Abstract

Résumé

Introduction

On an average day in Canada, there are about 140,000 adults under correctional supervision, with one third in prison and two thirds in a community program (Reitano 2017). The responsibility for the administration of adult correctional services is shared between federal and provincial and territorial governments. Provincial and territorial governments have jurisdiction over adults who are held in custody while awaiting trial or sentencing, sentenced to less than 2 years, or serving community sentences (Reitano 2017). Ontario has the highest count of adults in provincial prison, with about 8000 adults on an average day (Reitano 2017).

Risk behaviours for HIV are common among persons in prison in Canada. At the time of admission, most adults report recent drug use and about 10% report having injected drugs in the months before admission (Kouyoumdjian et al. 2016; Smith et al. 2013; Zakaria et al. 2010; A health care needs assessment of federal inmates in Burchell et al. 2003; Canada 2004; Courtemanche et al. 2018; Kouyoumdjian et al. 2014 Martin et al. 2005; Robinson & Mirabelli 1996; Wood et al. 2005). While in prison, sharing needles for injection drug use and sharing tattooing and piercing equipment present ongoing risks for the transmission of HIV (Kouyoumdjian et al. 2016; A health care needs assessment of federal inmates in Buxton et al. 2009; Canada 2004; Ford et al. 2000; Gagnon et al. 2007; Robinson & Mirabelli 1996; Thompson et al. 2011). While only a minority of persons report sexual activity while in prison, people commonly report sexual risk behaviours prior to admission, including a large number of lifetime sexual partners, sex with partners at increased risk of blood-borne infections (e.g., persons who inject drugs), inconsistent condom use and involvement in commercial sex (Burchell et al. 2003; Courtemanche et al. 2018; Dufour et al. 1996; Kouyoumdjian et al. 2016; Martin et al. 2005; Poulin et al. 2007; Robinson & Mirabelli 1996; Smith et al. 2013; Thompson et al. 2011; Zakaria et al. 2010).

These risk factors contribute to higher rates of HIV in persons in prison than in the overall Canadian population. Data from several sources demonstrate that the prevalence of HIV in people in jails and prisons in Canada is between five and over 25 times that of the general population (Bonnycastle & Villebrun 2011; Calzavara et al. 1995; Calzavara et al. 2007; Correctional Service Canada 2016; Courtemanche et al. 2018; De et al. 2004; Dufour et al. 1996; Ford et al. 1995; Poulin et al. 2007; Rothon et al. 1994) (Table ​(Table1),1), which was estimated as 0.18% in 2014. (Public Health Agency of Canada 2016; Statistics Canada 2014).

Understanding the prevalence of HIV in persons in prison is important for several reasons. First, achievement of the UNAIDS targets to eliminate HIV is predicated on 90% of persons with HIV infection being aware of their diagnosis (Public Health Agency of Canada 2016). Limited evidence indicates that a substantial proportion of people with HIV in prison are unaware of their infection. A 2015 systematic review of studies in Canada and the United States found that only 78−79% of people with HIV in jails and prisons were aware of their diagnosis (Iroh et al. 2015). In voluntary testing studies of people in provincial custody in Quebec, 20% of persons with HIV did not know about their infection in 2003, (Poulin et al. 2007), and in 2014 and 2015, 10.5% of people with HIV who injected drugs and 85.7% of people with HIV who did not inject drugs were unaware of their HIV infection (Courtemanche et al. 2018). We need an updated prevalence estimate for persons in prison in Ontario to assess current performance relative to this first UNAIDS target. Second, rates of linkage to and retention in HIV care are considerably lower among formerly imprisoned people with HIV relative to the general population of people with HIV (Iroh et al. 2015). Estimating the prevalence of HIV, as well as the subsequent steps of the HIV care cascade, is necessary for developing interventions to improve linkage to care in prison and post-release.

