HIV and Hepatitis B and C incidence rates in US correctional populations and high risk groups: a systematic review and meta-analysis

In 2008, nearly 2.4 million people were incarcerated in United States (US) jails or prisons [1, 2]. Furthermore, about 25% of persons living with the Human Immunodeficiency virus (HIV) and about 30% of those living with Hepatitis C Virus (HCV) infection spent time in correctional facilities [3]. As a result, many have expressed concerns that transmission of blood-borne infections among inmates may be a major source for further spread to the general population [4‐7].

Three lines of evidence support this view: (1) the prevalence of HIV infection, viral hepatitis, and sexually-transmitted infections (STI) is typically higher in incarcerated than in non-incarcerated populations; (2) illicit drug injection, unprotected sexual activity, and other risky behaviors are common in prisons and jails; and (3) correlative studies have repeatedly found independent associations between antecedents of incarceration and increased risk for infections such as HIV.

In more details, despite declines since the 1990s, it has been estimated in 2008 that 1.5% of the total US custody population in federal and state prisons was infected with HIV [8]; that 12.0% to 35.0% had chronic HCV infection [3]; and that 1.0% to 3.7% had serological markers of chronic HBV infection [3]. Based on these estimates, the prevalence of HIV was about four times higher among prison inmates than in the general population [9]. Likewise, the prevalence of chronic HBV infection was 2 to 6 times higher, and that of chronic HCV infection was up to 10 times higher, among prison inmates than in the community.

In contrast, more uncertainty remains about the extent of high-risk behavior taking place within US correctional facilities. Due to important differences between institutions in enabling factors such as overcrowding and understaffing, estimates of interest vary widely. Recent studies suggest that 3.0% to 28.0% of adult inmates use intravenous drugs while incarcerated; 4.0% to 65.0% engage in unprotected homosexual activities [3, 10, 11]; and 0.0% to 15.7% report sexual victimization during incarceration [12]. Prison entrants, incarcerated inmates, and intravenous drug users (IVDU) also tend to share a number of incarceration-related factors that predict HIV infection, including overall length of time spent incarcerated [13], repeated incarceration [14], tattooing in prison [15‐17], and history of syringe sharing in prison [18‐20]. Comparable observations have been made for other blood-borne infections and STI [14, 21‐26].

As just discussed, high HIV prevalence and high risk behavior within correctional systems are well documented. There is much less evidence, however, to support the notion that the correctional setting increases the incidence of HIV, and thus plays a central role in sustaining or increasing community rates when inmates are released. Many investigators, in fact, have suggested that inmates appear more likely to acquire infection outside than inside correctional facilities [3, 10, 27]. Since determining the role of incarceration in the epidemiology of HIV transmission is a crucial step toward formulating cost-effective public health policies and interventions for US HIV/AIDS control, we conducted a systematic and comparative literature review of HIV, HBV, and HCV incidence among residents of correctional facilities, released detainees, community-living IVDU and community-living men who have sex with men (MSM) in the US. Our aims were twofold: to summarize the published literature on HIV incidence rates and other key blood-borne infections, such as HBV and HCV, in US correctional facilities; and to compare these with infection rates among non-incarcerated individuals who bear high burdens of blood-borne infections and STIs. We hypothesized that inmates experience a lower incidence of HIV than community living risk groups that practice the behaviors which place incarcerated populations at increased risk. We further postulated that comparisons between viruses that share the same routes of transmission in prisons, jails, and high risk groups in the community will provide the comparative basis to more clearly elucidate the role the correctional setting may play in increased risk of infection and the likely avenues for further spread to the community.

Fueled by reports of HIV and STI outbreaks in correctional facilities in the US [11, 50, 97‐101] and in other high-income countries (Scotland [102, 103], Australia [17]), the debate about the magnitude of inmate-to-inmate transmission of HIV in the US has spanned more than two decades. The pattern of results that emerged from our comprehensive review and meta-analysis of HIV, HCV, and HBV incidence studies support the notion that the transmission of HIV and other blood-borne infections in US correctional populations occurs at alarmingly high rates during the periods that recidivists spend outside prison. In our study, HIV incidence among inmates released and reincarcerated was much greater (2.92/100py) than in the US general population (0.02 per 100 population in 2006) [104], while HIV incidence among inmates continuously incarcerated was more similar to the general populace (0.08/100py). Rates of HIV seroconversion among reincarcerated inmates were comparable to those typically observed among non-incarcerated individuals who engage in high-risk injecting and sexual behaviors (street-recruited IVDU, 2.78/100py; MSM, 2.12/100py). HIV incidence among IVDU enrolled in a drug treatment program (1.14/100py) laid between the lower bound observed in the general population and the higher bound observed in the high-risk groups. In contrast, intraprison incidences of HCV (0.75/100py) and HBV (1.71/100py) infection were also higher than in the US general population (0.01 and 0.02 per 100 population, respectively, in 2006) [105], but several-fold lower than among non-incarcerated IVDU (HCV, 18.53/100py; HBV, 16.54/100py) and MSM (HCV, no data available; HBV, 15.9%) [51]. The greater incidence of these infections compared with HIV probably reflects the higher prevalence rates of HCV and HBV infection among prison entrants [106] and higher infectivity of HCV and HBV compared with HIV [107, 108].

