Bodkin Et Al, 2019
Bodkin Et Al, 2019
Bodkin Et Al, 2019
Background. A history of childhood abuse may affect people’s health and confidence interval [CI] = 52.6, 77.7; range = 56.2% to 75.0%) among
criminal justice system involvement. Understanding the prevalence of women; only one study reported the prevalence among men (35.5%). The
childhood abuse among individuals in prison is important to inform ef- summary prevalence of sexual abuse was 50.4% (95% CI = 33.5, 67.2;
fective and appropriate correctional services. range = 9.9% to 77.3%) among women and 21.9% (95% CI = 15.7, 28.8;
range = 8.3% to 55.6%) among men. The prevalence of neglect was 51.5%
Objectives. To review and summarize data on the prevalence of (95% CI = 43.1, 59.7; range = 45.5% to 65.1%) among women and 42.0%
childhood abuse among people experiencing imprisonment in (95% CI = 12.7, 74.6; range = 6.8% to 99.0%) among men. The prevalence
Canada. of physical abuse was 47.7% (95% CI = 41.3, 54.0; range = 16.3% to 83.0%),
and the prevalence of emotional abuse was 51.5% (95% CI = 34.8, 67.9;
Search Methods. We searched for studies in bibliographic indexes, range = 8.7% to 96.0%); we did not find differences according to gender.
reference lists, and gray literature, and we consulted experts. Prevalence estimates for all types of abuse showed high and unexplained
variability across studies.
Selection Criteria. We included studies published since 1987 that re-
ported data on prevalence of a history of abuse before the age of 18 years Conclusions. Half of people in prisons in Canada experienced abuse in
among people in Canadian prisons, including any abuse, physical abuse, childhood.
sexual abuse, emotional abuse, and neglect.
Public Health Implications. Prisons should incorporate trauma-
Data Collection and Analysis. Two authors independently reviewed informed approaches. Research is required to understand the associa-
titles and abstracts for eligibility and reviewed full texts for eligibility. tion between a history of childhood abuse and criminal justice system
Analyses included summary estimates and meta-regression with random involvement and to prevent childhood abuse and mitigate its adverse
effects. effects.
Main Results. The search identified 1429 records. We included 34 unique Systematic Review Registration. PROSPERO CRD42017056192. (Am J
studies in our review and 29 nonoverlapping studies in our meta-analysis. Public Health. Published online ahead of print January 24, 2019: e1–e11.
The summary prevalence for any type of childhood abuse was 65.7% (95% doi:10.2105/AJPH.2018.304855)
PLAIN-LANGUAGE SUMMARY identified 34 relevant studies and included 29 with no significant differences between men
A history of childhood abuse may affect studies in our meta-analysis. The summary and women. Prevalence estimates varied
people’s health and criminal justice system prevalence of any child abuse was 65.7% substantially across studies for all types of
involvement. We reviewed data on the among women; only one study reported abuse. Overall, we conclude that half of
prevalence of childhood abuse among people prevalence among men (35.5%). The sum- people in prison in Canada experienced abuse
in prisons in Canada. We searched for studies mary prevalence of sexual abuse was 50.4% in childhood. Given this high prevalence,
published since 1987 reporting the prevalence among women and 21.9% among men, and prisons should incorporate trauma-informed
of any abuse, physical abuse, sexual abuse, the prevalence of neglect was 51.5% among approaches, and work should be done to
emotional abuse, or neglect before the age of women and 42.0% among men. The prev- prevent adverse health effects and criminal
18 years among people in prison in Canada alence of physical abuse was 47.7% and the justice system involvement among people
and summarized data via a meta-analysis. We prevalence of emotional abuse was 51.5%, who experience childhood abuse.
