Incarceration Rates and Incidence of Sexually Transmitted Infections in US Counties, 2011–2016

Abstract

Objectives. To examine rates of sexually transmitted infections as a function of jail and prison incarceration rates across US counties for the years 2011 to 2016.

Methods. We used data from several national databases. The outcomes were county-level chlamydia and gonorrhea incidence as reported by the Centers for Disease Control and Prevention (2012–2016). The exposures were lagged specifications of county-level jail and prison incarceration rates as reported by the Vera Institute of Justice (2011–2015). We estimated mixed models to account for the 3 sources of response variable variation occurring across repeated measures collected from counties nested within states.

Results. In the final model, jail and prison incarceration rates were associated with a rate increase of 10.13 per 100 000 and 8.22 per 100 000, respectively, of chlamydia incidence. The corresponding rate increases for gonorrhea incidence were 2.47 per 100 000 and 4.40 per 100 000.

Conclusions. These findings provide some evidence that the documented differences in chlamydia and gonorrhea incidence between counties may be partially attributable to differences in jail and prison incarceration rates.

Structural racism is a key determinant of population health in the United States,¹ causing widespread suffering not only for people of color but also for society as a whole.² One of the major institutional mechanisms that reproduces racial inequality is mass incarceration.³ According to a recent estimate,⁴ in 2012, a Black man born between 1965 and 1974 without a high-school diploma had a 80.9% chance of ever being incarcerated. Thomas⁵ argues that historic and contemporary social forces of racial subordination—from slavery to contemporary mass incarceration⁶—has led to sexual and care-seeking behaviors that favor the transmission of sexually transmitted infections (STIs) and, ultimately, the disproportionate documented rates of STIs across the US South where these systems of subordination have been concentrated. Thus, it is imperative to examine the contextual factors—poverty, discrimination, Black sex ratios, incarceration rates, and racial segregation—that promote patterns of sexual networks that facilitate transmission of STIs.^7,8 This article focuses on the global effects of mass incarceration on incidence of STIs by asking whether the rate of persons admitted to county jails and state prisons is associated with incidence of chlamydia and gonorrhea across US counties net of poverty, racial segregation, and Black sex ratios.

CORRELATION FROM INDIVIDUAL-LEVEL STUDIES

The vast majority of people who experience incarceration will return to their community, and it is likely that the postrelease period is a particularly high-risk time for returning citizens.⁹ In addition to the STI risks experienced before incarceration,^10,11 people returning to their communities are likely to engage in behaviors that may elevate their chances of acquiring an STI.^12–19 A retrospective cohort study of individuals released from jail documented significantly higher rates of STI for the cohort than in the general population in the 1 year following release.²⁰ Another study found that incarceration of less than 1 year (e.g., county jail) and more than 1 year (e.g., state prison) were both associated with STIs.²¹

In addition to the potential negative impact of incarceration on the person experiencing incarceration, the “forced migration” associated with incarceration has socially destabilizing effects on families, attributable to the removal of people from households,^22,23 and on communities by undermining social cohesion and control.²⁴ In this way, there may be “spillover” effects associated with mass incarceration for those who are not directly impacted, such as children and romantic and sexual partners.^7,25,26 For example, among children, experiencing parental incarceration at an early age is associated with sexual risk during adolescence and young adulthood²⁷ and later STI.²⁸ Race and class inequalities in parental incarceration may contribute to family complexity (e.g., noncustodial parenthood) and the reproduction of childhood disadvantage.²⁹ A higher ratio of women to men may lead to men having more power in opposite-sex sexual relationships, and lower levels of sexual relationship power among women is associated with engaging in riskier sex and having sex with higher-risk partners.³⁰ The gender‐ratio imbalance among African Americans attributable, in part, to mass incarceration is a contextual factor leading to increases in behaviors that place Black women at risk.³¹ Men recently released from prison perceive that they have power to negotiate sex with women,³² and some men who experience serial incarceration perceive relationships with their female partners and children to be complex and difficult to navigate.³³

