Differences in Hospitalization and Inpatient Death Patterns by Incarceration Status in 31 US Jurisdictions, 2021

Abstract

Objective:

Population-based studies that examine both hospitalization and inpatient death patterns by incarceration status are sparse. We sought to compare hospitalization and inpatient death patterns by incarceration status in the United States.

Methods:

In this retrospective study of adults aged 18 to 64 years, we used the 2021 State Inpatient Database files from 31 US jurisdictions, which included discharges from general acute care hospitals. We compared hospitalization and inpatient death patterns among incarcerated adults versus nonincarcerated adults using adjusted rate ratios (ARRs) and 95% CIs, estimated with negative binomial regression. We examined discharges overall and discharges by nonoverlapping hospital service line.

Results:

The study population included 6.3 million hospital discharges (incarcerated, 1.1%; mean [SD] age, 47.5 [12.7] y; women, 46.8%; non-Hispanic White race and ethnicity, 36.6%), with 2.2% inpatient deaths. For discharges overall, incarcerated adults had higher rates of hospitalization (ARR = 2.9; 95% CI, 2.2-3.8) than nonincarcerated adults, which was driven mainly by higher rates of inpatient admissions classified as mental health/substance use (ARR = 11.8; 95% CI, 8.5-16.8) and injury (ARR = 1.9; 95% CI, 1.6-2.2) among incarcerated adults than among nonincarcerated adults. However, surgical (ARR = 0.6; 95% CI, 0.5-0.6) and medical (ARR = 0.8; 95% CI, 0.7-0.9) admissions were lower among incarcerated adults than among nonincarcerated adults. Overall, incarcerated adults had lower inpatient mortality during their length of stay (ARR = 0.4; 95% CI, 0.4-0.5) than nonincarcerated adults, which was consistent by hospital service line.

Conclusions:

Considering hospitalization and mortality together offers a clear view of health care use among incarcerated adults and underscores the need for integrated correctional–hospital data systems to inform public health practice and policy.

In 2024, the incarcerated population totaled approximately 11.5 million worldwide, including 1.8 million in the United States. ¹ Despite these numbers, data on health care use and health care outcomes are not routinely collected and reported across jurisdictions.^2,3 Nevertheless, researchers have examined the effects of incarceration, including health-related outcomes among incarcerated adults compared with nonincarcerated adults. ⁴ Other studies have used alternative control groups to better distinguish between the effects of incarceration and preincarceration behaviors, environments, and access to care.^5,6 For example, Norris et al ⁵ compared mortality outcomes between incarcerated and nonincarcerated criminally charged defendants. Ecological studies have also linked incarceration rates with components of health care infrastructure, such as the number of hospital beds. Using data from 36 countries, Testa et al ⁷ found that increases in national incarceration rates were associated with declines in the number of hospital beds per capita over time. In explaining this inverse relationship, the authors suggested that a social control focus might lead to more public safety investment (eg, prisons, incarceration), whereas a social support focus might lead to more public health investment (eg, hospitals, bed capacity). Thus, with a growing population, fewer hospital beds will be available over time, depending on social priorities. Schnittker et al ⁸ found that US states with a high percentage of previously incarcerated individuals also had high rates of uninsurance and use of emergency department services. Moreover, for a given hospital bed capacity, studies have shown that hospitalization rates are higher among incarcerated populations than among nonincarcerated populations.^9-12

While increased health care use may indicate increased access to care, Kouyoumdjian et al ¹¹ suggested that high rates of hospitalization among incarcerated individuals may reflect elevated morbidity and poor access to quality health care, based on cohort data from Ontario, Canada. However, in a subsequent study in the same province, Kouyoumdjian et al ¹³ found that the adjusted 30-day readmission rate was lower but not significantly so among incarcerated individuals than among nonincarcerated individuals. Importantly, if incarcerated adults had higher disease acuity and/or received less effective clinical care than nonincarcerated adults, one might expect not only a higher 30-day readmission rate but also a higher case fatality rate (ie, number of deaths due to a disease for a period of time divided by the number of cases diagnosed with the disease during that time), assuming no other underlying differences. ¹⁴ As a rough short-term proxy for case fatality rates, other studies have compared in-hospital death rates by incarceration status, especially since the COVID-19 pandemic, using different methodologies and producing mixed results.^15-18

