many jurisdictions rapidly executed decarceration strategies that aspired to balance public safety and public health needs, although, admittedly, there was widespread confusion and misguidance about who should be released, as well as nonuniformity in overall pandemic response (Abraham et al., 2020; Hooks & Sawyer, 2020).
In May 2020, the WHO, United Nations Office on Drugs and Crime, Joint United Nations Programme on HIV and AIDS, and United Nations Office of the High Commissioner for Human Rights issued a joint statement urging political leaders to “consider limiting the deprivation of liberty, including pretrial detention, to a measure of last resort” (WHO, 2020b). Specifically, these organizations advised that release efforts should target individuals at risk for adverse outcomes due to COVID-19, including older individuals, individuals with preexisting health conditions, and individuals whose release would not jeopardize public safety, per se.
With concerns about infection control and mass exposure in close-quartered correctional facilities (James & Foster, 2020)—coupled with overcrowding issues in many institutions—there was an urgent, nationwide response to reduce overall jail census. Decarceration, especially for the pretrial population, occurred through various pathways as follows: city ordinances, jail/sheriffs’ administrative directives, prosecutor discretion, bail reform, compassionate release, and/or court orders (Alexander et al., 2020; Henry, 2020; Miller & Blumstein, 2020). Although some release determinations were based on index charges (e.g., nonviolent offenses, probation violations, misdemeanors), others were released due to medical vulnerabilities and increased susceptibility of contracting COVID-19, including those aged 55 or older and/or presenting with preexisting conditions (Tronick et al., 2022).
According to a Vera Institute of Justice report (Kang-Brown et al., 2021), by midyear and fall 2020, jail census across the nation was reduced by 24% and 10%, respectively. However, knowing how many individuals were released from custody does not necessarily shed light on who was released—or rather, who was detained—in the wake of COVID-19. For example, were there reduced numbers of detained individuals with chronic medical conditions, especially those aged 55 and older? If so, does the same hold true for individuals who endure behavioral health conditions, given this population’s shared vulnerability to adverse health-related outcomes, or does experience of severe mental illness and/or chemical addiction serve as the basis for prolonged detention, as some have suggested (e.g., Tronick et al., 2022)?
Purpose of Present Study
Using a serial cross-sectional design and following STROBE guidelines (Vandenbroucke et al., 2007), the current descriptive study aims to (1) determine whether decarceration trends aligned with agency mandates/expectations, (2) explore temporal changes in prevalence estimates of medical and behavioral health conditions (broadly inferred based on prescribing patterns) in a sample of jail detainees on specific days in 2019 through 2023, and (3) describe trends in context factors such as facility size, location, and demography. Findings will guide subsequent research efforts to better understand the needs of this population and inform policy development, resource allocation, and clinical care.
Method
Data Source
This study relied on data from 18 county and municipal jails nationwide between 2019 and 2023. Deidentified information about all individuals residing in each facility on June 30 of each year was extracted from the facilities’ electronic health record (EHR). This date aligns with practice of the Bureau of Justice Statistics, which “has long used…the end of June to measure jail populations, based on historical evidence that a June measure is more representative of an average daily population” (Kang-Brown et al., 2021, p. 11).
Each facility held contractual agreements with NaphCare, Inc. for the provision of medical/behavioral health services during the specified time frame, and under a confidential data sharing agreement, information decoupled from identifiers (i.e., name, booking numbers, medical record numbers, dates of birth, dates of arrest, and Social Security numbers) was provided for analysis. In light of strategies to anonymize the archival dataset, Salus IRB (Case ID # 23102-01) determined this project did not constitute human subjects research requiring institutional review board approval.
All individuals residing in facilities between the hours of 0001 and 2359 on June 30 of each respective year were eligible, including those booking into or releasing from the jail and those whose detention began before the date of interest. When booked into a facility, individuals undergo medical and behavioral health screenings. Individuals endorsing acute or chronic health care needs are triaged for further assessment and treatment by medical and behavioral health providers. When clinically indicated, individuals are prescribed pharmaceutical agents/products to manage health care needs, and all agents/products (even those considered “over the counter”) must be prescribed in correctional settings. The current study relied on medications prescribed rather than administered to broadly identify the conditions requiring pharmacological intervention.