ECOLOGICAL STUDIES

A handful of ecological studies have explicitly examined incarceration and documented infection rates. A study of 100 counties in the US state of North Carolina found that rates of STI (i.e., chlamydia, gonorrhea, syphilis, and HIV) and adolescent pregnancies consistently increased with increasing incarceration rates.³⁴ A longitudinal study of neighborhoods in Atlanta, Georgia, found that census tracts with increasing male incarceration rates experienced a more rapid increase in their rate of newly diagnosed STIs.³⁵ However, census tracts with medium and high baseline male incarceration rates actually experienced a decrease in newly diagnosed STIs over time. Finally, a study of census tracts in San Francisco, California, found a positive association between incarceration rates and chlamydia incidence in women aged younger than 25 years.³⁶ Although not focused on STI, a compelling study by Schnittker et al.³⁷ examined the spillover effects of state‐level incarceration rates on the functioning and quality of the health care system. Using state-level panel data, they found that for each percentage‐point increase in the formerly incarcerated population, there was a 0.32-percentage‐point increase in the uninsured population, and there were around 28 more emergency department visits per 1000 residents.

Taken together, research suggests that geographic areas with a high percentage of incarcerated persons have higher rates of STI not only because persons who experience incarceration are at an elevated risk for acquiring an STI but also because there is something about the geographic spaces where mass incarceration is concentrated (e.g., the disruption of sexual relationships, the change in sexual behaviors), which places everyone in the area at a higher risk whether they have personally experienced incarceration or not.^3,22,23 A handful of ecological studies document how living in an area characterized by high levels of incarceration leads to increases in STI in diverse, yet relatively small, geographic spaces (i.e., across counties in North Carolina and across census tracts in Atlanta and San Francisco).^34–36

We examined rates of STI as a function of jail and prison incarceration rates across US counties for the years 2011 to 2016. This was made possible by the recent release of a uniquely detailed historical data set of incarceration in both local jails and state prisons that was developed by the Vera Institute of Justice to explicitly examine incarceration at the county level across the United States (see Methods). The findings from this ecological study provide some evidence that the documented differences in chlamydia and gonorrhea incidence between counties may be partially attributable to differences in jail and prison incarceration rates.

METHODS

Data for this study were from multiple sources and were merged by using county-level Federal Information Processing Standards codes. All variables included single-year estimates for 2011 to 2016, except for urbanicity and segregation, which were time-invariant, and incarceration rates, which were annual estimates for 2011 to 2015. The unadjusted means, medians, and sample sizes are available in Table A (available as a supplement to the online version of this article at http://www.ajph.org) for each year for the primary independent and outcome variables, along with the baseline bivariate relationships using Pearson R (Table B, available as a supplement to the online version of this article at http://www.ajph.org). We obtained county-level surveillance incidence rates for chlamydia and gonorrhea for 2011 to 2016 from the Centers for Disease Control and Prevention’s (CDC’s) AtlasPlus,³⁸ which estimated incidence per 100 000 persons by using population denominators from the US Census Bureau.

County-level incarceration statistics were from the Vera Institute of Justice “In Our Own Backyard” Incarceration Trends (IOB) data set.³⁹ This novel data set was designed to motivate research on incarceration at the local level. Given that county officials—judges, prosecutors, jail administrators—are the primary actors deciding how communities use incarceration, having county-level data allows for more robust studies of the causes and impact of incarceration. Two authors (K. M. N. and M. O.) participated in the inaugural IOB symposium held during 2018. The IOB was collected from several Bureau of Justice Statistics data sources as well as crime data drawn from the Uniform Crime Report and Census estimates. The most recent year for which data were available was 2015. We used annual estimates of jail and prison admissions for county residents to calculate county-level jail and prison incarceration rates per 100 000 persons. Jail admissions are most often measured during a midyear week and then multiplied to get an annual estimate. Prison admissions are a count of the number of times people are sent to prison from each county excluding admissions such as returns from court and transfers from other jurisdictions. For more information, see the codebook (https://bit.ly/2ZH4znm) and the methodology report (https://bit.ly/2rLg0Oi). We also used the urban designations (urban, suburban, small/mid, rural) as provided in the IOB.