Prior to our research group’s current research (Khanna et al. 2018), the most recent HIV prevalence for persons in Ontario prisons was estimated in 2004 using biological sampling (Table ​(Table1)1) (Calzavara et al. 2007). The purpose of the current study is to compare the observed prevalence of HIV in this population using administrative health data with the expected prevalence based on the most recently available biological sampling data from 2004.

Methods

Results

We achieved valid data linkage for 52,313 persons who were released from Ontario provincial prisons in 2010 (Table ​(Table2)2) (Kouyoumdjian et al. 2018). The majority of persons had been in provincial prison prior to the admission leading to the initial release in 2010, and the majority of persons spent 1 week or longer in provincial prison during the admission leading to the initial release in 2010.

Table 2

Demographic characteristics of individuals released from Ontario provincial prisons in 2010, by HIV status on initial release in 2010

Characteristic		HIV-positive persons N=363 n (%)	HIV-negative persons N=51,950 n (%)
Age	15–29	57 (15.7)	22,024 (42.4)
	30–39	121 (33.3)	13,172 (25.4)
	40–49	144 (39.7)	11,062 (21.3)
	≥50	41 (11.3)	5692 (11.0)
Sex	Female	62 (17.1)	6295 (12.1)
	Male	301 (82.9)	45,655 (87.9)
Neighbourhood income quintile	Missing	9 (2.5)	3186 (6.1)
	1	156 (43.0)	19,017 (36.6)
	2	75 (20.7)	11,005 (21.2)
	3	53 (14.6)	8109 (15.6)
	4	38 (10.5)	6181 (11.9)
	5	32 (8.8)	4452 (8.6)
Time in provincial prison, index admission*	<1 week	107 (29.5)	21,036 (40.5)
	1 week to <2 weeks	59 (16.3)	6515 (12.5)
	2 weeks to <4 weeks	44 (12.1)	5678 (10.9)
	4 to <12 weeks	72 (19.8)	8819 (17.0)
	≥12 weeks	81 (22.3)	9902 (19.1)
Number of admissions to provincial prison between 2005 and index admission*	0	83 (22.9)	22,885 (44.1)
	1	42 (11.6)	8559 (16.5)
	2	29 (8.0)	5595 (10.8)
	3	21 (5.8)	3971 (7.6)
	4	27 (7.4)	2927 (5.6)
	≥5	161 (44.4)	8013 (15.4)

Open in a separate window

*Index admission is defined as the period in custody leading to initial release in 2010

Applying the validated algorithm to this population, we identified 363 persons with HIV infection, of whom 17.1% were women. The overall observed HIV prevalence was 0.69% (Table ​(Table3),3), and the prevalence was 0.65% in men and 0.98% in women. The total expected number of cases of HIV in the 2010 cohort after applying and summing age-stratum-specific prevalence percentages was 1244 (95% CI, 474–2132), equivalent to a prevalence of 2.38%. The indirect age-standardized prevalence ratio was 0.29 (95% CI, 0.17–0.77).

Table 3

Observed and expected cases of HIV in persons released from Ontario provincial prison in 2010 by age stratum and age-standardized prevalence ratio

Age group (years)	Persons released from prison in 2010	Observed*		Expected†		Age-standardized prevalence ratio‡ (95%CI)
		Cases	Prevalence (%)	Cases (95% CI)	Prevalence (95% CI)
≤29	22,081	57	0.26	44.2 (2.2–198.7)	0.2 (0.01–0.9)	–
30–39	13,293	121	0.91	358.9 (159.5–558.3)	2.7 (1.2–4.2)	–
40–49	11,206	144	1.29	336.2 (134.5–549.1)	3.0 (1.2–4.9)	–
≥50	5733	41	0.72	504.5 (177.7–825.6)	8.8 (3.1–14.4)	–
Total	52,313	363	0.69	1243.8 (473.9–2131.7)	2.4 (1.3–2.8)	0.29 (0.17–0.77)

Open in a separate window

*Applying a validated algorithm to administrative health data at ICES (Antoniou et al. 2011)