Data Extraction
Data extracted from the EHR were pipelined to an Amazon Web Services EC2 SQL instance. Databases deployed to this cloud-based solution allowed developers to extract predefined datapoints and securely transmit data transfer to a private instance, which required separate credentials (by a limited number of research team members) to access.
Data extraction tools included scripts/queries written in SQL and Python. SQL scripts extracted data for all individuals in custody on each date and yielded drug classification codes for all active prescriptions on the days of interest. SQL results were exported to CSV files. Python was used to encode demographic information from CSV files and combine data records. This process was first automated to minimize errors and ensure consistency, then executed site-by-site. Data were housed and versioned using the aforementioned secured private instance. Ethical and legal guidelines for accessing and handling data—including compliance with data protection regulations and obtaining necessary permissions/approvals when working with sensitive or confidential data—were followed.
Measures
Facility Indicators
Categorical variables were created to identify facility size and location. The American Jail Association (n.d.) offers a framework for jail size based on bed capacity as follows: small (1–49), medium (50–249), large (250–999), and mega (1,000+). No small jails were included in this sample. We elected to divide the “large” category into medium-large (250–499) and large (500–999) based on experience providing services in facilities of these sizes. The U.S. Census Bureau’s (2020 update) census regions—Northeast, Midwest, South, and West—served as the basis for coding location.
Clinical Indicators
The Master Drug Database (MDDB) is a coding and ontological system of pharmaceuticals developed and maintained by MediSpan (Saitwal et al., 2012). The MDDB includes a proprietary drug identification system (i.e., the Generic Product Identifier or GPI) to provide descriptive information about medication names (brand and generic), drug class, active ingredients, and so forth. In full form, the 14-digit GPI denotes drug groups, therapeutic class/subclass, and granular agent/product information (e.g., base name, dose form, and strength).
However, because some individuals may receive two different dosages of the same medication, we elected to extract a 10-digit GPI (to avoid artificial inflation of prevalence estimates) for each agent/product prescribed. These values were then coded at the therapeutic and drug class levels (see Table 1). Although others have used pharmaceutical data to approximate prevalence of medical conditions in the community (Cossman et al., 2010), this is the first study to examine prescription rates in jails.
Description of Generic Product Identifier (GPI) Therapeutic and Drug Classes
Only drug classes represented in the dataset are included here. A comprehensive list can be found at http://propharmoneplus.com/gpi-classification-system.html
Master Drug Database (MDDB) includes lithium as an antipsychotic medication; however, given the restricted use of this agent to treat mood instability associated with bipolar disorder, we considered lithium as an independent drug class for the current study.
Analyses
We first examined trends in census data to determine whether there was an overall decline in jail population as expected given the aforementioned decarceration recommendations, guidelines, and mandates. Then, we examined unadjusted rates of prescriptions by pooling data for all sites, grouping samples by year. Dichotomous indicators were created for therapeutic and agent/product classes to determine the number of unique individuals prescribed each class of medication. We documented the total number of agents prescribed for each therapeutic and agent class, as some individuals were prescribed multiple agents from the same category.
We then calculated prevalence estimates of prescribed agents for therapeutic and agent classes, as well as descriptive statistics regarding the demography—biological sex, age, and race/ethnicity—of individuals prescribed various therapeutic agents. SPSS (version 21) and R/Python were used to validate data and calculate descriptive statistics.
Results
Site/Facility Information
On June 30, 2019, facilities were, collectively, at 99.5% capacity (i.e., of the 17,704 “beds” available 2 , 17,607 were occupied). The following year, facilities operated at 68.7% capacity, an approximate 30.8% reduction in overall census. However, each subsequent year evidenced increases in jail census, and by 2023, facilities were 88.5% occupied. One facility in the South increased its capacity by 450 beds, so percentages from 2023 are based on total capacity of 18,154.