We used Black female-to-male sex ratio and Black–White neighborhood segregation as additional indicators of structural racism. We calculated Black female-to-male population ratios and percentage working age population from the US Census Bureau’s file “Annual County Resident Population Estimates by Age, Sex, Race, and Hispanic Origin: April 1, 2010 to July 1, 2017.” We measured county-level residential racial segregation by using Black-to-White isolation index scores as calculated by the Brown University American Communities Project.⁴⁰ The isolation index is a common dimension of racial/ethnic segregation calculated as the minority-weighted average of the minority proportion in a given locale.⁴¹ Index scores ranged from 0 (no integration of Blacks with Whites) to 1 (complete integration of Blacks with Whites). In calculating isolation scores, the American Communities Project defined “neighborhood” as a given census tract plus each adjacent tract to account for spatial effects.

The Robert Wood Johnson Foundation County Health Rankings and Roadmaps database provided the remaining variables. High-school graduation rates represent the estimated proportion of ninth graders that graduate from high school within 4 years. We sourced county-level percentages of children in poverty in the County Health Rankings and Roadmaps from US Census Bureau’s Small Area Income and Poverty Estimates. Percentage of working-age population included persons aged 18 to 65 years.

We estimated random effects models for repeated measures data⁴² with Stata version 15 (StataCorp LP, College Station, TX) to examine the differences across counties that may influence STI incidence. Given the short time period for the study, we did not anticipate unexpected variation or special events that may affect the outcome variable, but we did include a control for time. Given the documented regional variation in STI, we also included a control for US Census region. We adjusted standard errors for state-level clustering. A Breusch and Pagan Lagrangian multiplier test for random effects showed that the data need to be modeled as panel data (χ² = 6575.32; P < .001).

We estimated ordinary least squares models to assess for multicollinearity, including combinations of total, jail, and prison incarceration rates, and their interaction with time. The interaction for incarceration and time caused elevated variance inflation factors (> 6.0), so we centered time for use in interactions. After centering time, the mean variance inflation factor was 1.54 with the highest variance inflation factors for urban designation (2.28 and 2.79). We estimated multiple models by using 1- and 2-year time lags for incarceration including total jail or prison incarceration, jail-only incarceration, and prison-only incarceration. The 1-year lagged variable for incarceration means that the outcome is regressed on the incarceration rate for the previous year. We included time lags following Thomas and Torrone’s³⁶ ecological examination of whether there is a lag in time between incarceration and a number of community health outcomes among counties in North Carolina. They found the strongest and most consistent correlations for a 1-year lag.

We then estimated mixed models to directly model the random effects for counties and states. We used a model-building approach involving estimating models in a sequential stepwise fashion to properly account for the 3 sources of response variable variation occurring across repeated measures collected from counties that were nested within states.⁴³ We mean centered all continuous control variables in the fixed portion of the models. We used the Akaike information criterion, Bayesian information criterion, and log likelihood ratio to determine the best model fit. The estimation sample for chlamydia with a 1-year lag for incarceration rates included 10 003 total observations across 5 years nested within 2439 counties, nested within 44 states (81.2% of US counties). For gonorrhea, the estimation sample was 9995 over 5 time years nested within 2437 counties nested within 44 states. Counties were missing if they did not have reliable estimates for the study variables. We used 6 years of data to estimate the 5 time points (incarceration rates were from 2011 to 2015; STI and other variables were from 2012 to 2016). The estimation sample for a 2-year lag was smaller because it only included 4 years of data.

RESULTS

Across counties, a 1-standard-deviation increase in total incarceration was associated with an increase of 12.19 new cases of chlamydia per 100 000 persons net of relevant county-level demographic factors and other indicators of structural racism (Table 1). This represented a 4.89% increase from the model average of 249.22 per 100 000 chlamydia incidence. When we lagged incarceration rates by 2 years—incarceration rates predicted the rate increase in chlamydia incidence 2 years later—the estimate was 8.81 per 100 000. A different specification of total incarceration rates found that a 10% increase from the mean incarceration rate was associated with a 3.94 per 100 000 increase in chlamydia incidence the following year, with 95% of counties experiencing a 10% increase in total incarceration showing between 1.95 and 5.93 per 100 000 increase in chlamydia incidence. The corresponding rate for a 20% increase in total incarceration was 6.96 (95% confidence interval [CI] = 3.45, 10.47) per 100 000. Table 1 also shows incarceration rates disaggregated by jail and prison. The rate increases of chlamydia incidence for a 1-standard-deviation increase in jail incarceration were 10.98 using a 1-year lag and 7.96 using a 2-year lag. For prison incarceration rates, the corresponding rate increases were 8.79 per 100 000 and 8.19 per 100 000.