March 2019, Vol 109, No. 3 AJPH Bodkin et al. Peer Reviewed Systematic Review e1
AJPH OPEN-THEMED RESEARCH
e2 Systematic Review Peer Reviewed Bodkin et al. AJPH March 2019, Vol 109, No. 3
AJPH OPEN-THEMED RESEARCH
Data Analysis bias. For the purposes of summarizing data, Egger test with logit and standard error of
Two of the authors (C. B. and F. K.) in- we considered data from multiple publica- logit prevalence to perform regression-based
dependently extracted data using a template tions as a single study if the study data matched statistical tests for publication bias. We gen-
that we developed, piloted, and modified. on key variables. erated contour-enhanced funnel plots25 to
We resolved disagreements regarding In our quantitative synthesis, we included minimize oversensitivity of our bias tests
extracted data through discussions. We data from the study with the larger sample size with respect to binary outcomes and small
extracted data on study characteristics: pub- when studies overlapped in terms of pop- samples.33
lication year, study period, study location, ulation, time frame, and type of abuse re- We analyzed 3 study population charac-
name of facility, type of facility (provincial– ported. In the case of studies that reported teristics that might predict abuse prevalence
territorial or federal), study type, recruitment more than one prevalence estimate for a single and account for between-study heterogene-
strategy, data collection method, and sample type of abuse (e.g., separate reports of physical ity: gender, adult versus youth populations,
size. We also extracted data on participant abuse by the mother and physical abuse by the and federal versus provincial or territorial
characteristics (age, gender, race/ethnicity, father), we used the greater value to indicate facilities. We selected these 3 factors on the
morbidity) and childhood abuse (types of the minimum proportion of participants ex- basis of the published literature,8 theoretical
abuse, definition, prevalence by type of abuse). periencing that type of abuse. considerations, and available data. We assessed
Two authors (C. B. and F. K.) indepen- We used Stata 14 (StataCorp LP, College gender differences for all types of abuse and
dently assessed each included study for Station, TX) to perform our quantitative plotted results by gender. We used mixed-
meta-analysis and produce graphics.24–27 We effects logistic regression models to test
risk of bias in 2 domains: selection and out-
generated forest plots for each type of abuse; whether these factors were significantly asso-
come measurement.19–21 We resolved dis-
estimates are presented as prevalence per- ciated with study prevalence; we also used
agreements regarding bias assessment through
centages with exact binomial confidence random intercept logistic models with an un-
discussions. We classified risk of bias in each
intervals (CIs). structured covariance matrix. We explored
domain as unclear, low, or high, consistent
For summary estimates, we assumed ran- whether apparent associations were unduly
with principles of bias assessment in the
dom effects given the heterogeneity in study influenced by specific studies by examining
Cochrane handbook20 and related guides.22,23
samples, populations, methods, and rigor.28 model diagnostic statistics for influence, in-
With respect to selection, we considered
In our meta-analysis, binary outcome data cluding Cook’s D statistic and delta-B estimates
the risk of bias high if the participant re-
were transformed via the Freeman–Tukey (and outcome estimates if a study was removed),
cruitment or selection method was not sys-
double arcsine transformation (and then back and visually inspecting data using subgroup
tematic, the inclusion and exclusion criteria
transformed to show estimates and confi- sorted forest plots (data not shown). Studies
were not specified, or there was a suggestion
dence intervals as percentages). When sum- with the largest amounts of influence are dis-
of bias in terms of participants in comparison
marizing just 3 study estimates, we used the cussed but were not excluded from our analysis
with the eligible population; we considered metan procedure27 along with a logit trans- according to any specific cut point. Only studies
the risk low if methods of participant selection formation of the proportion and standard with single-gender reporting were included in
were well described (including explicit in- error of the transformed proportion. Trans- our mixed-effects regression modeling.
clusion and exclusion criteria), there was a formations ensured that no studies with a In cases in which gender was significantly
high participation rate, and there was no prevalence nearing 100% were excluded and associated with abuse prevalence, we plotted
evidence of bias in participation; and we that confidence intervals did not fall outside of gender-specific but not overall estimates.
considered the risk unclear if methods for the valid range.29,30 We used the DerSimo- When abuse prevalence did not vary signifi-
defining the population and inclusion and nian and Laird method (assuming random cantly according to gender, we plotted
exclusion criteria were not specified or the effects with inverse-variance weights) to gender-specific and overall estimates, ensuring
participation rate was not specified. conduct tests of heterogeneity of estimates that study subjects were counted only once.
