The prevalence of non-communicable disease in older people in prison: a systematic review and meta-analysis

Abstract

Background

people in prison often experience poor health. Those aged 50 and over are the fastest growing age-group in prison and present particular challenges to criminal justice systems around the world. Non-communicable diseases (NCDs) account for two-thirds of deaths globally and no estimate of the prevalence of NCDs in this vulnerable population exists.

Methods

we searched PubMed, Medline, CINAHL, EMBASE and Global Health databases to identify original research papers that met our pre-defined inclusion criteria. No date or language restrictions were applied. Two authors undertook full-text screening as well as quality assessment and data extraction for all included studies. A random effects model was used to calculate pooled prevalence of any disease that was reported in two or more articles.

Results

the initial search identified 2,712 articles. 119 underwent full-text screening with 26 meeting the inclusion criteria. This provided prevalence data on 28 NCDs in 93,862 individuals from prisons in 11 countries. Pooled prevalence for the most significant NCDs was a follows; cancer 8% (95% CI 6–10%), cardiovascular disease 38% (95% CI 33–42%), hypertension 39% (95% CI 32–47%), diabetes 14% (95% CI 12–16%), COPD prevalence estimates ranged from 4% to 18%. Heterogeneity across studies was high.

Conclusions

those in prison over 50 years of age experience a high burden of NCDs which is often higher than younger prison and age-matched community peers. This health inequality is influenced by lifestyle, environmental and societal factors. Prison services should be adapted to serve the needs of this growing population.

non-communicable Disease (NCDs), older people, prison, prevalence, systematic review

Key points

• Those aged 50 years and over are the fastest growing population group in prisons.

• People in prison already experience well documented health inequalities.

• Prevalence of Non-Communicable Disease in older people in prison is higher than younger prisoners and higher than age-matched co.

• Prison regimes and services should be adapted to support this population and provide equivalence of care.

Introduction

There are 10.4 million people in prisons or other prescribed places of detention across the world [1] and this population experience poorer health than their community peers [2]. For example, the odds ratio for prisoners developing major non-communicable diseases (NCDs) such as hypertension, cancers, asthma and arthritis is between 1.2 and 1.8 when compared to age-matched community peers [3] and rates of cardiovascular disease, stroke and Chronic Obstructive Pulmonary Disease (COPD) three times higher [4]. This health inequality is compounded by the ageing process [5], with older people in prison often having morbidity rates and functional abilities similar to peers in the community 10 years their senior [6–8]. It is likely the physical environment also adversely impacts on the functional abilities of older people in prison because most prison buildings around the world are constructed without consideration of age or infirmity. In prisons, therefore, those aged 50 years and over are widely considered to be ‘older prisoners’ by practitioners [6, 9], policy-makers [10–12], researchers [13, 14] and national advocacy groups [12, 15, 16].

In many countries, the number of people in prison defined as “older” is increasing at a disproportionately high rate. In the UK, the proportion of people in prison in this age group has increased by 169% over the last 15 years to 15.6% of the whole prison population [17]. Although this is a lower figure than proportion of people aged 50 or over in the UK overall (36%), the 169% rate of growth far exceeds the 21% growth of those aged 50 and over seen overall in this population over the same period [18]. There will be an estimated 14,800 older prisoners by 2021, equivalent to nearly 1 in 5 prisoners being aged 50 years and over [19, 20]. In the USA, the proportion of prisoners who are considered older has increased from 10.3% to 19.2% [21, 22].

Whilst the ageing population in the UK and the USA partly explains the increasing older population, it is also a result of factors including; historical sentencing practices, increases in sentence length and increases in late-in-life prosecutions, often for historic sex offences [23]. This presents challenges for healthcare teams and policy-makers. The complexity of providing care to these older people is significant, with up to 90% diagnosed with at least one long-term health condition [7]. These individuals are managed in an environment designed for fit, young men that struggles to adjust to the unplanned roles of care home or hospice [23]. Furthermore, the healthcare cost of incarcerating these individuals is estimated to be 250% more than for younger people [24].

Systematic reviews have estimated the prevalence of NCD risk factors in the prison population [25], substance abuse and addiction [26] and the psychiatric morbidity of older people in prison [27]. No review exists of the prevalence of non-communicable physical disease within this vulnerable population. We therefore have synthesised data on the prevalence of NCDs in older people in prison.