Most of these pandemic-era studies of inpatient deaths by incarceration status focused on COVID-19 admissions.^15-17 As others have reported, earlier studies based on electronic health records were unable to distinguish between primary COVID-19 hospitalizations and incidental SARS-CoV-2 infections among patients hospitalized for other reasons.^19-21 The higher COVID-19 testing rates and the presence of on-site staff trained in cardiopulmonary resuscitation (CPR) and first aid at correctional facilities likely enhanced the triage and prompt hospital referrals of incarcerated patients.^22,23 Therefore, the misattribution of primary COVID-19 admissions and the failure to account for prehospital acuity likely contributed to conflicting reports. In a comparison of 32 state prisons with their statewide general populations, prison populations had a higher COVID-19 case rate (ie, cases per population) and death rate (ie, deaths per population) than the general population; however, the estimated case fatality rate (ie, death rate per case rate) was lower in the prison population than in the general population. ²²

Notably, prior studies of inpatient deaths did not consider hospitalization rates, which could offer additional insights. Considering both inpatient deaths and hospitalization rates may help separate the effects of access to care from those of clinical acuity that might be influenced by incarceration. In this study, our aim was to compare hospitalization and inpatient death patterns between incarcerated and nonincarcerated adults in 2021, the second year of the COVID-19 pandemic, when inpatient care for COVID-19 was more established and COVID-19 vaccines had just become available in the United States. To reduce the potential for misattribution bias and provide a broader context, we examined hospitalizations overall and by major hospital service line.

Methods

Study Population

In this retrospective cohort study, we used data from the 2021 State Inpatient Databases (SID), Healthcare Cost and Utilization Project, through the Agency for Healthcare Research and Quality (AHRQ). ²⁴ SID files include data on discharges from general acute care hospitals, including Veterans Health Administration and Indian Health Service hospitals, from 31 US jurisdictions: Alaska, Arizona, Arkansas, Colorado, Delaware, District of Columbia, Florida, Georgia, Hawaii, Indiana, Iowa, Kansas, Kentucky, Maryland, Massachusetts, Michigan, Minnesota, Mississippi, Missouri, New Jersey, New Mexico, New York, North Carolina, Oregon, Rhode Island, South Dakota, Utah, Vermont, Washington, West Virginia, and Wisconsin. We selected the SID files based on what was available at the time of our data request to AHRQ in March 2025. We considered the National Inpatient Sample file, also maintained by AHRQ, which is designed to be nationally representative of nonfederal acute care general hospitals. ²⁵ However, since 2012, the National Inpatient Sample has combined “court/law enforcement” and “home” as a single category for discharge disposition, which prevents the separate identification of incarcerated individuals. For our study population, we included adults hospitalized for at least 24 hours for nonmaternal/neonatal causes. We limited our study population to adults aged 18 to 64 years to (1) lessen the potential effects of postdischarge deaths that might occur more frequently among older adults and (2) reduce differences in age distributions between the incarcerated and nonincarcerated groups. We also limited our study population to the following major racial and ethnic groups, based on available data routinely reported by the US Department of Justice for prisons and jails: Hispanic, non-Hispanic American Indian/Alaska Native, non-Hispanic Asian/Pacific Islander, non-Hispanic Black, and non-Hispanic White.

Study Measures

For outcome measures, we used 3 negative binomial regression models to estimate rates: (1) hospital discharges per population, (2) inpatient deaths per length of stay (days), and (3) inpatient deaths per population. For the population denominators, we used multiple sources, including the US Census Bureau, the US Department of Justice, and a compilation of secondary data.^26-29 When cross-tabulated frequencies for age, sex, and race and ethnicity groups were not given, we assumed statistical independence to derive estimates (eg, probability [age group 1 AND female AND race/ethnicity group 1] = probability [age group 1] × probability [female] × probability [race/ethnicity group 1]). The primary comparison was current incarceration status, which we defined by using the point-of-origin/disposition fields listed as “court/law enforcement” or ICD-10 (International Classification of Diseases, Tenth Revision ³⁰ ) codes listed as Z65.1 (imprisonment and other incarceration) or Y92.14* (prison as the place of occurrence of the external cause). Covariates included age, sex, and race and ethnicity. In addition, we examined discharges overall and by hospital service line (ie, mental health/substance use, injury, surgical, medical). AHRQ uses a hierarchical service line variable with a single code per hospital discharge, based on the principal discharge diagnosis. Therefore, there is no overlap in the service line codes.