In regard to outcome measurement, we between studies.24 We defined the research protocol a
considered the risk of bias high if the in- We also report I2 statistics, interpreted as priori and registered the protocol with
vestigators used a measure of child abuse that approximately the proportion of total esti- PROSPERO.34 We revised the protocol to
was not validated, low if the investigators re- mate variability that is due to inconsistency include a meta-analysis prior to conducting
ported use of a validated measure (e.g., a between studies.31,32 We conducted het- our analyses.
measure validated against a gold standard or erogeneity analyses to explore inconsistency
assessed for content validity), and unclear if rather than to exclude outliers.31 We used the
there was no specification of whether the Stata metaprop_one procedure to generate
measure was validated. We classified the overall forest plots with confidence intervals and RESULTS
risk of bias for each study as low if the risk was heterogeneity statistics.24 Our search identified 1429 records overall
low for both domains, high if the risk was high Potential publication bias was assessed via and 1130 records without duplicates (Figure
in either domain, and unclear otherwise.20 graphical and regression-based statistical tests 1). Of the records without duplicates, 92 were
We qualitatively summarized studies (and only when summary results were based eligible for full review. We were unable to
according to study characteristics and risk of on at least 10 study estimates). We used the retrieve 2 articles.35,36 Of the 90 full articles
March 2019, Vol 109, No. 3 AJPH Bodkin et al. Peer Reviewed Systematic Review e3
AJPH OPEN-THEMED RESEARCH
reviewed, 39 were eligible for inclusion,37–75 women; 17 studies reported data only on men. organic brain disorders. All of the included
representing 34 unique studies. We included Half of the studies were conducted in British studies were cross sectional. Data were col-
29 nonoverlapping samples in our quantita- Columbia, Ontario, or Quebec (n = 17), with lected via interviews (n = 17), administrative
tive synthesis. another 8 involving individuals in federal file reviews (n = 17), and questionnaires
Table 1 summarizes the characteristics of custody in facilities across Canada. In 6 studies, (n = 8). Studies differed in terms of definitions
the 34 included studies. Most studies examined there was no specification of study location. of age of abuse, with some specifying only
the prevalence of childhood sexual abuse Three studies were conducted in the Prairies, abuse during childhood and others specifying
(n = 26) or childhood physical abuse (n = 23). Territories, or Maritimes. ages of abuse ranging from before 10 years to
Fewer studies assessed the prevalence of any Most studies included the general pop- before 18 years.
abuse (n = 7), emotional abuse (n = 12), and ulation of people in prison; however, several Regarding overall risk of bias, we found a
neglect (n = 8). Most studies focused on adults included only individuals convicted of sex low risk in only one study. Risks were unclear in
(n = 27) and individuals in federal facilities offenses (n = 6), and some excluded in- 5 studies (15%) and high in 29 studies (85%;
(n = 22). Only 5 studies provided data on the dividuals on the basis of language, literacy, Table A, available as a supplement to the online
prevalence of childhood abuse solely among active psychiatric illness, or presence of version of this article at http://www.ajph.org).
Studies included in
quantitative synthesis
(meta-analysis)
(n = 29)
FIGURE 1—Flow Diagram of Study Selection for Our Systematic Review and Meta-Analysis of Childhood Abuse Among People in Canadian
Prisons
e4 Systematic Review Peer Reviewed Bodkin et al. AJPH March 2019, Vol 109, No. 3
AJPH OPEN-THEMED RESEARCH
TABLE 1—Studies Included in Systematic Review of Prevalence of Exposure to Childhood Abuse Among People in Canadian Prisons (n = 34)
For each type of abuse, overall and 3; Figures A–C, available as supplements to 80.9%. In terms of any childhood abuse, the
gender-specific estimates showed very high the online version of this article at http:// summary prevalence among women was
between-study heterogeneity (Figures 2 and www.ajph.org), with I2 values all above 65.7% (range = 56.2% to 75.0%; Figure A).
March 2019, Vol 109, No. 3 AJPH Bodkin et al. Peer Reviewed Systematic Review e5
AJPH OPEN-THEMED RESEARCH
Only one study reported childhood abuse prevalence of abuse reported among female estimates are imprecise and the ranges of es-
among men, with a prevalence of 35.5%. The than male populations. However, the timates are broad, it is clear that people in
difference according to gender was apparent Wanklyn et al.43 and Fraser and Roesch48 prison commonly report many forms of
and statistically significant. studies had an unacceptably high influence on abuse. Our summary estimates indicate that
The summary prevalence estimate for the results. After exclusion of the Wanklyn approximately half of people in Canadian
physical abuse (Figure 2) was 47.7% (range = et al. study, there was no significant difference prisons have experienced at least one type of
16.3% to 83.0%), and study heterogeneity was between men and women. On the basis of childhood abuse. Women were more likely
very high (I2 = 94.6% overall; P < .01). There these analyses and a visual inspection of the than men to have experienced sexual abuse
was no strong evidence that gender, provincial data, we did not find sufficient evidence of a (50.4% vs 21.9%) and neglect (51.5% vs
versus federal facility, or youth versus adult difference between men and women. We 42.0%). We did not find sufficient evidence of
population explained the heterogeneity in therefore considered all studies, yielding a a difference between men and women in the
between-study estimates. Bias analyses and summary estimate of 51.5% (range = 8.7% prevalence of physical abuse (47.7% overall)
contour-enhanced funnel plots (data not shown) to 96.0%; Figure B) that was highly or emotional abuse (51.5%).