TABLE 1—

Rate of Sexually Transmitted Infections per 100 000 by County as a Function of County-Level Incarceration Rates: United States: 2011–2016

	Model 1: 1-Year Laga			Model 2: 2-Year Laga
	County-Years	Interclass Correlation	Coefficient (95% CI)	County-Years	Interclass Correlation	Coefficient (95% CI)
Chlamydia
Total incarceration	10 003	0.76	12.19 (7.40, 16.97)	8 119	0.77	8.81 (2.49, 15.13)
Jail incarcerationb	10 003	0.75	10.98 (6.30, 15.65)	8 119	0.77	7.69 (1.37, 14.01)
Prison incarcerationb			8.79 (4.09, 13.49)			8.19 (2.62, 13.76)
Gonorrhea
Total incarceration	9 995	0.64	3.03 (1.59, 4.47)	8 110	0.63	3.31 (1.20, 5.42)
Jail incarcerationb	9 995	0.63	2.22 (0.71, 3.72)	8 110	0.66	2.49 (0.33, 4.65)
Prison incarcerationb			5.52 (2.68, 8.35)			5.58 (2.57, 8.59)

Note. CI = confidence interval. Coefficients are the rate differences in sexually transmitted infection observed when the incarceration rate is increased by 1 standard deviation, according to county of residence. Coefficients were adjusted for county distributions of age and race, percentage of children living in poverty, percentage of residents with a high-school degree, urbanicity, and data year and region. The standard errors used to calculate 95% CIs were adjusted for clustering within states.

^aRandom effects models with lagged specifications for standardized incarceration rates.

^bCoefficients for jail and prison incarceration were estimated in the same model.

The random effects models for gonorrhea indicated that a 1-standard-deviation increase in total incarceration was associated with a 3.03 per 100 000 rate increase in gonorrhea incidence the following year representing a 9.36% increase from the model average of 32.39 (Table 1). The 2-year lagged total incarceration rate was associated with a 3.31 per 100 000 rate increase. A 10% increase in total incarceration was associated with a 1.18 per 100 000 (95% CI = 0.52, 1.83) rate increase from the mean gonorrhea incidence the following year, with a 20% increase corresponding to a 2.15 per 100 000 (95% CI = 0.94, 3.36) rate increase. When we disaggregated total incarceration into jail and prison incarceration, we found that prison incarceration had a stronger association with gonorrhea incidence.

The best-fitting mixed models are presented in Table 2. The models used for model building are available in the Tables C and D, available as supplements to the online version of this article at http://www.ajph.org). The fixed effects in model 1 indicated that both jail (10.11/100 000) and prison (7.06/100 000) incarceration were associated with higher chlamydia incidence, on average, across US counties, and that chlamydia incidence, overall, has increased over time (11.21/100 000). The rate increase of chlamydia incidence per a standard deviation increase in jail incarceration reflected a 4.27% increase in chlamydia incidence from the model average for counties. The corresponding rate increase for prisons was 2.69%. The fixed effects in model 2 indicated that counties with higher jail and prison incarceration rates have higher incidence of gonorrhea (2.47/100 000 and 4.40/100 000, respectively) and that, on average, gonorrhea incidence has increased by 6.49 per 100 000 each year from 2012 to 2016. The rate increases represented a 4.39% and 7.81% increase over the model average gonorrhea incidence for jail and prison incarceration, respectively.