Data Analysis

We used weighted Pearson χ ² tests and t tests to compare categorical and continuous variables, respectively, by incarceration status. We defined significance as 2-sided P < .05. For our multivariate analysis, we compared currently incarcerated adults with nonincarcerated adults by using adjusted rate ratios (ARRs) and 95% CIs, estimated with negative binomial regression models because preliminary results suggested overdispersion in the data. ³¹ To account for potential clustering of discharges at the state level, we estimated SEs while incorporating a cluster indicator variable. We defined ARRs as significant when the corresponding 95% CIs did not include 1. We evaluated model fits by using pseudo-R ² (higher values indicating better fit), Akaike information criterion (lower is better), and Bayesian information criterion (lower is better). We performed all analyses using Stata version 18 (StataCorp LLC).

Ethical Considerations

For the analysis, we used only deidentified, publicly available data; therefore, institutional review board approval was not required based on AHRQ’s determination that Healthcare Cost and Utilization Product databases are consistent with the definition of limited datasets under the Health Insurance Portability and Accountability Act of 1996 Privacy Rule and contain no direct patient identifiers. We present only aggregate results; therefore, informed consent was not required.

Results

The study population included approximately 6.3 million hospital discharges (1.1% incarcerated; mean [SD] age, 47.5 [12.7] y; 46.8% women; 63.6% non-Hispanic White, 21.5% non-Hispanic Black, 11.6% Hispanic, 2.2% non-Hispanic Asian/Pacific Islander, and 1.3% non-Hispanic American Indian/Alaska Native) (Table 1), with 2.2% inpatient deaths among them. Compared with nonincarcerated adults, incarcerated adults were more likely to be younger, be male, be non-Hispanic White, have inpatient service lines classified as mental health/substance use and injury, have longer lengths of stay, and have few inpatient deaths (P < .01 for all variables). For the surgical inpatients who died, the most common principal discharge diagnosis was sepsis (ICD-10 code A419) for both incarcerated (15.7%) and nonincarcerated (13.4%) people. For the medical inpatients who died, the most common principal discharge diagnosis was COVID-19 (ICD-10 code U071) for both incarcerated (21.2%) and nonincarcerated (23.3%) people.

Table 1.

Demographic and hospital discharge characteristics of study population of nonincarcerated adults versus incarcerated adults aged 18 to 64 years, United States, 2021 ^a

Characteristic	No. (%) of adults aged 18 to 64 years			P value b
	Nonincarcerated	Incarcerated	Total
No.	6 270 018	68 856	6 338 874	—
Age group, mean (SD), y	47.6 (12.7)	39.4 (12.6)	47.5 (12.7)	<.001
Sex				<.001
Male	3 323 901 (53.0)	49 921 (72.5)	3 373 822 (53.2)
Female	2 946 117 (47.0)	18 935 (27.5)	2 965 052 (46.8)
Race and ethnicity				<.001
Hispanic	731 128 (11.7)	6 959 (10.1)	738 087 (11.6)
Non-Hispanic American Indian/Alaska Native	78 959 (1.3)	1248 (1.8)	80 207 (1.3)
Non-Hispanic Asian/Pacific Islander	137 988 (2.2)	924 (1.3)	138 912 (2.2)
Non-Hispanic Black	1 346 601 (21.5)	19 212 (27.9)	1 365 813 (21.5)
Non-Hispanic White	3 975 342 (63.4)	40 513 (58.8)	4 015 855 (63.4)
Hospital service line				<.001
Mental health/substance use	915 543 (14.6)	40 227 (58.4)	955 770 (15.1)
Injury	327 741 (5.2)	5841 (8.5)	333 582 (5.3)
Surgical	1 473 390 (23.5)	5197 (7.5)	1 478 587 (23.3)
Medical	3 553 344 (56.7)	17 591 (25.5)	3 570 935 (56.3)
Length of stay, mean (SD), d	6.0 (8.8)	9.6 (19.6)	6.0 (9.0)	<.001
Inpatient mortality				<.001
Survived	6 128 218 (97.7)	68 413 (99.4)	6 196 631 (97.8)
Died	141 800 (2.3)	443 (0.6)	142 243 (2.2)