Regardless of size, all facilities demonstrated initial decreases in census. Medium-sized jails showed the greatest initial reduction (43.5%), while mega-sized jails evidenced an initial 29.1% reduction. In 2023, all facilities operated at lower capacity than in 2019; however, facilities of all sizes increased census compared with 2020 rates. Mega jails were over maximum census in 2019 and 2023.
Similarly, when considering location, Western facilities demonstrated the greatest initial census reduction (37.1%) from 2019 to 2020, whereas Northeastern facilities evidenced an initial 21.4% reduction. However, the Northeast operated at 70.8% capacity in 2019, whereas the West operated at 92.2%. The Midwest and South were over capacity in 2019 (i.e., 101.2% and 107.9%, respectively). Table 2 details census changes from 2019 to 2023 by facility size and region.
Jail Census and Capacity by Facility Size and Facility Location for 2019–2023
Total capacity for all facilities was 17,704 beds for years 2019 to 2022 and 18,154 beds for 2023.
Facility size based on bed capacity: medium = 50–249 (n = 2); medium-large = 250–499 (n = 3); large = 500–999 (n = 7); and mega = 1,000+ (n = 6).
Facility location coded according to the U.S. Census Bureau’s (2020) census regions: Midwest (n = 3); Northeast (n = 1); South (n = 4); and West (n = 10).
Demographics
The overall sample (N = 75,201) was predominantly male detainees (87.8%) who were, on average, 36.1 years old (SD = 11.6). Female detainees (12.2%) were, on average, 35.9 years old (SD = 10.4). There were minimal fluctuations in average age from 2019 to 2023, and geriatric detainees (i.e., 55 and older) accounted for approximately 9% of the sample each year with the exception of 2020 (i.e., 7.9%).
Race identification for 5.2% of the sample was not known. The number of male detainees identified as Black increased from 2019 (46.5%) to 2023 (51.7%), whereas the number of White male detainees decreased from 2019 (45%) to 2023 (40.2%). A similar trend was observed for females as follows: those identified as Black increased from 24.2% in 2019 to 37.8% in 2023, whereas females identified as White decreased from 66.7% in 2019 to 53.3% in 2023. Table 3 provides additional demographic information.
Demographics and Overall Agent/Product Prescription Patterns by Year
Biological sex data are imported into the electronic health record (EHR) by the jail management system (JMS) and cannot be altered by health care staff.
Race data are imported into the EHR by the JMS and cannot be altered by health care staff. Facilities that do not routinely capture race are represented as “Unknown” category. “Other” category includes individuals identified as Indigenous (e.g., Alaska Native, American Indian, Asian, Pacific Islander, South Asian), Middle Eastern/North Africa (MENA), and multiracial.
See Diaz et al. (2021), for discussion of age categories in 5-year intervals.
Percentage based on number of individuals prescribed at least one agent/product.
Prescribing Patterns of Agents/Products by Therapeutic and Drug Class
As described, data show census decreases in June 2020, then gradual increases in total population census each subsequent year. Although facilities still operated below census in June 2023, there was an overall increase from 2019 to 2023 in the number of individuals prescribed at least one agent/product (i.e., 47.5% to 57.7%, respectively). On average, individuals across years were prescribed approximately three agents/product each, although the number of individuals prescribed polypharmacy (i.e., five or more medications; Varghese et al., 2023) increased from 18.9% in 2019 to 21.2% in 2023. There were minimal changes in the average number of targeted conditions/organ systems from 2019 to 2023 (2.0 and 2.2, respectively) or the range of number of targeted systems (1 to 9 both years). See Table 3.