Despite significant heterogeneity among included studies, our results were consistent across meta-analyses and multiple meta-regression analyses. Sensitivity analyses indicated that results were only sensitive to the exclusion of four studies, two in which incidence was inferred from self reported data [55, 56], and two in which MSM with early syphilis infection were tested retrospectively [95, 96]. Furthermore, our summary estimates of HIV incidence among inmates were comparable with estimates that were published before 1990 [109‐112]; did not report data in the desired format [48]; and with studies conducted in Europe [113‐115].

Low incidence rates of HIV transmission in prison (range, 0.0 to 0.4/100py) have been reported in three US studies published in the late 1980s [109‐111]. The extent to which these older studies further our understanding of the current dynamic of HIV transmission in prison is unclear, since HIV prevalence at prison intake was considerably lower in the 1980s compared with the 2000s. Of note however, study start year and publication year did not significantly contribute to the meta-regression analysis model, suggesting that calendar time was not an important source of heterogeneity among HIV studies. Also, in a US-based study that did not provide annualized incidence density estimates of HIV infection, but retrospectively followed 5,265 male inmates from their entry into custody in 1978 until 2000, 0.63% of the detainees were diagnosed with HIV infection during incarceration and 4.6% after release from prison [48]. Bias may have inflated the difference in incidence during the incarceration and released periods in this study, but no other source of information was found that estimated HIV incidence among inmates released from a US prison as compared to inmates undergoing periods of continuous incarceration.

In three European studies, HIV incidence among detainees ranged between 0 and 1.0/100py [113‐115], and was highest among male and female IVDUs recruited in 1987-1988 at a prison remand centre in Sweden (the authors of these studies did not clearly indicate what percentage of inmates had been continuously incarcerated) [115]. In a small Australian study (n = 90), including inmates of both genders, a higher incidence of HCV seroconversion was found among inmates who underwent a period of release before reincarceration compared with inmates who had been continuously incarcerated (10.8 vs. 4.5/100py; P = 0.07) [116]. Although the results from these developed nations and from the US seem to be consistent in their documentation of low HIV incidence rates within the prison system and higher rates during post release, international comparisons should be made with great caution given the differences in correctional systems and epidemiological contexts across countries [117]. Hence, two studies identified from the developing world reported higher HIV incidence rates - Brazil (2.8/100py) [118] and Thailand (4.18/100py) [16]. It is clear that an important gap still exists in our understanding of HIV and blood-borne infection transmission in correctional environments throughout the world.

Among MSM, our pooled estimate of HIV incidence is similar to that calculated in a recent study which used a fixed-effects model to calculate a weighted average (2.39%) [119]. In another study, estimates calculated for MSM (0.7/100py) and IVDU (1.5/100/py) differed in comparison to our pooled estimates [120], but this citation [120] did not use meta-analysis methods to estimate HIV incidence in these risk groups.

Our findings are consistent with studies of risky behaviors in correctional populations and hypotheses proposed to explain the apparent paradox of low incidences of HIV, HBV and HCV in the prison system, and high incidence of HIV, HBV and HCV during the post-release periods. As mentioned earlier, the US correctional system offers conditions seemingly favorable to the transmission of blood-borne viruses. There is a large reservoir of potential transmitters in the prison system at any time, and many inmates engage in sexual and drug-mediated risk behaviors regardless of the general lack of condoms [49, 50] and sterile injection material [10, 48]. For instance, although the frequency of drug use in prison is typically lower than in the general community [48, 113, 121], there is clear evidence that incarcerated drug users often continue to inject; that injecting equipment is frequently shared among inmates; and that the risk of equipment contamination by parenterally-transmitted viruses is higher within the prison system than outside of it [20, 113].