indicated very high heterogeneity, but there was heterogeneous. Similar to national studies of prison pop-
not strong evidence of bias (P = .06). In exploratory analyses controlling for ulations in the United States and the general
The estimated mean prevalence of a his- gender, we found a significant association population in Canada, we found high abso-
tory of sexual abuse was 50.4% (range = between emotional abuse and a pair of group lute proportions of all types of childhood
9.9% to 77.3%) among women and 21.9% categories (adults vs youths and individuals in abuse. Across types of abuse, our mean
(range = 8.3% to 55.6%) among men (Figure provincial–territorial vs federal facilities), with prevalence estimates were substantially higher
3). Heterogeneity was high (above 94% for a higher prevalence for youths and for those in than the estimates revealed in the 2014 Ca-
both genders; P < .01). In mixed-model re- provincial–territorial facilities (Figures F–G, nadian General Social Survey, with a 4-fold
gression analyses incorporating only gender- available as supplements to the online version higher prevalence of sexual abuse among men
specific estimates, the odds ratio was 4.5 (95% of this article at http://www.ajph.org). As and women; that 2014 survey showed that
CI = 3.0, 6.9) in comparisons of studies in- with the sexual abuse model, we could not 35% of men and 31% of women had expe-
volving women with those involving men. examine the effects of facility type and age rienced any form of child maltreatment, 31%
Two of the studies included in the regression group independent of each other because of men and 22% of women had experienced
model, Marshall et al.57 and Swihart,72 had a they were collinear. childhood physical abuse, and 4% of men and
large influence on the results (Figure 3), and The mean prevalence of neglect was 51.5% 12% of women had experienced childhood
exclusion of these studies resulted in a stronger (range = 45.5% to 65.1%) among women and sexual abuse.76
gender difference; we retained all studies in the 42.0% (range = 6.8% to 99.0%) among men Our prevalence estimates were also higher
final forest plots and gender-specific summary (Figure C). Estimate heterogeneity was ex- than those found in national surveys con-
estimates presented. Egger tests for publication tremely high (I2 = 99%) within the male es- ducted in US correctional facilities between
bias were nonsignificant for each gender and timates; heterogeneity could not estimated 1995 and 1997, which revealed rates of any
for cases in which the genders were combined. for female populations because there were form of childhood abuse of 14.4% among
Exploratory analyses suggested that type of only 2 studies. Using mixed-model regression men and 36.7% among women in state
facility predicted the prevalence of a reported and including only gender-specific studies, prisons, 5.8% among men and 23.0% among
history of sexual abuse; the prevalence was we found an odds ratio of 1.8 (95% CI = 1.2, women in federal prisons, and 11.9% among
higher among individuals in federal facilities, 2.8) among women (relative to men). Al- men and 36.6% among women in jails.77
even after control for gender. When the though the Fraser and Roesch study48 had an Several other US studies have also identified a
analysis was restricted to data on men (Figure unacceptably high influence, we did not ex- high prevalence of childhood abuse among
D, available as a supplement to the online clude that study; its exclusion would not have people in jails and prisons, with varying results
version of this article at http://www.ajph. had a large impact on the odds ratio (which that may be due to differences in sampling and
org), the results showed that those in federal would have been 1.9). There were too few measurement.4,78–82
facilities were more likely than those in studies of gender-specific neglect to examine
provincial facilities to have a history of sexual publication bias statistics. There was no strong
abuse. As federal facilities are also adult fa- evidence of a difference between adult and Limitations