The most frequently prescribed agents across all years fell in the central nervous system. (CNS), cardiovascular, and neuromuscular therapeutic classes, with 25.7%, 16.6%, and 11.0% (respectively) of detainees prescribed medications in these classes in 2019. Increases in prescription frequencies were observed in every therapeutic class compared with 2019. The largest increase was observed in the CNS therapeutic class, increasing from 25.7% of the sample in 2019 to 32.5% in 2023.
To better understand the clinical landscape of behavioral health conditions and symptoms treated, we extracted drug class and agent-specific data for these therapeutic classes. As shown in Table 4, the frequency of CNS-antipsychotic prescriptions increased from 13.5% in 2019 to 17.7% in 2023; CNS-antidepressant prescriptions increased from 18.7% (2019) to 22.9% (2023); and CNS-antianxiety prescriptions increased from 8.6% (2019) to 10.5% (2023).
Prevalence Estimates (Percentages) and Demography of Individuals Prescribed Select Pharmaceutical Agents/Products by Year
Note: Therapeutic and drug classes are reported as frequency counts and percentage of total sample for respective year. Individuals may be prescribed more than one medication from a particular therapeutic class and may be represented in multiple drug class categories. Demographic data (i.e., age, sex, and race) are reported as frequency counts and percentage of unique individuals prescribed an agent from the respective therapeutic class. Given the infrequency of Latinx representation in therapeutic classes, those individuals’ data were included in the “Other” category. Dashes (“—”) indicate that there were no individuals meeting the respective demographic criterion.
The next largest prescription increases occurred within the analgesic/anesthetics and neuromuscular therapeutic classes, increasing 3.6% and 3.4%, respectively. In 2019, approximately 21.1% of individuals prescribed analgesic/anesthetics were prescribed a narcotic agent. Of those prescribed narcotics, 91.7% were medication-assisted treatment (MAT) medications (e.g., buprenorphine, methadone, naltrexone, sublocade, and suboxone) used to safely withdraw and/or maintain individuals with opioid use disorders. By comparison, in 2023, 33.2% of prescribed analgesic/anesthetics were narcotics, with 98.7% of those classified as MAT agents.
In 2019 and 2023, the most frequently prescribed neuromuscular agents were anticonvulsants, accounting for 72.4% and 72.7%, respectively, of prescriptions in this therapeutic class. However, only approximately 34.0% and 28.4% of individuals prescribed an anticonvulsant (in 2019 and 2023, respectively) were identified as having a history of seizure events/disorder. In the same years, approximately 19.5% and 21.9% of individuals prescribed an anticonvulsant agent with no seizure history were prescribed levetiracetam (used to safely manage risk for alcohol or benzodiazepine withdrawal).
In 2019, approximately 34.4% of individuals prescribed an anticonvulsant (other than levetiracetam) did not have a seizure history, suggesting off-label use of these medications for mood stabilization and/or substance use disorder (Donovan & Nunes, 1998; Small, 1990). This rate decreased to 28.4% in 2023.
Discussion
As expected, population census data align with previous findings describing initial—and substantial—decreases in jail populations, then gradual, steady return to almost prepandemic levels (Kang-Brown et al., 2021, 2023; Prison Policy Initiative, 2023). Despite the reduction in population census, data contained herein show an overall increase in prescribed medications across all major therapeutic categories. As noted, the greatest increases occurred in drug categories often used to manage behavioral health concerns, such as severe mental illness and chemical addiction. The notable rise in CNS prescriptions from 2019 to 2023 underscores an increasing recognition and response to mental health issues within jail populations.
The pandemic’s impact on jail populations, including decarceration efforts and changes in arrest policies, may have influenced the detainee population and their health needs (e.g., the increased medication needs evidenced herein may reflect heightened stress and health concerns). In addition, decades of research have pointed to an increasing frequency of medical and behavioral health conditions in the inmate-patient population (Harzke & Pruitt, 2018; Lamb & Weinberger, 1998; Maruschak & Berzofsky, 2016; NCCHC, 2002; Torrey et al., 2010). Even before the pandemic in 2019, 25% of individuals in the current study were prescribed CNS medications, exceeding the estimated 16.5% of noninstitutionalized individuals who reported taking prescription medication for mental health in 2020 (Terlizzi & Norris, 2021).