Method

Search strategy and selection criteria

A protocol for this study was registered at PROSPERO (http://www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42017075171). We searched PubMed, Medline, CINAHL, EMBASE and Global Health electronic databases for relevant articles in March 2018. The full search strategy for PubMed is available in Appendix 1. We hand-searched reference lists of included studies, performed a forward citation search and a grey literature search via Google, the websites of the UK, the USA and Australian governments and of relevant third sector organisations.

Included studies contained; participants currently serving a custodial sentence, prevalence data on any NCD (as defined by the Global Burden of Disease Study [28]) excluding oral health, and data stratified into an older person group (minimum age for this group was 50 years). We imposed no language or date restriction on the search (full version in Appendix 2 in the supplementary data, available at Age and Ageing online). D.M. and E.P. independently screened title and abstracts (n = 2,712) and the full-text of 119 articles. We resolved any disagreements by discussion and J.L. resolved any conflicts.

Quality assessment and data extraction

D.M. and E.P. independently quality assessed and extracted data from the 26 included studies using a modified version of the Newcastle–Ottawa scale [29] scores for all studies across all nine domains are presented in Appendix 3 in the supplementary data, available at Age and Ageing online. Authors were contacted for further primary data where this might allow inclusion of their study. Results were recorded according to PRISMA guidelines [30].

Statistical analysis

The outcome measure of interest was the proportion of prisoners with a specific NCD. We undertook meta-analyses for each disease where prevalence was reported in two or more studies using MetaXL (http://www.epigear.com/index_files/metaxl.html) with a random effects model [31]. We generated Forest plots for each disease with a weighted pooled prevalence estimate and 95% confidence intervals (CIs). I² and Cochran’s Q (with associated P-values) were used to investigate heterogeneity between studies. Sub-group analysis was used to investigate potential sources of heterogeneity [31]. Pre-specified sub-group analysis was undertaken for each disease according to participants’ gender (Male vs. Female), age group (50–59 vs. 60+) and country of study, where sufficient studies reported data. Sensitivity analysis was undertaken to investigate the impact of bias within the prevalence estimates by removing studies judged to be weak in the outcome assessment domains and we re-analysed data using a quality effects model. Where available, comparative data estimating disease prevalence in younger people in prison and prevalence in age-matched community peers, was extracted See Table 2a. We identified general community prevalence data for older people from high-quality international sources [28, 52]. These data are presented descriptively in Table 2b.

Table 2a.

Comparator data between older and younger prisoners

Disease	Study	Prevalence in older prisonersa		Prevalence in younger prisonersb
		%	(95% CI)	%	(95% CI)
Arthritis	Stewart (2015)	18	(14–23%)	6	(5–7%)
Asthma	Harzke (2010)	5	(4–5%)	5	(5–5%)
	Moschetti (2015)	3	(1–7%)	2	(1–3%)
	Stewart (2015)	11	(8–15%)	14	(13–16%)
	Voller (2011)	1	(0–2%)	2	(1–3%)
	Voller (2016)	2	(1–3%)	2	(1–2%)
Cancer (any site)	Nolan (2014)	10	(1–24%)	6	(3–8%)
	Stewart (2015)	7	(4–10%)	1	(1–1%)
Cardiovascular disease (all)	Fazel (2001)	35	(29–42%)	3	(2–4%)
	Moschetti (2015)	33	(25–41%)	4	(3–5%)
	Nolan (2014)	47	(29–65%)	15	(10–19%)
COPD	Bania (2016)	14	(9–19%)	2	(1–4%)
	Harzke (2010)	8	(7–8%)	1	(1–1%)
	Voller (2011)	4	(2–6%)	1	(0–2%)
Diabetes	Bautista-Arredondo (2015)	15	(13–17%)	2	(1–2%)
	Harzke (2010)	17	(17–18%)	3	(3–3%)
	Moschetti (2015)	13	(8–1%)	1	(0–2%)
	Nolan (2014)	17	(5–32%)	4	(1–6%)
	Stewart (2015)	11	(8–15%)	3	(2–4%)
	Voller (2016)	10	(9–11%)	1	(0–1%)
Ischaemic Heart Disease	Davoren (2015)	14	(9–19%)	2	(0–4%)
	Harzke (2010)	13	(13–14%)	1	(1–1%)
	Tagirova (2010)	65	(52–77%)	28	(23–32%)
Hypertension	Bautista-Arredondo (2015)	27	(24–29%)	4	(4–5%)
	D’Souza (2005)	32	(18–48%)	11	(8–13%)
	Harzke (2010)	57	(57–58%)	16	(16–16%)
	Moschetti (2015)	26	(19–33%)	4	(3–5%)
	Stewart (2015)	23	(19–28%)	6	(5–7%)
	Tagirova (2010)	70	(57–82%)	45	(40–50%)
	Voller (2016)	22	(20–23%)	1	(1–2%)
Neurological illness	Davoren (2015)	23	(17–29%)	11	(7–16%)

^aLower limit of age for older prisoners was 50 for all studies in this table except for Harzke 2010 (55 years) and Davoren 2015 and Fazel 2001 (60 years).