TABLE 2—

Mixed Models for Sexually Transmitted Infection Incidence per 100 000: Repeated Measures Nested Within Counties (Level 2), Nested Within States (Level 3): United States, Years 2011–2016

	Model 1: Chlamydia	Model 2: Gonorrhea
Fixed effects, coefficient (95% CI)
Intercept	262.16 (230.65, 293.66)	40.30 (33.46, 47.13)
Year	11.21 (7.56, 14.86)	6.42 (4.91, 7.92)
Jail incarceration ratea	10.11 (7.28, 12.93)	2.54 (1.51, 3.58)
Prison incarceration ratea	7.06 (3.82, 10.30)	4.42 (3.23, 5.61)
Black female–male sex ratiob	15.61 (9.07, 22.14)	4.66 (2.54, 6.79)
% working age populationb	11.16 (9.74, 12.59)	1.51 (1.08, 1.94)
Black–White segregationb	8.36 (7.84, 8.88)	3.11 (2.96, 3.26)
% children living in povertyb	1.48 (0.94, 2.03)	0.37 (0.18, 0.55)
% high-school graduatesb	−0.35 (−0.66, −0.03)	−0.14 (−0.26, −0.02)
Urban designation (Ref: urban/suburban), coefficient (95% CI)
Small or midsize designation	58.12 (43.27, 72.97)	16.27 (11.88, 20.65)
Rural designation	39.75 (25.28, 54.23)	3.46 (−0.86, 7.78)
US census region (Ref: Midwest), coefficient (95% CI)
Northeast	−48.34 (100.71, 4.04)	−10.40 (−21.24, 0.44)
South	−22.17 (−61.24, 16.89)	−8.92 (−16.72, −1.12)
West	18.13 (−37.07, 63.34)	−12.48 (−22.00, −2.96)
Random effects, coefficient (95% CI)
Level 1: intercept variance	3 164.89 (3 044.74, 3 289.79)	546.08 (525.42, 576.55)
Level 2: intercept variance	11 199.05 (10 266.38, 12 216.45)	782.48 (691.15, 885.88)
Level 2: time variance	298.80 (245.17, 351.17)	48.48 (41.07, 57.23)
Level 2: covariance	−667.05 (−843.09, −491.01)	−60.87 (−84.42, −37.31)
Level 3: intercept variance	2278.38 (1359.19, 3819.18)	71.34 (40.11, 126.92)
Level 3: time variance	121.19 (72.37, 202.95)	21.42 (13.10, 35.02)
Model summary
Level 2: intraclass correlation	0.81	0.61
Level 3: intraclass correlation	0.14	0.05
Log likelihood ratioc	−58 311.0	−48 624.1
Degrees of freedom	20	20
AIC	116 662.1	97 288.1
BIC	116 806.3	97 432.3

Note. AIC = Akaike information criterion; BIC = Bayesian information criterion; CI = confidence interval. The number of county-years for the estimation sample is 10 003 with 2439 counties nested in 44 states. The number of counties per state ranges from 2 to 846 with a mean of 227.3. The number of years per county ranges from 1 to 5 with a mean of 4.1.

^aIncarceration rate is lagged by 1 year and standardized.

^bVariable is mean-centered.

^cLog likelihood ratio tests of successive models are significant at the P < .001 level.

Overall, the fixed-effects estimates from the mixed models were similar to the estimates found in the random-effects models. However, the random effects in the mixed models provided additional insights. First, there was more variation in STI incidence between counties than within counties over time. Second, there was substantial heterogeneity across counties and states that was unexplained by the model. The final model specification for chlamydia decreased the county-level variation in STI by 54.78% across counties and 68.07% across states. The corresponding numbers for gonorrhea were 72.43% for county-level variation and 93.85% for state-level variation. The model covariates accounted for a lot of the variation, but not all.

DISCUSSION

We used a mixed-model approach to examine the relationship between STI incidence and total incarceration rates—and incarceration disaggregated into county jail and state prison incarceration—at the county level from 2011 to 2016. The findings showed that counties with higher rates of incarceration have higher incidence of both chlamydia and gonorrhea and that both jail and prison incarceration were independently associated with rate increases in STI. We accounted for time-order by using lagged specifications for incarceration, strengthening the argument for a potential causal relationship. Disaggregation of incarceration data showed that chlamydia rates were most strongly associated with jail incarceration rates, and gonorrhea rates were most strongly associated with prison incarceration rates. Overall, these findings provide some evidence that the effects of mass incarceration are manifest at the population level across US counties and suggest possible “spillover” effects associated with mass incarceration for those who are not directly impacted. There was, however, substantial variation in the association between incarceration and STI incidence over time and across counties and states, net of relevant demographic factors (e.g., poverty) and other measures of structural racism (i.e., Black female-to-male sex ratio, Black–White segregation).