^aData source: 2021 State Inpatient Databases. ²⁴ Hospital discharge data were used from 31 jurisdictions: Alaska, Arizona, Arkansas, Colorado, Delaware, District of Columbia, Florida, Georgia, Hawaii, Indiana, Iowa, Kansas, Kentucky, Maryland, Massachusetts, Michigan, Minnesota, Mississippi, Missouri, New Jersey, New Mexico, New York, North Carolina, Oregon, Rhode Island, South Dakota, Utah, Vermont, Washington, West Virginia, and Wisconsin.

^bComparison between nonincarcerated and incarcerated populations; significant at P < .05 for the weighted Pearson χ ² tests and t tests.

For hospital discharges overall, compared with nonincarcerated adults, incarcerated adults had higher discharges per population (ARR = 2.9; 95% CI, 2.2-3.8) (model 1), lower inpatient deaths per length of stay (ARR = 0.4; 95% CI, 0.4-0.5) (model 2), and lower inpatient deaths per population (ARR = 0.4; 95% CI, 0.3-0.5) (model 3), after adjusting for demographic factors (Table 2, Figure). The hospital service line–specific results were qualitatively similar, except that incarcerated adults had lower surgical (ARR = 0.6; 95% CI, 0.5-0.6) (model 1) and medical (ARR = 0.8; 95% CI, 0.7-0.9) (model 1) admissions than nonincarcerated adults, with nonsignificant differences in mental health/substance use inpatient deaths per population (ARR = 1.7; 95% CI, 0.7-4.0) (model 3).

Table 2.

Comparison of incarcerated adults versus nonincarcerated adults aged 18 to 64 years using adjusted rate ratios (ARRs) and 95% CIs, estimated with negative binomial regression models, by hospital service line, United States, 2021 ^a