^bAge range for younger comparison group varied between studies.

Table 2b

Comparator data between older prisoners and community peers of the same age

	50+ Prisoners		Males, Global high income countriesa
			50–69-year-olds
	%	(95% CI)	%				(95% CI)
Cancer	8	(6–10%)	3				(3–4%)
CVD	38	(33–42%)	20				(19–21%)
Diabetes	14	(12–16%)	14				(13–15%)
	50+ Prisoners		Males, Western high income countriesb
			50–54		55–59		60–64		65–69
	%	(95% CI)	%	(95% CI)	%	(95% CI)	%	(95% CI)	%	(95% CI)
Hypertension	39	(32–47%)	27	(22–33%)	33	(27–39%)	38	(32–44%)	42	(36–48%)

^aData from Global burden of disease study (2017) [28].

^bData from NCD Risk Factor Collaboration (2017) [52] is only available in 5-year-age bands and by smaller geographical region.

Findings

Search results

We identified 2,712 articles with 119 proceeding to full-text review and 26 meeting the inclusion criteria. See PRISMA flowchart in Figure 1. Included studies varied in the number of older participants they included (n = 30 to n = 28,580). Half [13] reported outcomes on people in the USA [3, 4, 10, 11, 32–35, 38, 41, 42, 50, 51] with the rest from elsewhere in North America [37, 45, 55], or from the UK [5, 6], Ireland [39], Continental Europe [36, 43, 44, 47–49] or Australasia [40]. The majority of studies defined older prisoners as those aged at least 50-years-old [3, 6, 10, 11, 32–34, 36–38, 42, 44, 45, 47–49, 52, 55], although some used 55 [4, 40, 41, 43, 50, 51] or 60 years [5, 39] as the lower age limit. See Table 1 for full characteristics.

Table 1.