Thus, a possible explanation for the low transmission of HIV, HCV, and HBV within the prison system is that inmates' risk networks are on average considerably smaller and more closed within correctional facilities than in the community. Given the de facto segregation of detainees by age, sex, race, category of offense and, historically in some states, by HIV status [10], it is plausible that the lack of bridges between intra-prison networks, and the small size of the networks, lead to the rapid saturation of the susceptible inmates who have effective contacts with a transmitter [106]. However, the formation of bridges between inmates' risk networks when an adequate proportion of susceptible inmates exists may lead to efficient infection transmission [52, 99]. In contrast, studies have shown that many inmates, following re-entry in the community, revert to pre-incarceration habits and engage in high rates of unsafe sexual and intravenous drug use behaviors [122‐126], as suggested, for instance, by high frequency of anal sex reporting [124], excess occurrence of drug overdose [125‐127], and high risk for mortality [80, 128‐130] at post release. Among the four studies on inmates released and reincarcerated, those reporting the highest post-release incidence rates followed recidivist female IVDU [34] and recidivist MSM [36]. As such, sub-populations of inmates with risky pre-incarceration behaviors may be at particularly high risk during periods between release and reincarceration.

There was a notable difference in the reported proportion of IVDU and MSM in recidivist inmate populations [34, 36, 53, 62] as compared to continuous inmates [34, 35, 60, 83]. Three of the four recidivist studies reported the proportion of IVDU (59.1% and 100%) or MSM (0.0% and 100%); while two continuous inmate studies reported the proportion of IVDU (2.9% and 11%) or MSM (0.0% and 4.5%) (Additional file 2). These data suggest that the proportion of IVDU and MSM might be larger in recidivist studies than in continuous inmate studies. If real, such a difference would be another explanation for the observed differences in incidence among the studies in continuous inmates, recidivists and community-living populations.

A meta-analysis of pooled incidence rates stratified by risk behaviors would have further clarified whether the recidivist groups are at increased risk of infection compared to their continuous inmate or community living counterparts. However, available data were not sufficient for this type of analysis. Only one study reported the incidence of HBV (8.2/100py) and HCV (5.5/100py) among continuously incarcerated inmates who reported injection drug use [60]. For both infections, the incidence in this group of inmates was greater than the pooled estimate for CIIP cohorts, but was lower than the pooled estimate for community IVDU; the study did not report whether injection drug use occurred during or prior to incarceration.

Our study is subject to limitations. Despite the crucial importance of characterizing the relations between incarceration and the HIV epidemic in the US, we found only five incidence studies published between 1990 and 2004, and one published between 2005 and 2009, that reported primary data on the transmission of HIV in US correctional populations. The recent literature on incidence of blood-borne infections among MSM was also sparse. The importance of further studies on these outcomes in these populations cannot be over emphasized. Our observation of high infection rates among inmates who were released and reincarcerated was based on four studies only and incidence of HIV was relatively low in two of these studies (Figure 2). Three studies evaluated predominantly IVDU women only; one evaluated MSM only; none measured HBV and HCV incidence; and none included inmates who were not reincarcerated after their release.

Egger's test suggested that publication bias might have affected our results. Possible sources of publication bias include citation bias, poor methodological quality of smaller studies, and true heterogeneity. Although study selection criteria were clearly defined and study selection was done by two independent reviewers, we cannot entirely exclude the possibility that some studies were missed due to low citation frequency. Most reviewed studies, in particular small studies, shared one or several important methodological shortcomings, including purposive or convenience samples, inconsistent operational definition of risk populations, short follow-up times, high attrition rates, and inappropriate periods of risk assessment. As already indicated, there was significant heterogeneity across included studies, part of which is likely to have been true heterogeneity. Finally, Egger's test can be sensitive to extreme observations and large sample size, both of which were present in this meta-analysis.

Our findings support the notion that comprehensive strategies are needed to control the spread of parenterally and sexually transmitted viruses in US correctional populations. Examples are prevention programs to reduce transmission within prison systems [129]; transition programs to better prepare inmates for life after discharge; and interventions to ensure continuity of care in the community. While incarceration does not appear to increase the risk for HIV or other blood-borne disease infection for the average inmate, and while rates of transmission in US correctional settings appear to be lower than would be expected outside prison or jail, the lack of sufficient data for meta-regression and sub-group analyses made it difficult to draw definitive conclusions about the increased risk incarceration poses on high risk groups such as MSM and IVDU, or about the increased risk to these groups and the community during periods of release. In addition to supporting innovative intervention studies, our findings also point to the need for further research to update our understanding of the transmission of blood-borne and sexually-transmitted infections in inmate populations, and, most importantly, to clarify the role of the post release period in infection risk and further spread to the general community. The consistency of our results across infectious agents confirms that HBV and HCV infections may be used as sentinel indicators of risk for HIV infection in correctional settings.