Characteristic	ARR (95% CI)
	Model 1 b	Model 2 c	Model 3 d
Overall
Incarceration status
Incarcerated	2.89 e (2.18-3.83)	0.39 e (0.30-0.50)	0.42 e (0.33-0.54)
Nonincarcerated	1 [Reference]	1 [Reference]	1 [Reference]
Race and ethnicity
Hispanic	0.66 e (0.61-0.72)	0.96 (0.85-1.10)	0.74 e (0.63-0.86)
Non-Hispanic American Indian/Alaska Native	1.74 e (1.49-2.04)	1.13 e (1.00-1.27)	2.56 e (2.12-3.10)
Non-Hispanic Asian/Pacific Islander	0.98 (0.66-1.45)	0.95 (0.87-1.03)	0.52 e (0.41-0.67)
Non-Hispanic Black	1.25 e (1.17-1.34)	0.83 e (0.77-0.88)	1.40 e (1.26-1.55)
Non-Hispanic White	1 [Reference]	1 [Reference]	1 [Reference]
Sex
Female	1.42 e (1.26-1.59)	0.76 e (0.73-0.79)	0.66 e (0.62-0.69)
Male	1 [Reference]	1 [Reference]	1 [Reference]
Age group, y
18-44	1 [Reference]	1 [Reference]	1 [Reference]
45-64	1.62 e (1.52-1.74)	1.71 e (1.64-1.79)	4.92 e (4.58-5.29)
Pseudo R 2	0.021	0.018	0.047
Akaike information criterion	48 586.2	16 272.5	17 611.3
Bayesian information criterion	48 642.8	16 329.1	17 667.9
Mental health/substance use
Incarceration status
Incarcerated	11.93 e (8.47-16.79)	0.26 e (0.12-0.55)	1.68 (0.72-3.96)
Nonincarcerated	1 [Reference]	1 [Reference]	1 [Reference]
Race and ethnicity
Hispanic	0.49 e (0.43-0.56)	0.85 (0.59-1.22)	0.51 e (0.36-0.73)
Non-Hispanic American Indian/Alaska Native	1.41 e (1.11-1.79)	1.98 e (1.16-3.38)	3.32 e (1.92-5.72)
Non-Hispanic Asian/Pacific Islander	1.04 (0.63-1.71)	0.56 (0.29-1.06)	0.19 e (0.10-0.35)
Non-Hispanic Black	1.11 e (1.01-1.23)	0.66 e (0.46-0.95)	1.04 (0.73-1.50)
Non-Hispanic White	1 [Reference]	1 [Reference]	1 [Reference]
Sex
Female	1.50 e (1.33-1.70)	0.43 e (0.34-0.55)	0.32 e (0.26-0.41)
Male	1 [Reference]	1 [Reference]	1 [Reference]
Age group, y
18-44	1 [Reference]	1 [Reference]	1 [Reference]
45-64	0.67 e (0.61-0.73)	4.52 e (3.38-6.04)	3.89 e (2.91-5.21)
Pseudo R 2	0.082	0.157	0.166
Akaike information criterion	12 508.3	911.5	911.2
Bayesian information criterion	12 553.1	956.3	956.0
Injury
Incarceration status
Incarcerated	1.86 e (1.56-2.22)	0.28 e (0.20-0.40)	0.32 e (0.22-0.47)
Nonincarcerated	1 [Reference]	1 [Reference]	1 [Reference]
Race and ethnicity
Hispanic	0.70 e (0.62-0.79)	0.78 e (0.67-0.91)	0.64 e (0.53-0.78)
Non-Hispanic American Indian/Alaska Native	1.75 e (1.48-2.06)	0.81 (0.63-1.04)	1.92 e (1.48-2.49)
Non-Hispanic Asian/Pacific Islander	0.37 e (0.31-0.43)	0.82 e (0.69-0.96)	0.33 e (0.22-0.48)
Non-Hispanic Black	1.25 e (1.11-1.42)	0.86 e (0.79-0.95)	1.64 e (1.42-1.90)
Non-Hispanic White	1 [Reference]	1 [Reference]	1 [Reference]
Sex
Female	0.60 e (0.56-0.64)	0.84 e (0.79-0.89)	0.36 e (0.34-0.39)
Male	1 [Reference]	1 [Reference]	1 [Reference]
Age group, y
18-44	1 [Reference]	1 [Reference]	1 [Reference]
45-64	1.08 e (1.01-1.15)	0.97 (0.91-1.05)	1.45 e (1.32-1.59)
Pseudo R 2	0.072	0.039	0.144
Akaike information criterion	8428.3	2520.6	2709.5
Bayesian information criterion	8471.8	2564.1	2753.0
Surgical
Incarceration status
Incarcerated	0.56 e (0.49-0.63)	0.60 e (0.45-0.79)	0.32 e (0.22-0.45)
Nonincarcerated	1 [Reference]	1 [Reference]	1 [Reference]
Race and ethnicity
Hispanic	0.70 e (0.63-0.78)	1.05 (0.96-1.15)	0.82 e (0.71-0.94)
Non-Hispanic American Indian/Alaska Native	1.55 e (1.37-1.76)	1.39 e (1.17-1.65)	2.93 e (2.36-3.63)
Non-Hispanic Asian/Pacific Islander	0.48 e (0.41-0.56)	1.17 e (1.08-1.27)	0.67 e (0.52-0.85)
Non-Hispanic Black	1.14 e (1.07-1.22)	1.06 e (1.00-1.11)	1.85 e (1.66-2.07)
Non-Hispanic White	1 [Reference]	1 [Reference]	1 [Reference]
Sex
Female	1.10 e (1.05-1.15)	0.81 e (0.78-0.84)	0.63 e (0.60-0.66)
Male	1 [Reference]	1 [Reference]	1 [Reference]
Age group, y
18-44	1 [Reference]	1 [Reference]	1 [Reference]
45-64	2.73 e (2.63-2.84)	1.55 e (1.48-1.62)	5.35 e (4.99-5.73)
Pseudo R 2	0.099	0.076	0.185
Akaike information criterion	9856.4	3435.8	3869.2
Bayesian information criterion	9900.0	3479.5	3912.9
Medical
Incarceration status
Incarcerated	0.79 e (0.68-0.91)	0.60 e (0.48-0.75)	0.39 e (0.32-0.48)
Nonincarcerated	1 [Reference]	1 [Reference]	1 [Reference]
Race and ethnicity
Hispanic	0.70 e (0.63-0.77)	0.99 (0.91-1.07)	0.73 e (0.64-0.84)
Non-Hispanic American Indian/Alaska Native	1.68 e (1.43-1.97)	1.35 e (1.17-1.56)	2.94 e (2.41-3.59)
Non-Hispanic Asian/Pacific Islander	0.52 e (0.43-0.62)	1.02 (0.95-1.10)	0.53 e (0.41-0.68)
Non-Hispanic Black	1.38 e (1.28-1.49)	0.80 e (0.77-0.84)	1.54 e (1.40-1.70)
Non-Hispanic White	1 [Reference]	1 [Reference]	1 [Reference]
Sex
Female	1.04 (0.99-1.10)	0.84 e (0.82-0.86)	0.70 e (0.67-0.72)
Male	1 [Reference]	1 [Reference]	1 [Reference]
Age group, y
18-44	1 [Reference]	1 [Reference]	1 [Reference]
45-64	2.31 e (2.18-2.45)	1.95 e (1.85-2.04)	5.68 e (5.27-6.12)
Pseudo R 2	0.062	0.114	0.169
Akaike information criterion	12 412.0	5040.1	5682.5
Bayesian information criterion	12 456.6	5084.7	5727.0

