Search strategy and selection criteria

A search string was agreed to identify articles on the association between cell spatial density and infectious and communicable diseases and mental health outcomes (table 2). For the present paper, articles reporting on infectious and communicable disease outcomes were extracted from the final list of eligible studies. Reviews of the mental health outcome studies are reported elsewhere.29 Infectious and communicable diseases are defined ‘as an illness due to a specific infectious agent or its toxic products that arises through transmission of such agent or products from an infected person, animal or reservoir to a susceptible host, either directly or indirectly through an intermediate plant or animal host, vector or inanimate environment’.30

The scope of the search was limited to the availability of an English abstract and to human subjects. Publication date was not restricted. Studies were excluded if they did not specify cell size or a measurement of cell spatial density, examined only temporary accommodation within custodial settings such as segregation cells, police cells, lockups, prison transport/transfer units and prison clinic accommodation. Studies examining combined living and sleeping space preventing the differentiation of sleeping space were also excluded from the current review.

Fourteen medical, social science and architectural databases were searched: EMBASE (1947–2018), PubMed (≈1800–2018), Medline (1946–2018), Scopus (1976–2018), PsycINFO (1806–2018), PsycExtra (1908–2018), Web of Science (1900–2018), ProQuest Databases (1763–2018), ProQuest Dissertations and Theses Global (1716–2018), Index to Legal Periodicals (1985–2018), Informit Online (1920–2018), Cochrane Library (≈1900–2018), Criminal Justice Abstracts (1910–2018) and ICONDA (1976–2018). Grey literature was also identified from these database searches; and a further search for grey literature was conducted using Google and selected websites including those of the United Nations, International Committee of the Red Cross, Amnesty International, International Centre for Prison Studies, WHO, Campbell Collaboration, the Criminal Justice Reference Service, the US National Library of Medicine and websites of the Australian Inspector of Custodial Services. A number of international experts in the prisoner health area were contacted to inquire as to whether they were aware of additional material of potential interest to the review.

To manage records identified, articles were exported from databases into EndNote V.X7.1. Grey literature from non-database sources were entered into Endnote manually. PDF versions of articles were located and attached to Endnote items.

Study eligibility

All publications identified by the search were first screened by two independent teams of two reviewers (PLS, AA and MS, LG). To ensure consistency in the screening process and selection of eligible studies across the teams, titles and abstracts of the first 100 studies were independently reviewed by both teams and discussed as a group to ensure consistency. After achieving a consensus on eligibility for full-text review across both teams, the remaining articles were divided between the two teams of reviewers. Each person within the team independently reviewed their assigned literature, and on finishing, discussed the eligibility outcome with the other team member. In those instances where there was disagreement within a team, both teams met to resolve the issues and decide through consensus if the article was eligible for full-text review.

Articles were excluded from full-text review if the study design did not feature in the Australian National Health and Medical Research Council’s (NHMRC) Hierarchy of Evidence.31 This hierarchy includes systematic reviews of prospective cohort studies (level I evidence), prospective cohort study (level II evidence), ‘all or none’ study (ie, either all or none of the people with the risk factor(s) experienced the outcome of interest) (level III-1 evidence) retrospective cohort study (level III-2 evidence), case–control study (level III-3 evidence), or cross-sectional study or case series (level IV evidence).31 Articles were also excluded if they did not conduct an analysis that examined the relationship between the exposure variable of prison cells accommodating one or more persons with a specified cell spatial density or cell dimensions and an outcome variable of an infectious and/or communicable disease.

Critical appraisal of eligible studies

Critical appraisal of eligible studies involved three stages: (1) data extraction,32 (2) assessment of bias, chance and confounding33 and (3) an aggregated assessment of all studies using the NHMRC FORM Matrix and Evidence Statement to produce a final rating and evidence statement.31

An adapted version of the NHMRC’s standardised data summary table was used to extract data.32 Data extraction tables were prepared and cross-checked by two independent reviewers (PLS and MS). Extracted information included: general study details (citation, study design, length of follow-up, level of evidence and location and setting); study exposure and control descriptions; population characteristics; internal and external validity considerations; and study outcomes and results.

Assessment of bias, chance and confounding was guided by the NHMRC’s checklist to critically appraise aetiology or risk factor studies.33 Checklist items used to guide the assessment of studies included: exposure misclassification, outcome misclassification, selection bias, confounding and chance. Items were assessed on one of two 3-point scales: low (0), medium (1), high (2) risk; or poor (0), fair (1), good (2).