Characteristics of studies

Study	Study design	Location	N	Age range	Gender	Method of Ax	Comparative data
Aday (2014)[32]	Cross-sectional	South West USA	327	50 and over	F only	Self-completed questionnaire	Nil
Baillargeon (2000)[33]	Cohort	Texas, USA	13,027	50 and over	M & F	Assessment at reception and subsequent medical encounters	Nil
Baillargeon (2007)[34]	Cohort	Texas, USA	28,580	50 and over	M & F	Assessment at reception and subsequent medical encounters	18–39-year-old prisoners
Baillargeon (2009)[35]	Cohort	Texas, USA	26,820	50 and over	M only	Assessment at reception and subsequent medical encounters	18–39-year-old prisoners
Bania (2016)[36]	Cross- sectional	Greece	180	50 and over	M & F	Spirometry testing according to international guidelines	30–49-year-old prisoners
Bautista-Arredondo (2015)[37]	Cross- sectional	Mexico City, Mexico	1,117	50 and over	M & F	Clinical measurements of BP and DM and medical histories	Gen. pop from Mexican via Health and Nutrition Survey
Binswanger (2009) [3]	Cross- sectional	USA	2,313	50-65	M & F	Structured interviews	USA Gen. population via National Health Interview Survey
Colsher (1992)[38]	Cross- sectional	Iowa, USA	119	50 and over	M only	Standardised questionnaire	Nil
Davoren (2015)[39]	Retro cohort	Republic of Ireland	213	60 and over	M &F	Medical notes review	Prisoners less than 60-years-old
D’Souza (2005)[40]	Cross- sectional	NSW, Australia	57	55 and over	M & F	Clinical measurements and medical history at interview	Australian gen. population, via AusDiab study and 25–54-year-old prisoners
Fazel 2001 [5]	Cross- sectional	Southern England	203	60 and over	M only	Structured interviews and medical notes review	65–74-year-old community dwelling men and prisoners aged 18–49
Greene 2015[41]	Cross- sectional	San Francisco, USA	238	55 and over	M only	Not defined	USA Gen. population via Health and Retirement Study
Harzke 2010 [4]	Cohort	Texas, USA	13,500	55 and over	M & F	Assessment at reception and subsequent medical encounters	US 2000 standard population and prisoners under age of 55
Hayes 2012 [6]	Cross- sectional	North West England	262	50 and over	M only	Structured interview using the Burvill Grid	Nil
Leigey 2012[42]	Cross- sectional	USA	1,139	50 and over	M & F	Semi-structured interview	Nil
Marquart 2000 [10]	Cross- sectional	Texas, USA	69	50 and over	M only	Structured interviews	Nil
Meyer 2016[43]	Cross- sectional	Rheinland-Pfalz, Germany	113	55 and over	M only	Self-completed questionnaire	Gen. pop of Germany via national survey on ageing (DEAS)
Moschetti 2015[44]	Cohort	Canton of Vaud, Switzerland	136	50 and over	M & F	Clinical measurements and medical history at reception	General Swiss pop and prisoners under age of 50
Nolan (2014)[45]	Cross- sectional	Canada	30	50 and over	F only	Structure screening interview with nurse, and clinical measurements	Prisoners less than 50 years-old
Nowotny (2016) [11]	Cross- sectional	USA	1,206	50 and over	M only	Computer assisted personal interview	Nil
Stewart (2015)[46]	Cross- sectional	Canada	302	50 and over	M only	Structure screening interview with nurse, and clinical measurements	Prisoners less than 50-years-old
Tagirova (2010)[47]	Cross- sectional	Russia	54	50 and over	F only	Clinical measurements and self-reported	Prisoners less than 50-years-old
Voller (2011)[48]	Cross- sectional	Tuscany, Italy	380	50 and over	M & F	Review of medical notes	Gen population of Italy via national survey and 18–49-year-old prisoners
Voller (2016)[49]	Cross- sectional	Italy	3,167	50 and over	M & F	Review of medical records	Prisoners aged 18–39 and aged 40–49
Williams (2006)[50]	Cross- sectional	California, USA	120	55 and over	F only	Structured interview	Nil
Williams (2014)[51]	Cohort	California, USA	210	55 and over	M & F	Structured interview and medical note review	Nil

Figure 1

PRISMA flow diagram. *For eligibility criteria see Appendix 1 in the supplementary data, available at Age and Ageing online.

Pooled prevalence

Prevalence data from 93,862 participants aged 50 or over, reported across the 26 studies, covering 28 different NCDs were included in the quantitative analysis. These data are presented in Appendix 3. Eleven diseases were each only reported in one study and were not included in the meta-analyses.

Nine studies reported cancer prevalence (in any site) [3, 11, 38, 42, 43, 45, 46, 50, 51]. Pooled prevalence was 8% (95% CI 6–10%; I² 84%) (Figure 2). Prevalence appeared higher in females (10%; 95% CI 6–15%) than males (7%; 95% CI 4–11%) but this difference was not statistically significant. Analysis by age was not possible due to a lack of data. Cardiovascular disease (CVD) was reported as an umbrella diagnosis in four studies [5, 44–46]. Pooled prevalence was 38% (95% CI 33–42% I²; 30%) (Figure 2) and there was insufficient data to undertake sub-group analyses. Hypertension was the most commonly reported single disease, with 21 out of 26 studies reporting prevalence [3–5, 32, 33, 37, 38, 40, 41, 46, 47, 49–51] (Figure 2). Prevalence estimates appeared higher in females [4, 32, 33, 37, 42, 47, 50] than males [4–6, 10, 11, 33, 37, 38, 41–43, 46]; 55% (95% CI 47–63%) vs. 36% (95% CI 27–46%). As heterogeneity was high, results should be interpreted with caution.

Figure 2

(a) Forest plot of hypertension prevalence. (b) Forest plot of diabetes prevalence. (c) Forest plot of cancer prevalence. (d) Forest plot of CVD (all) prevalence

Diabetes was the second most commonly investigated disease, in 19 out of 26 studies [3–6, 10, 11, 33, 37, 38, 41–46, 49–51]. Pooled prevalence was 14% (95% CI 12–16%; I² 94%) (Figure 2) Sub-group analysis demonstrated female prevalence of 17% (95% 13–21%) [4, 33, 37, 42, 45, 50] and male prevalence of 14% (95% CI 12–17%) [4–6, 10, 11, 33, 37, 38, 41–43, 46]. In those aged 50–59 prevalence was 10% (95% CI 7–13%) [6, 37, 38, 49] compared to 17% (95% CI 12–23%) in 60+ year olds [5, 6, 37, 38, 49]. These differences were not statistically significant.