Strengths and Limitations

The strength of this study was the inclusion of counties from across the United States. However, there were several data limitations. First, chlamydia and gonorrhea estimates reflect the minimum likely incidence rates, because not all infections are diagnosed and reported to national surveillance systems. Data from the CDC indicate state-level variation in chlamydia screening rates, with southern states having the lowest rate of screening.⁴⁴ This suggests regional variation in screening practices. Although estimates for screening are not available at the county level, the state-level adjustments included in the models helped to account for this variation. In fact, including a random effect for states in the mixed models greatly reduced the variation in STI among counties (Tables C and D).

Related, reporting and surveillance practices differ between and within states, so national reporting data are incomplete.³⁸ Although all diagnoses are impossible to ascertain, it is estimated that 40.3% and 32.2% of chlamydia and gonorrhea cases are undiagnosed, respectively.⁴⁵ As chlamydia is often asymptomatic, trends in the number of diagnosed cases may be influenced by trends in incidence of infections or in diagnostic, screening, and reporting practices. Therefore, our findings are likely an underestimation. The ecological fallacy would be to infer from our findings that incarceration leads to higher STI incidence. Given the absence of individual-level data, it is impossible to conclude that incarceration causes infection with sexually transmitted diseases.

There are also limitations associated with the other county-level variables. For example, the estimation of high-school graduates may vary across states because of suppression requirements and classification and inclusion rules for marked groups.⁴⁶ Finally, the incarceration and STI data are not sex- and race/ethnicity–specific. Future research should model the impact of incarceration on disparities directly or for the health of Black people, specifically, across different geographic spaces.

Public Health Implications

Our study supports the growing body of research examining the complex relationship between mass incarceration and STI. Previous studies have found, for example, that, for women, dating a man who has been incarcerated is positively associated with ever having an STI, infection with an STI is associated with a personal history of incarceration, incarceration disrupts partnerships and ends romantic and sexual relationships, and previously incarcerated people have increased relative risk for chlamydia and gonorrhea.^12,20,21 Our findings reinforce the need to move beyond the individual to examine how mass incarceration creates conditions that favor disease transmission, similar to a handful of other studies. These studies document how living in an area characterized by high levels of incarceration leads to a more rapid increase in new STI diagnoses in diverse, yet relatively small, geographic spaces (e.g., a single state or city).^34–38

The enormous scale of imprisonment for Black men is undeniable, and recent scholarship identifies contemporary criminalization and incarceration as an important representation of structural racism.^6,47,48 Community harms resulting from incarceration create a social dynamic that increases the negative consequences to those who are removed from and later returned to specific neighborhoods in concentrated numbers.³ Concentrated imprisonment further damages the social bonds that sustain life, especially for poor communities, as incarceration removes the benefits that individuals bring to their familial networks such as providing money, childcare, and emotional support and benefits to their larger social networks including social capital.⁴⁹ That is, imprisonment was once a concern that affected the individual, but the scale of incarceration today is such that the effects are felt much more broadly and can be detected globally at the population level, as suggested by the findings in the present study. Simply put, without personally experiencing incarceration or having personally engaged in high-risk sex or other behaviors, people may still suffer the consequences from residing in, and forming relationships within, a community structure that fosters unequal opportunity for healthy behaviors because of mass incarceration.

We argue that a multilevel and multisector approach is needed to ameliorate the deleterious health effects of mass incarceration. For example, implementation research to develop an integrated health care model between the state- and county-level justice and public health systems would help communities to frame criminal justice policy as health policy. This collaboration could also support screening and treatment during the reintegration period, which may be crucial for decreasing new STI cases. Finally, findings from this ecological approach have important implications for directing limited national health resources to areas with high incarceration rates to geographically focus prevention interventions and provide improved access to STI services in these areas both within and outside correctional spaces.