^aData source: 2021 State Inpatient Databases. ²⁴ Hospital discharge data were used from 31 jurisdictions: Alaska, Arizona, Arkansas, Colorado, Delaware, District of Columbia, Florida, Georgia, Hawaii, Indiana, Iowa, Kansas, Kentucky, Maryland, Massachusetts, Michigan, Minnesota, Mississippi, Missouri, New Jersey, New Mexico, New York, North Carolina, Oregon, Rhode Island, South Dakota, Utah, Vermont, Washington, West Virginia, and Wisconsin.

^bModel 1: discharges per population.

^cModel 2: inpatient deaths per length of stay (days).

^dModel 3: inpatient deaths per population.

^eSignificant because 95% CI excludes 1.

Figure.

Adjusted rate ratios (ARRs) and 95% CIs (error bars) for incarcerated adults versus nonincarcerated adults aged 18 to 64 years estimated with negative binomial regression models, by hospital service line, United States, 2021. All models were adjusted for age, sex, and race and ethnicity; rate ratios >1 indicate higher rates for incarcerated adults than for nonincarcerated adults; ARRs <1 indicate lower rates for incarcerated adults than for nonincarcerated adults; 95% CIs excluding 1 indicate significant differences in ARRs. Data source: 2021 State Inpatient Databases. ²⁴ Hospital discharge data were used from 31 jurisdictions: Alaska, Arizona, Arkansas, Colorado, Delaware, District of Columbia, Florida, Georgia, Hawaii, Indiana, Iowa, Kansas, Kentucky, Maryland, Massachusetts, Michigan, Minnesota, Mississippi, Missouri, New Jersey, New Mexico, New York, North Carolina, Oregon, Rhode Island, South Dakota, Utah, Vermont, Washington, West Virginia, and Wisconsin.

For other predictors, non-Hispanic American Indian/Alaska Native people, women, and older adults had the highest rate of admissions overall (Table 2, model 1). For inpatient deaths per length of stay overall (model 2), risk of inpatient death was highest among non-Hispanic American Indian/Alaska Native people, men, and older adults. While the patterns varied by hospital service line, they were generally consistent with the overall results, except for inpatient admissions classified as injury, which were higher among men than among women (model 1), and no elevated risk of inpatient deaths per length of stay for non-Hispanic American Indian/Alaska Native people (model 2).

Discussion

In this retrospective study of data from 31 US jurisdictions, we examined both hospitalization and inpatient death patterns among adults aged 18 to 64 years by incarceration status. After adjusting for demographic factors, we found higher hospitalization rates among incarcerated adults than among nonincarcerated adults, which was driven mainly by admissions classified as mental health/substance use and injury service lines rather than surgical or medical admissions. We also found that, overall, incarcerated adults had lower inpatient mortality rates during their length of stay than nonincarcerated adults, which was consistent across all hospital service lines. Our findings suggest that while incarcerated adults may have risk factors that increase their use of hospital services compared with nonincarcerated adults, their inpatient mortality rates were lower than those of nonincarcerated adults.

