What Can We Infer About Incarceration and Sexually Transmitted Diseases?

A handful of recent ecologic and cross-sectional studies have found evidence of a positive association between incarceration rates and sexually transmitted disease (STD) rates.1–6 In this issue of STDs, Dauria et al.7 publish the first ecologic study of the association between incarceration and STD rates over time. As expected, Dauria et al.7 found evidence supporting a positive association between higher baseline incarceration rates and higher STD incidence rates. Intuitively, this makes sense, and other studies have observed a similar association. One simple explanation of the association is that high incarceration rates indicate neighborhoods with high crime and illicit activity, such as drugs and commercial sex, all of which favor STD transmission. Another more complex theory is that when a large number of men in a community are removed, for example, because of incarceration, the ratio of men to women becomes unbalanced. This imbalance destabilizes the community and shifts power dynamics, giving the remaining men in the community more power, and women less power, to negotiate sexual partnerships, relationship dynamics, and sexual behaviors, creating to a high-risk sexual environment.1–5,8,9 There is also evidence that correctional populations have higher rates of infectious diseases than do the general population, including HIV and other STDs.10 It is unclear when infections are acquired, but presumably, many men are infected before they are incarcerated, and many become infected while incarcerated. Incarcerated men who become infected while in the system may spread infection when they return to their home community at the end of their sentence, or while on probation or parole.2,9,11 By extension, the in-community sexual partners of incarcerated men may have other sexual partners during the incarceration period, complicating the sexual network and increasing risk of STD acquisition and transmission.12,13 In addition, men with a history of incarceration and unstable housing situations have more sex partners and more unprotected sex,6,9,14 increasing their risk for STD acquisition and transmission. Unexpectedly, Dauria et al.7 also observed decreases in STD rates with increases in incarceration rates over time. Although they cannot explain the mechanism(s) underlying this trend given their data, they hypothesized that removing men from the community who are infected with an STD before they are incarcerated may reduce STD prevalence. Sexually transmitted disease rates could remain lower in the longer term if these men receive STD screening and treatment while incarcerated,10 suggesting that these results could have meaningful policy and practice implications. Before getting too excited and making grand interpretations and recommendations, we need to examine the results in more depth for their correctness. Incarceration and STD rates were calculated using administrative data, which have the advantage of being collected from entire populations and thus theoretically have zero sampling variance. Administrative data also have the advantage of being routinely collected over time, allowing for trend analysis. Administrative data definitions can change, especially over time, as seen with definitions of HIV/AIDS and incarceration (prison or prison and jail sentences), and data quality and completeness can also vary, as seen with STD risk factor data collected on reportable disease reports.15 Dauria et al. limited STDs to include chlamydial infection, gonorrhea, and syphilis, the definitions, detection, and reportability of which remained consistent over the study period. They also limited their definition of incarceration to prison sentences. Dauria et al. compare longitudinal incarceration and STD data by census tract, and this design creates conditions for variables in the analysis to be both temporally and spatially correlated (i.e., between variables) and auto-correlated (i.e., within variable); in other words, they are spatially and temporally dependent. Sexually transmitted diseases are inherently temporally auto-correlated because current STD rates are dependent on past STD rates, and future STD rates are dependent on current STD rates.16 Similarly, STD rates are spatially auto-correlated because STD rates near each other in space are more similar than STD rates further apart.16–18 There is some evidence suggesting that incarceration rates are also spatially auto-correlated,2 and it is reasonable to assume that incarceration rates may also be temporally auto-correlated. Dauria et al. did account for temporal clustering, that is, the dependence of measuring the same individual say 5 times over the life of the study with multilevel modeling; however, adjusting for temporal clustering is not the same as adjusting for temporal correlation or temporal autocorrelation, that is, the temporal sequence of events happening a day apart, a month apart, or a year apart. Random-effects models assume independence between cluster-specific baseline risk and exposure and is not going to alleviate the problem of temporal dependence without a proper temporal correlation structure. Dauria et al. accounted for the spatial correlation by controlling for covariates in their analysis, which assumes that the spatial correlation of the outcome will be fully explained by the covariates. However, Dauria et al. did not check for residual spatial autocorrelation to validate their assumption, which is a crucial part of the model. To date, only one study of incarceration rates and STDs has accounted for spatial dependence due to both correlation and autocorrelation2 and found the spatial dependence to be small, suggesting that the result of Dauria et al. may only be slightly biased if spatial autocorrelation does exist. Ignoring spatial (probably minimal) and temporal (probably more substantial) dependence is more likely to affect confidence intervals than point estimates. Point estimates are likely to be affected by control for confounding variables. Dauria et al.7 controlled for a reasonable set of social determinants of health using data from the census. To date, most (if not all) sociospatial ecologic health studies have been highly dependent on census data to control for sociodemographic confounding. Census data are appealing because they are readily available and aggregated to geopolitical administrative units, the underlying population is known facilitating rate estimation, and they represent the population, making the sampling variance theoretically zero. The census is very good at capturing data on the general population, much better than any research study could accomplish; however, there are “hard-to-count” groups and circumstances that lead to undercounting in any census. The US Census Bureau estimated that 3% of the US population was undercounted in the 2010 census (http://www.census.gov/newsroom/releases/archives/2010_census/cb12-95.html accessed March 30, 2015). Undercounted groups included renters, blacks, Hispanics, American Indian/Alaska Natives living on-reserve, and men. All of these undercounted groups share characteristics with groups at high risk for incarceration and STDs, suggesting that those individuals included in incarceration and STD rates may not be included in census derived social determinants of health measures. Dauria et al. have given a solid effort to analyzing a very important and challenging-to-measure relationship. They used many methods to reduce bias and error; however still, the results must be interpreted with caution given insufficient control of measured confounders, unmeasured confounders, and concurrent interventions that will all affect the observed associations both in direction and in magnitude. Although the data for this study were collected over time, the ecologic design means we cannot know for certain which came first, high incarceration rates or high STD rates. Ecologic fallacy teaches us that we cannot downscale the association to infer how incarceration and STDs are related at the individual level, and the lack of individual level data prevents examination of the cross-level associations between incarceration and STDs. So, what do we know? We know that there is increasing evidence suggesting a complex relationship between incarceration and STDs. Regardless of the direction, magnitude, or underlying mechanisms of that relationship, prisons and jails stand to be important STD intervention points.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.