†Data on prevalence from a biological sampling study by Calzavara et al. in 2003/2004 (Calzavara et al. 2007). Number of expected cases derived by multiplying age-specific prevalence rates from Calzavara et al. by the number of persons released from prison in 2010

‡For men and women combined

The observed HIV prevalence accounting for the sensitivity of the case-finding algorithm (i.e., 96.2%) was 0.72% (Fig. 1). Further adjustment for the expected number of individuals who do not know they are infected with HIV, which is estimated to be 20% (Public Health Agency of Canada 2016), increased the prevalence estimate to 0.90%. The indirect age-standardized prevalence ratios increased to 0.30 (95% CI, 0.27–0.34) and 0.38 (95% CI, 0.35–0.42), respectively, using the prevalences calculated from the sensitivity analyses.

Open in a separate window

Fig. 1

Prevalence of HIV in adults released from Ontario prisons in 2010, according to method of generating estimate. *Applying a validated algorithm to administrative health data at ICES (Antoniou et al. 2011) †From a biological sampling study by Calzavara et al. (2007) in 2004. ‡Sensitivity of the algorithm was 96.2% (Antoniou et al. 2011) §PHAC estimates that 20% of persons infected with HIV do not know they have HIV infection (Public Health Agency of Canada 2016)

Discussion

Using administrative health data, we estimated that 0.69% of persons released from Ontario prisons in 2010 were living with HIV, substantially lower than the expected prevalence of 2.38%. Considering the HIV care cascade, our expected prevalence reflects the number of persons infected with HIV, and the observed prevalence reflects the number of persons diagnosed with HIV and successfully linked to care. We found an indirect standardized prevalence ratio of HIV infection of 0.29 (95% CI, 0.17–0.77), meaning a validated algorithm identified less than one third of the persons with HIV that we would expect to identify based on the most recent biological sampling study (Calzavara et al. 2007; Antoniou et al. 2011). Sensitivity analyses adjusting for the sensitivity of the case-finding algorithm (Antoniou et al. 2011) and for individuals who do not know they are infected with HIV did not substantially alter the standardized prevalence ratio. The observed HIV prevalence was substantially lower than any estimate from prior studies conducted in prisons in Canada (see Table ​Table1).1). The observed and expected HIV prevalences were higher than the prevalences in the general Canadian population by four and 13 times, respectively (Public Health Agency of Canada 2016; Statistics Canada 2014).

Our study has several strengths. The sample was large and population-based, as we included all adults released from provincial prison in 2010 and achieved high rates of linkage to administrative health data. To generate the observed cases, we used a validated, sensitive and specific case-finding algorithm for identifying individuals with HIV infection from Ontario administrative health data (Antoniou et al. 2011). For the expected prevalence, we used data from a representative study sample of persons in Ontario provincial prisons, such that the population was comparable to our study population (Calzavara et al. 2007). The use of indirect standardization enabled an appropriately age-adjusted comparison of the observed and expected HIV prevalences in our study cohort. We think that the finding of a substantially lower HIV prevalence identified in administrative health data compared with biological sampling is likely generalizable to other jurisdictions that have not implemented universal HIV screening in prison.

Our study has several potential limitations. The algorithm used to identify persons with HIV was validated in primary care settings with a high prevalence of HIV of 23.1% (Antoniou et al. 2011), which was higher than that observed in this or previous studies of HIV prevalence in federal or provincial prisons (Table ​(Table1).1). Applying the algorithm in a population with a lower prevalence may increase the number of false positives, leading to overestimating the prevalence and moving the standardized prevalence ratio closer to 1. Administrative health data would not identify certain people with HIV, including recent immigrants who have not accessed provincially funded care, people who had been in federal prison (during which time they would not have accessed provincially funded healthcare) and people who had been diagnosed with HIV but were not accessing care. In Ontario, anonymous diagnostic testing for HIV is available at select clinics in the community, but not in prisons, and persons who tested positive anonymously may not have gone on to have nominal HIV testing or have accessed care for HIV. Any of these issues would lead to misclassifying people with HIV as not having HIV in our study, which would lead to an underestimate of the observed prevalence and move the standardized prevalence ratio away from the null. Laboratory and public health reporting data would identify more people who had been diagnosed with HIV infection, but we did not have access to those data in this study.