In one of the most carefully conducted studies estimating the causal effect of incarceration on mortality in the United States, Norris et al ⁵ compared incarcerated defendants with criminally charged, nonincarcerated defendants, using administrative data from Ohio. They found that incarceration reduced mortality risk among incarcerated individuals by 61% overall, driven primarily by large decreases in deaths from homicide (99%), overdose (53%), and natural causes (28%), including heart disease (43%). In discussing their findings, which run counter to conventional wisdom, the authors highlighted bias in earlier studies that used the general population as a control group, because that approach does not account for the risky behaviors and environments that existed before individuals were incarcerated. Indeed, in explaining their findings, they suggested that incarcerated individuals may have less access to firearms and illicit drugs than their criminally charged, nonincarcerated counterparts but more access to health care services. Our findings are consistent with that Ohio-based study in several ways. We found more hospital use for mental health/substance use and injury among incarcerated adults than among nonincarcerated adults, which might reflect the higher risk among incarcerated adults from overdose, self-harm, and interpersonal violence that warranted a referral to a higher level of care. On the other hand, lower-than-expected hospital use for surgical/medical care among incarcerated adults than among nonincarcerated adults might reflect access to in-house health care services (eg, wound care, chronic disease management).

In a Swedish study, Hjalmarsson and Lindquist ⁶ examined the health effects of prison reforms that kept sentence lengths fixed but increased the proportion of the sentence that incarcerated individuals had to serve. They found that longer time served reduced postrelease mortality risk, especially suicide among people with a history of psychiatric hospitalization before incarceration and circulatory diseases among those aged ≥33 years at the time of incarceration. The authors also noted that in-prison health care use and program participation increased with time served, which they suggested was the major mechanism for the observed mortality reductions. Thus, increased length of stay in the prison system may have increased opportunities for accessing on-site health care services. That Swedish study may help to better explain our findings. In the United States, prisons generally house people sentenced to >1 year, whereas jails house people who are unsentenced or sentenced to <1 year. Because the SID files do not report the length of incarceration, we were unable to distinguish between shorter-term and longer-term stays. Nevertheless, the risks of alcohol, benzodiazepine, and opioid withdrawal, as well as violence, are likely greater early rather than later in a person’s incarceration.^32,33 Therefore, the higher rates of inpatient admissions classified as mental health/substance use and injury might be more representative of those with shorter-term (vs longer-term) incarcerations, especially if such behaviors influenced their arrest, whereas the lower rates of inpatient admissions classified as surgical and medical might be more representative of those with longer-term (vs shorter-term) incarcerations. Together, these patterns might reflect the differing health care needs of incarcerated patients.

Some survey-based studies have suggested that at least 90% of correctional facilities have access to health care services (eg, on-site, on-call, telehealth), which may vary by jurisdiction.^34,35 In addition, in Estelle v Gamble, the US Supreme Court held that deliberate indifference by prison personnel to a prisoner’s serious illness or injury constitutes cruel and unusual punishment under the Eighth Amendment. ³⁶ Thus, incarcerated individuals are entitled to receive appropriate medical care, whether in-house or at an outside facility such as a regional or community hospital. A working group of experts convened by the RAND Corporation and the University of Denver acknowledged that many convicted and sentenced people arrive at correctional facilities in poor health, which reflects the reality of inadequate access to care and treatment in the community before incarceration. ³⁷ For our study, it seems plausible that the need for a higher level of care might be more readily recognized in correctional facilities than in the community, which employ not only nurses and clinicians but also many other staff members such as correctional officers who are trained in CPR and first aid. ²³ Therefore, earlier detection of clinical deterioration may lead to more frequent hospital referrals. While this could result in a relative overuse of hospital services for incarcerated individuals, our analysis limited the potential of such effects by restricting the study population to patients hospitalized for ≥24 hours. Not only did this strategy eliminate same-day surgeries, it also reduced the number of less acute admissions that did not warrant longer lengths of stay. Together, our findings suggest that more prompt prehospital intervention for incarcerated adults may have contributed to lower inpatient mortality.