Although trends observed within our data align with broader national and global concerns about the COVID-19 pandemic’s impact on behavioral health (Substance Abuse and Mental Health Services Administration, 2023; WHO, 2022), there is a dearth of robust, publicly disseminated data regarding frequency of behavioral health conditions in jail detainees. Existing literature relies on a heterogenous amalgamation of samples, conditions, and measures to infer prevalence estimates of behavioral health conditions.
Limited pre-COVID-19 data prevent identifying the pandemic as a clear catalyst for increases in behavioral health pharmacological interventions. Nevertheless, data herein suggest a growing need for mental health and substance abuse services in correctional facilities since at least 2019, increased number of individuals seeking care in the wake of a pandemic, and/or improved care and access to such services in jails as evidenced by increased prescription frequencies.
Limitations
Because “the health care infrastructure within correctional facilities frequently creates barriers, limiting access to medical care” (Novisky et al., 2021, p. 1632), our data may underrepresent the extent and severity of medical and behavioral health conditions due to access to care, rapid turnover/release from custody without first obtaining clinical information, and bias in self-report (i.e., underreporting) during screenings. Thus, data herein may underemphasize the growing need and demand for medical and behavioral health services in correctional environments, which, in turn, may impact resource allocation for health care services.
We are also cautious about the overall increase in prescribing patterns without confirmatory diagnostic indicators for care. We recognize that many medications may be continued upon booking based on prescriptions from the community without independent evaluation of need or efficacy. Many medications may have off-label benefit that skewed our ability to infer certain conditions without corroborating diagnosis based on therapeutic or drug class alone.
Finally, we recognize our findings are based on one company’s policies and procedures, and data may not generalize to organizations following other operational strategies. However, prescribers maintain independent clinical discretion to order medically appropriate/necessary medications. Although there should be a modicum of universality in prescribed treatments, we anticipate as much variation in prescribing practices as there are providers.
Nevertheless, the current findings are the first to be generated from a large-scale, nationwide dataset obtained through provision of routine correctional health care instead of relying on public health surveys to estimate prevalence.
Future Directions
Further investigations are warranted to better understand the health care needs of jail populations. The health of detainees is intrinsically linked to broader public health concerns. Improved services in jails could impact public health positively when considering the rates of recidivism and the transitional nature of jail populations. We also detected an increase in the proportion of Black detainees, highlighting concerns about racial dynamics broadly regarding arrest rates and emphasizing the need for culturally competent health services in jails.
Conclusion
The comprehensive data from this multisite study provide a snapshot into the changing landscape of prescribing patterns for medical and behavioral health conditions among jail detainees from 2019 to 2023, suggesting an evolving pattern of health care needs within this population.
In conclusion, this study illuminates the dynamic nature of health care needs in jail populations—needs that have steadily increased over the past several decades and may have been accelerated by the COVID-19 pandemic. The findings highlight the importance of continuous monitoring and updating of health services in jails to adapt to the evolving demographic and health profiles of detainees. The increased pharmacological treatment for mental health conditions and the need for comprehensive, tailored health care approaches are evident.
These insights are vital for policymakers, health care providers, and correctional administrators to improve the quality of health care in jails and address the unique challenges faced by this population. As we move forward, a focused approach on mental health, addressing health care disparities and ensuring adequate health care resources, is critical to meet the diverse needs of the jail detainee population.
Footnotes
Authors’ Contributions
A.H.S.: conceptualization, methodology, project administration, formal analysis, data curation, and writing of the original article. W.J.: data curation, software, supervision, writing—review and editing, and project administration. A.A.: writing, review, and editing of the original article. Y.A.P.: data curation and software.
Author Disclosure Statement
The authors disclosed no actual or potential conflicts of interest with respect to the research, authorship, or publication of this article.
Funding Information
The authors received no financial support for the research, authorship, or publication of this article.