The finding of a standardized HIV prevalence ratio of 0.29 can have several possible interpretations with important implications for public health professionals, clinicians and policymakers. First, the true prevalence in 2003−2004 could have been lower than the prevalence identified in the Calzavara et al. study. We think this is unlikely given the large, representative sample used and rigorous methods of that study, as well as the consistency of the study findings with previous Canadian studies of HIV prevalence. Second, it is possible that the true HIV prevalence in persons in provincial prison decreased between 2003−2004 and 2010. A recent study in provincial correctional facilities in Quebec found no significant difference in HIV prevalence in men between 2003 and 2014−2015, though there was a significant decrease in HIV prevalence in women (Courtemanche et al. 2018); we note that women represent only 12% of persons released from provincial prison in Ontario. We think it is unlikely that the HIV prevalence decreased substantially over this period in people in this population in Ontario in the absence of significant policy and program changes in provincial prison and considering trends in the general Canadian population (Public Health Agency of Canada 2014), and especially unlikely to explain a 71% decline in HIV prevalence. If we assume the Calzavara et al. prevalence was accurate and our expected prevalence closely approximates the true prevalence of HIV in persons in Ontario prisons, several other issues may explain the standardized prevalence ratio we found. One possible explanation is that biological sampling identifies individuals who do not know they are infected with HIV, who would not be identified in administrative health data. Adjusting our estimate for the algorithm’s sensitivity and the expected number of individuals who might not know they have HIV infection, however, only increased the standardized prevalence ratio to 0.38. This suggests that the enhanced detection of biological sampling is not the sole factor and that the proportion of persons released from provincial prisons with HIV infection who do not know their status could be much higher than that in the general population. Another possible explanation is that persons with HIV infection in provincial prisons who know their HIV status may be less likely than the general population to access care related to their HIV infection. This would decrease the sensitivity of administrative health data-based approaches to identifying persons with HIV. HIV cascade of care estimates in persons released from prisons in the US and Canada compared with the general population provide support for this interpretation (Iroh et al. 2015). In related research, we found no difference in the median time to accessing HIV-specific ambulatory care for people after prison release compared with people in the general population (Khanna et al. 2018), though those data would not include persons with HIV who did not meet the algorithm’s threshold for number of times accessing care for HIV within 3 years (Antoniou et al. 2011).

Conclusion

The prevalence of HIV infection continues to be higher in persons in prison in Ontario than in the general population. The findings of our study suggest that high proportions of persons with HIV recently released from provincial prisons either do not know they have HIV infection or know about their infection but are not engaged in care, and that these proportions are much higher than those in the general population. Our study has explored the first two steps of the HIV care cascade. Describing all steps of the HIV care cascade for persons in prison in Ontario will form a critical research foundation for understanding inequities in HIV care for this population and for evaluating the performance of interventions to close health equity gaps. Time spent in prison presents an important opportunity for diagnosing HIV, linking people with care, and implementing evidence-based interventions to support indicated care (Iroh et al. 2015; Rich et al. 2013; Kamarulzaman et al. 2016; Rich et al. 2016; World Health Organization 2016). Prevention strategies should be adapted to be accessible for people with short lengths of stay, which are common in provincial prison, for example, providing universal opt-out testing on admission with low barrier testing options such as point-of-care and dried blood spot testing (Kamarulzaman et al. 2016). Enhanced HIV diagnosis and better understanding of barriers to accessing HIV services for persons in provincial prison are needed in order to meet the UNAIDS targets for the HIV care cascade (Public Health Agency of Canada 2016).

Footnotes

References