Notably, the hospitalizations in our study occurred during the second year of the COVID-19 pandemic, when inpatient care for COVID-19 was more established and COVID-19 vaccines had just become available. Reported associations between incarceration status and inpatient outcomes during the COVID-19 pandemic have been mixed, likely due to substantial methodological differences across studies.^15-18 For example, a study of a Michigan-based private hospital system found that inpatient mortality for COVID-19 was higher among incarcerated adults than among nonincarcerated adults. ¹⁶ However, the Michigan Department of Corrections operates its own hospital/infirmaries, and, therefore, outside hospital referrals likely represented more acute cases, which was not addressed in the main analysis. ³⁸ Furthermore, according to supplemental data, the 30-day readmission rate was lower for the incarcerated group, suggesting that discharged incarcerated patients had less severity or better access to appropriate care on-site at the Michigan Department of Corrections.

Beyond our main findings, we noted significant patterns by race and ethnicity. In particular, we found that non-Hispanic American Indian/Alaska Native adults had much higher rates of hospitalization and inpatient death than adults from other racial and ethnic groups. The inclusion of federal hospitals in our dataset enabled us to capture those participating in the Indian Health Service safety-net system. The Indian Health Service operates a network of hospitals and clinics across the United States, serving members of the American Indian/Alaska Native community, including in rural areas. Our analysis highlights the health care vulnerability of this indigenous patient group, which has been previously documented. ³⁹ Higher health care use and higher inpatient mortality among non-Hispanic American Indian/Alaska Native adults than among non-Hispanic White adults suggest that expanded efforts are needed to improve early detection and disease management in this group. Importantly, federal funding per enrollee is much lower for the Indian Health Service than for other major safety-net programs such as Medicaid and the Veterans Health Administration. ⁴⁰

Limitations

Our study had several limitations. First, we relied on administrative data to define incarceration status, which may have led to some misclassification. In the future, data collection organizations should collaborate with correctional agencies to more accurately identify incarcerated people. Therefore, when interpreting our results, caution is warranted. Second, we considered only inpatient mortality rather than 30-day mortality, given our available data. However, by limiting our study population to adults aged 18 to 64 years, the potential effects of non-inpatient deaths were lessened. Future studies should examine 30-day mortality and readmission rates. Third, we relied on hospital discharge data rather than unique patients, and, therefore, a single patient may have had multiple discharges. However, data from the National Readmission Databases indicate that approximately 80% of inpatients have only a single hospital visit in a given year. ⁴¹ Furthermore, our large dataset likely mitigated such effects. Fourth, we accounted for only basic demographic factors in our multivariate models, given the available data for population denominators. Future studies should account for other potential confounders, such as prehospital socioeconomic (eg, poverty level, education), behavioral (eg, exposure to illicit drugs, violence), and clinical (eg, multimorbidity, COVID-19 vaccination status) factors.

Strengths

Our study also had several strengths. First, to our knowledge, this is the first population-based study to simultaneously compare hospitalization and inpatient death patterns by incarceration status, which addresses an important gap in the literature. Second, we disaggregated discharges by major hospital service lines, which offers enhanced understanding about the drivers of health care use patterns without exacerbating misattribution bias, which is more likely when relying solely on the primary discharge diagnosis. Third, with 6.3 million hospital discharges from 31 US jurisdictions, this is one of the largest studies to examine the association between incarceration and in-hospital outcomes. Fourth, our data included discharges from both federal and nonfederal acute care hospitals, enabling us to capture patients served by the Veterans Health Administration and Indian Health Service safety-net programs.

Conclusions

This multistate study demonstrated that incorporating both hospitalization and inpatient mortality patterns provided a more nuanced understanding of health care use among incarcerated adults. By examining service-line–specific hospitalizations alongside mortality, we offer new evidence that challenges assumptions about health outcomes in correctional settings and highlights the role of prehospital recognition and access to care. These findings underscore the need for future work that links correctional health services with community data systems to clarify mechanisms driving hospital use and to inform evidence-based policies aimed at improving care continuity for incarcerated and nonincarcerated populations.