cilities (Figure E, available as a supplement to youth populations or between individuals in There are several potential limitations to
the online version of this article at http:// federal and provincial–territorial facilities. the included studies. The high risk of bias in
www.ajph.org), the effect of the 2 charac- most studies in the domains of selection and
teristics cannot be separated. We could not outcome measurement may have affected
examine the impact of facility type among study validity. Selection of a sample that is
women because there was only one estimate DISCUSSION neither random nor representative of the total
available (from a youth facility study). Our meta-analysis revealed a very high population could lead to a higher or lower
For emotional abuse, there was a statisti- prevalence of childhood abuse among people prevalence than the true prevalence for the
cally significant gender effect, with a higher in prison in Canada. Although summary population. Outcome measurement issues
e6 Systematic Review Peer Reviewed Bodkin et al. AJPH March 2019, Vol 109, No. 3
AJPH OPEN-THEMED RESEARCH
Female
Allenby et al.37 Adult, Federal 39.3 (26.5, 53.2)
Colantonio et al.41 Adult, Provincial 53.5 (43.2, 63.6)
Heney50 Adult, Federal 72.7 (57.2, 85.0)
Heney and Kristiansen51 Adult, Federal 45.2 (27.3, 64.0)
Robeson Barrett et al.66 Adult, Federal 54.3 (46.0, 62.4)
Stewart and Wormith70 Adult, Federal 46.5 (36.5, 56.7)
Swihart73 Adult, Both 27.9 (19.8, 37.2)
Wanklyn et al.43 Youth, Provincial 55.8 (39.9, 70.9)
Subtotal (I 2 = 80.9%; P < .001) 49.0 (39.7, 58.3)
Male
Beauregard et al.38 Adult, Unspecified 68.8 (57.3, 78.9)
Chubaty39,40 Adult, Federal 61.5 (50.8, 71.6)
Colantonio et al.41 Adult, Provincial 36.0 (27.6, 45.1)
Dutton and Hart45,46 Adult, Federal 31.1 (27.5, 35.0)
Forth and Tobin47 Youth, Provincial 35.8 (26.2, 46.3)
Fraser and Roesch48 Youth, Provincial 83.0 (74.2, 89.8)
Johnston53 Adult, Federal 50.0 (37.2, 62.8)
Milcent and Granger60 Adult, Federal 73.9 (51.6, 89.8)
Mowat-Leger62 Adult, Provincial 44.8 (36.8, 53.0)
Reckdenwald et al.64 Adult, Federal 49.8 (45.7, 54.0)
Robinson and Taylor67 Adult, Federal 34.7 (31.6, 37.8)
Smale68 Youth, Provincial 50.0 (21.1, 78.9)
Vitelli75 Adult, Provincial 36.4 (27.8, 45.8)
Wanklyn et al.43 Youth, Provincial 52.2 (39.7, 64.6)
Subtotal (I 2 = 93.5%; P < .001) 49.8 (42.0, 57.7)
Reported combined
Johnston54 Adult, Federal 45.2 (40.8, 49.6)
Joubert et al.55 Adult, Provincial 16.3 (13.2, 19.7)
Smith and Corrado69 Youth, Provincial 60.4 (54.4, 66.2)
Ulzen and Hamilton74 Youth, Provincial 30.6 (18.3, 45.4)
Subtotal (I 2 = 98.5%; P < .001) 37.4 (17.3, 60.2)
0 10 20 30 40 50 60 70 80 90 100
Note. CI = confidence interval; ES = estimate. Summary estimates for prevalence estimates are presented for information purposes, despite statistically significant
heterogeneity statistics (see the Discussion section for rationale).
FIGURE 2—Forest Plot From Meta-Analysis of Studies Reporting Prevalence of a History of Childhood Physical Abuse Among People in
Canadian Prisons, by Gender
such as social desirability and forgetting could sensitivity.85 Validated tools (the Childhood Note that the only estimate of the preva-
lead to underestimation of childhood abuse, Trauma Questionnaire and the Clarke lence of any childhood abuse among
whereas the narrowness or breadth of the Parent-Child Relations Questionnaire) were men72 that we identified was significantly
definition of childhood abuse could lead to used in the 2 studies that were at low risk of lower than the prevalence of physical
underestimation or overestimation.83,84 outcome measurement bias,42,43,60 but these abuse38,40,43,48,53,60,62,64 or emotional
Specific challenges include asking only studies involved relatively small sample sizes. abuse38,47,48,62,64 among men in several
about a single type of abuse, collecting data Using multiple sources of data may also im- studies, indicating that the prevalence of any
only on part of childhood or on abuse per- prove sensitivity; for example, Joubert et al. child abuse revealed in that study may be
petrated by certain individuals (e.g., parents), used institutional files, medical charts, and lower than the true prevalence.