A systematic assessment of the overall body of the evidence was conducted using a modified (for study population relevance) NHMRC FORM Matrix and Evidence Statement covering34: (1) evidence base, (2) consistency of results, (3) population health impact and (4) generalisability. Each component was rated from excellent to poor (online supplementary appendx A) to allow a final overall rating and evidence statement to be derived ranging from ‘A’ to D’: ‘A’ indicating that the body of evidence can be trusted to ‘D’ indicating that the body of evidence is weak, and findings cannot be trusted (online supplementary appendx B).31 The final overall rating and evidence statement were undertaken by one reviewer (PLS) and cross-checked by another (MS).

Following the completion of the review by the team, an expert, independent of the reviewers and advisory panel, peer reviewed the methodology and findings.

Patient and public involvement

Patients and the public were not involved in the review.

Evidence of an association between cell spatial density and infectious and communicable diseases

While the review found mostly consistent evidence that cell spatial density is associated with clinically verifiable infectious and communicable diseases, with five articles reporting a statistically significant association between cell spatial density and infectious and communicable diseases (table 3), all but one of these associations could possibly be explained by chance, study bias or confounding.

In the study which investigated MTB infection in a random sample of 425 male prisoners in five Pakistan prisons, the association between cell spatial density and MTB was reported to be statistically significant (adjusted OR (aOR) 2.6; 95% CI 1.6 to 4.3) after adjusting for age, education, length of incarceration, tobacco use and other variables.36 However, caution is warranted when interpreting this result due to possible exposure and outcome misclassification. That is, the cell spatial density measure was a dichotomised average floor area per person (ie, >5.6 m² per person and ≤5.6 m² per person) and thus may be a crude measure. The study does not report on how this measure was determined. Additionally, the 100% response rate to the screening suggests possible coercion to participate in the study that may have affected the reliability of self-report data ta (eg, household income, occupation).

The LTBI study was conducted in 46 prisons in Chile and included 418 prisoner and non-prisoner contacts of 33 active TB cases recruited 12 months following the identification of active TB cases.39 The study found high TB incidence (123.9 per 100 000 prisoners) and high LTBI prevalence (29.4%) among contacts, with LTBI rates significantly higher in prisoners than in non-prisoners (33.2% vs 15.6%). Multivariate analyses showed a significant association between cell spatial density and LTBI (aOR 3.5; 95% CI 1.1 to 11.5) after adjusting for gender, age, illicit drug use, history of corticosteroid use and malnutrition, and length of incarceration. Assessing the extent of potential exposure misclassification was difficult as the spatial density measure used in analysis is not reported (unclear if continuous or categorical variable used). Cell type (ie, multiple-person cells, dormitories) across the different prisons was also not reported.

One US study investigating a pneumococcal disease outbreak35 examined the disease attack rate and its association with cell spatial density. Although more clarity and consistency in reporting of the analysis is needed to properly assess the findings, the results indicate that the attack rate was highest in four-person cells (2.9 m² per person) compared with the two other cell types described (single cells at 4.2 m² per person and open dormitories at 2.6 m² per person). The study also reported that detainees housed in shared and single areas ≥7.4m² per person (8% of the inmate population) had significantly fewer cases of pneumococcal disease than all other detainees (attack rate: 0 vs 4.7 per 1000 persons; p=0.03). This lower attack rate for detainees housed in shared areas ≥7.4 m² per person may in part be explained by time spent outside cells as those housed in this area were engaged in work programmes and left their cells to undertake such work. Results were inconsistently reported across cell types with both univariate and aORs reported for the different cell types suggesting that some results were adjusted for and some not.

In the two studies conducted with mostly male (97%) prisoners in Nigeria, the prevalence of infectious and non-infectious skin conditions among prisoners living in single cells (0.9 m² per person) and dormitories (2.4 m² per person) was significantly different at 61.7% vs 43.2%37 and 82.9% vs 69.7%.37 38 However, the quality for these studies was assessed as ‘poor’ due to the absence of any multivariate analysis, no details of whether those conducting the physical examinations were blinded to the cell assignments of the prisoners, no response rates or sampling method described, and cell type details derived from self-report questionnaires.