Asthma was reported in nine studies [3–6, 11, 33, 38, 42, 44, 46, 48, 49] and COPD in six [4–6, 36, 38, 48]. The prevalence of asthma ranged from 1% [48] to 17% [6] and COPD from 4% [48] to 18% [38]. Musculoskeletal diseases were reported as a group in four studies [5, 44, 45, 48]. Pooled prevalence was 23% (95% CI 20–26%; I² 0%). Arthritis as a singular disease was reported in 13 studies [3, 5, 6, 10, 11, 32, 33, 38, 42, 43, 46, 50, 51]. Pooled prevalence was 37% (95% CI 24–51%; I² 99%). Sub-group analysis by method of case ascertainment (self-reported and [3, 5, 6, 10, 11, 32, 38, 42, 43, 50, 51] vs. medical notes review [33, 55]) showed pooled prevalence of 41% (95% CI 34–50%; I² 97%) and 13% (95% CI 7–22%; I² 94%) respectively. The prevalence of any neurological condition was 14% (95% CI 4–27%; I² 98%) [5, 6, 39, 44–46, 49], but heterogeneity was high. Two studies reported ‘kidney problems’ [11, 42] with a pooled prevalence of 7% (95% CI 2–15% I²; 97%). The prevalence of ‘prostate problems’ varied from 6% [48] to 20% [38]. The prevalence of incontinence was 22% (95% CI 17–27% I² 0%) [38, 50].

The prevalence of impaired vision ranged from 20% [11] to 84% [32], whilst 31% (95% CI 14–50% I²; 99%) of older people had impaired hearing [6, 10, 11, 32, 38, 41, 50].

Sensitivity analysis to explore the impact of poor study quality on our findings did not reveal significant changes in prevalence rates.

Comparator data is presented in Table 2a and b demonstrating that disease prevalence is consistently higher in the older compared to younger people in prison; COPD was eight times higher (8% vs. 1%) [4], diabetes ten times higher (10% vs. 1%) [49], ischaemic heart disease 13 times higher (13% vs 1%) [4] and hypertension over 20 times higher (22%% vs 1%) [49]. Some studies compared older prisoners with age-matched community peers. When our pooled prevalence rates are compared to similar age-group from the community (males aged 50–69); prevalence is at least double among prisoners.

Discussion

This is the first study to examine the prevalence of NCDs across the ageing prison population globally. Despite the aim to review data on the global ageing prison population, only one study from a low or middle income country was included. The findings demonstrate that those diseases responsible for the greatest morbidity and mortality in the general population [28] have particularly high prevalence rates in this population: 38% of older prisons had cardiovascular disease, 8% had been diagnosed with cancer and diabetes estimates varied between 8 and 21%. Impaired vision affected 59% of people and estimates of hearing loss varied from 14 to 66%. The burden of diseases like ischaemic heart disease, diabetes and COPD were up to 20 times higher in the older compared to younger people in prison.

Rates of NCDs were higher in older prisoners than community peers of the same age, from similar high income countries. For example; cancer (8% compared to 3–4%) and CVD (38% vs. 18–20%) were around twice as high. Hypertension was higher in the older prisoner cohort (39%) than community 27–42%. Diabetes prevalence among 50–69-year-olds in prison and those in the community is the same (14%) although this masks variation in community rates between UK (7%) and N. America (17%), a trend not seen in the older prison population where there is no difference in prevalence rates between these countries.

The high prevalence of NCDs in this population is striking. This may be explained by the risk-factor profile of this group. Alcohol abuse disorder is estimated to be 24% in adult prisoners globally [53] and age-specific rates are highest in older prisoners [54, 55]. Opportunities for physical activity are limited in prison with as many as 9 in 10 not sufficiently active enough to benefit their health [56]. Poor diet compounds this with excessive fat and sodium intake reported [25], whilst 86% of people in prison do not eat at least five daily portions of fruit and vegetables [56]. Smoking prevalence is high in prisons- up to three times that of community levels [57].

Mental ill-health is associated with physical ill-health [58]. The estimated prevalence of psychiatric illness in older prisoners is 38% [27] and although the interaction between the two disease groups is complex, it offers a further possible explanation for our findings.

The nature of incarceration means that people in prison have less control over certain risk factors than their community peers; they have little choice over what they eat or for how long they exercise. Furthermore, many of these risk-factors are socially determined and are not simply lifestyle choices. The environment people are born and live in is a strong determinant of the prevalence of these risk-factors [59]. The health inequalities people in prison experience may be because those from the lowest socioeconomic groups are over represented in the prison population [60], with such groups generally experiencing a higher burden of disease than those with higher socioeconomic status [59]. The community prevalence data used by studies as a comparator represented a cross-section of society and not just lower social-economic groups, but even when a comparison is made between older prisoners and age-matched community members from the lowest socioeconomic status quintile, prisoners’ disease prevalence remains higher [41]. It is important therefore that the unique opportunity that prisons present to effectively promote health, as well as address the wider determinants of health, with this vulnerable group of people should be fully utilised [7].

To the best of our knowledge, this is the first study to calculate pooled prevalence of NCDs in the ageing prison population. This will provide valuable information for policy-makers, commissioners and researchers. However, several studies were excluded because of the use of 50-years-old and older as a criterion for inclusion, excluding large datasets from the UK, the USA and Australia government surveys. Furthermore, it was not always possible to obtain original data from authors which meant that studies that had potentially relevant prevalence data, but had not published it as age-stratified, were excluded and the extent to which sub-group analysis could be undertaken was limited. The absence of consistent comparative data in the included studies precluded formal statistical analysis. Descriptive comparison between prison and community prevalence in 50–69-year-olds was made but it is acknowledge that only ~88% of older prisoners are within this age range (12% are aged 70+). Many studies were also underpowered to demonstrate a significant result between older people and younger or community peers. Lack of data meant we could not examine differences in prevalence between men and women with the exception of cancer, diabetes and hypertension. Although the prevalence of all three diseases appeared to be higher in women, high heterogeneity or overlapping CIs precluded meaningful conclusions about differences. This is not unexpected given that women comprise only 6% of the prison population worldwide [1] but also highlights a vulnerable minority population within prison meriting further attention [61].

It was not always possible to identify the causes of heterogeneity, despite undertaking sensitivity analysis. Study quality was not an important source. However, within some diseases, such as arthritis and asthma, that case ascertainment methodology (self-reported diagnosis vs review of medical notes) appeared to determine the prevalence rates observed. This phenomenon has previously been noted elsewhere [5]. Similarly, participants were defined as hypertensive in varying ways, e.g. one-off measurements, 24-h monitoring and self-report. These differences might have produced different prevalence estimates. Furthermore, some ‘diseases’ lacked clear definitions. For example, it was not clear if “prostate problems” related to a diagnosis of benign prostatic hyperplasia, or the presence of urinary symptoms.

Differences exist between penal systems globally and people detained in different geographical locations may vary in terms of pre-incarceration determinants of health, healthcare access within and prior to prison and exposure to risk-factors whilst detained. It was not however possible to explore this here.

Providing effective support for the ageing prison population poses a significant challenge to criminal justice systems across the world. Our data contributes to the evidence base on which policies and approaches to tackle these challenges should be based. The prevalence data should inform the development of appropriate, needs-based health services, and to ensure services are accessible, acceptable and comprehensive enough to reflect the complex needs of this population. The data on conditions such as back pain, sensory impairment and incontinence demonstrate the need for a “whole prison” approach to health, as the majority of the required management and adaptation for these conditions sits not only with healthcare services, but the wider prison regime. This includes providing age-appropriate physical activity, purposeful activity, methods of communication and adaptations of the built environment. Further prison-based research is required to determine effective interventions in this unique setting and should be informed by what is known to be effective in addressing health inequalities and improving health in older people in the community.

Conclusion

This systematic review identified a high prevalence of NCDs in the ageing prison population. This growing population is particularly vulnerable and experiences significant health inequalities. However, it is possible to address these through the provision of appropriate health, social care and custodial services and policy-makers and commissioners of prison services should ensure this happens. The principle of equivalence of care for those in prison is fundamental in this regard [62, 63]. In line with a number of other systematic reviews, this review highlights the dearth of evidence from low and middle income countries and the need for further research examining the prevalence of disease and health needs of older people in prison globally.

Acknowledgements: The authors would like to acknowledge the input of Nia Wyn Roberts for advice on building the search strategy and Dr Emma Munday and Dr Gracia Fellmeth for translating non-English language journal articles.

Declaration of Conflict of Interest: None.