Two other US articles reporting on the association between cell spatial density and illness reporting at the prison clinic were from a research programme spanning approximately 10 years reporting on data collected from 1400 prisoners from six federal prisons.27 40 Illness reporting rates were determined by dividing the number of days a participant was in a cell type (with a specific cell social or spatial density) by the number of visits to a clinic. Not all outcomes were measured at and/or reported for all prison sites, and prisoner population characteristics, sampling methods, response rates and analytical procedures are not consistently reported across and within publications. One article, the Danbury prison study, reported results on illness reporting by contagious status,27 covering: venereal disease; eye, ear, nose, throat infections; cough, cold, influenza; gastrointestinal, stomach problems; skin conditions; virus; chills and fever. However, the lack of detail on the statistical analysis means it is unclear if potential confounders and effect modifiers were adjusted for. Nonetheless, the findings reported a significant difference (although p values were not reported) in contagious illness reporting between those in single occupant cells of 4.5 and/or 5.6 m² per person (0.03 reports per week) and those in dormitories with a spatial density of 4.6 and/or 5.5 m² (0.06 reports per week) per person. As the reported cell spatial area range of these two cell types are nearly identical, this is likely to say more about the effect of social density (the number of individuals per cell) or cell type (dormitories vs singles) than it does on cell spatial density.

The second US study from this research examined cell spatial density in single occupancy cells alone in two federal prisons.40 Using multiple linear regression analysis, statistically significant effects were reported in only one of the prisons. However, for contagious illness, the cell spatial density variable was reported as not significant (p=0.88). However, the validity and reliability of the outcome measure used are questionable as the measure may capture other parameters of health service use besides health effects.

Implications for researchers and policy-makers

Further research is needed that addresses the confounding, bias and chance elements in studies examined by this review. We recommend that future studies include clinically verifiable health outcomes to ensure outcome misclassification is prevented or minimised and include adequate detail on how cell floor spatial density was determined to facilitate a clear assessment of exposure misclassification.

In most articles reviewed, inadequate attempts were made to adjust for personal characteristics of prisoners and/or prison-related factors as potential confounders. This statement acknowledges that cell spatial density, as an objective condition of crowding, is likely to be embedded in a complex interplay of psychological, social, cultural and institutional factors, and that variations of any health effects cannot be fully understood without their consideration. Thus, future research should seek to collect more extensive data on prospective confounding and mediating factors to examine how such factors interact with cell spatial density and health effects. To assist this, we reviewed the 12 eligible articles identified in this review (7) and the mental health and well-being review (5),29 in addition to key international reports on prisoner conditions, prison crowding and prisoner health to identify what factors might mediate the association between prison cell spatial density and adverse health effects.42 From this review, we recommend that future research on cell spatial density and health effects consider the following factors: (1) personal characteristics of prisoners such as age, education level, gender, ethnicity, underlying medical conditions and health risk behaviours (eg, current intravenous drug use); (2) physical environment measures such as air ventilation and privacy afforded to the prisoner and (3) social and policy environment of prison such as time confined in cell per day, cell allocation policy and practice, health service access, length of incarceration and custody and security classification of prisoner.42

Although the body of evidence assessed indicated cautious support for the association between prison cell spatial density and infectious and communicable diseases, we do not advocate increasing social and spatial density and there is likely to be reasonable grounds for policy responses to address prison crowding in the absence of scientific certainty. As one US Court of Appeals43 ruling stated:

Undoubtedly, certainty is the scientific ideal—to the extent that even science can be certain of its truth. But certainty in the complexities of environmental medicine may be achievable only after the fact, when scientists have the opportunity for leisurely and isolated scrutiny of an entire mechanism. Awaiting certainty will often allow for only reactive, not preventive regulation.

In the absence of scientific certainty, international standards and recommendations provide guidance by way of mitigating prison and prison cell crowding and any potential infectious and communicable disease effects this may bring. The International Red Cross Association (ICPA) handbook on Water, Sanitation, Hygiene and Habitat in Prisons recognises that any technical focus on prisons to address potential adverse health effects, such as cell spatial density, cannot be separated from addressing other factors that mediate prisoner health and well-being.17 Some ICPA guidelines and recommendations are feasible in terms of implementation in the short term to address immediate need (eg, access to healthcare and sanitary facilities and extending time ‘out-of-cell’ to participate in meaningful activities). A public health approach to addressing any adverse health effects associated with prison cell crowding should include prevention strategies ranging from the micro to the macrolevel. The United Nations’ Office of Drugs and Crime Handbook on Strategies to Reduce Overcrowding in Prisons13 acknowledges that to address the health effects associated with crowding, ‘decarceration’ strategies to prevent prison crowding occurring in the first instance is required. Decarceration strategies involve processes that remove people from prisons and prevent them from (re)entering prison. This is likely to be a longer term strategy that will require extensive work involving multiple agencies from different sectors, and changes in political and societal attitudes to incarceration.