classifying abuse as a binary variable, and interviews to examine childhood abuse There are also limitations to the review
relying on administrative data or tools that among 522 individuals,55 although such a itself. We included only studies published
have not been validated and may lack strategy may be relatively resource intensive. in the past 30 years because we wanted to
March 2019, Vol 109, No. 3 AJPH Bodkin et al. Peer Reviewed Systematic Review e7
AJPH OPEN-THEMED RESEARCH
Female
Allenby et al.37 Adult, Federal 57.1 (44.1, 69.2)
Colantonio et al.41 Adult, Provincial 45.5 (36.0, 55.2)
Heney50 Adult, Federal 77.3 (63.0, 87.2)
Heney and Kristiansen51 Adult, Federal 58.1 (40.8, 73.6)
Robeson Barrett et al.66 Adult, Federal 66.4 (57.4, 74.3)
Stewart and Wormith70 Adult, Federal 49.5 (40.0, 59.1)
Swihart73 Adult, Both 9.9 (5.6, 16.9)
Wanklyn et al.43 Youth, Provincial 46.5 (32.5, 61.1)
Subtotal (I 2 = 94.2%; P < .001) 50.4 (33.5, 67.2)
Male
Beauregard et al.38 Adult, Unspecified 18.2 (11.2, 28.2)
Chubaty39,40 Adult, Federal 16.5 (10.3, 25.4)
Colantonio et al.41 Adult, Provincial 12.8 (8.0, 19.8)
Dhawan and Marshall44 Adult, Federal 46.2 (34.6, 58.1)
Dutton and Hart45,46 Adult, Federal 11.4 (9.1, 14.2)
Forth and Tobin47 Youth, Provincial 16.8 (10.6, 25.6)
Fraser and Roesch48 Youth, Provincial 11.0 (6.3, 18.6)
Johnston53 Adult, Federal 21.9 (13.5, 33.4)
Marshall et al.57 Adult, Federal 55.6 (42.4, 68.0)
Marshall and Marshall58 Adult, Federal 35.0 (22.1, 50.5)
Mowat-Leger62 Adult, Provincial 16.2 (11.2, 22.9)
Nunes et al.63 Adult, Federal 31.2 (25.8, 37.2)
Reckdenwald et al.64 Adult, Federal 41.3 (37.4, 45.4)
Robinson and Taylor67 Adult, Federal 12.0 (10.1, 14.2)
Smale68 Youth, Provincial 8.3 (1.5, 35.4)
Vitelli75 Adult, Provincial 16.1 (10.6, 23.8)
Wanklyn et al.43 Youth, Provincial 19.4 (11.7, 30.4)
Subtotal (I 2 = 94.6%; P < .001) 21.9 (15.7, 28.8)
Reported combined
Johnston54 Adult, Federal 21.2 (17.9, 25.0)
Joubert et al.55 Adult, Provincial 10.9 (8.5, 13.9)
Smith and Corrado69 Youth, Provincial 20.5 (16.2, 25.6)
Ulzen and Hamilton74 Youth, Provincial 10.2 (4.4, 21.8)
Subtotal (I 2 = 88.3%; P < .001) 15.9 (10.1, 22.7)
0 10 20 30 40 50 60 70 80 90 100
Prevalence of Childhood Sexual Abuse, %
Note. CI = confidence interval; ES = estimate. Summary estimates for prevalence estimates are presented for information purposes, despite statistically significant
heterogeneity statistics (see the Discussion section for rationale). Overall summary estimates are not shown owing to apparent and statistically significant differences
according to gender.
FIGURE 3—Forest Plot From Meta-Analysis of Studies Reporting Prevalence of a History of Childhood Sexual Abuse Among People in
Canadian Prisons, by Gender
include relatively recent data and ensure that For consistency with the World Health inherently abusive and traumatic nature
our prevalence estimates would remain valid. Organization’s definition of childhood abuse, of residential schools89 and the over-
Evidence is mixed regarding whether the we did not include witnessing abuse in our representation of Indigenous peoples in Ca-
prevalence of childhood abuse has changed review, despite its inclusion in many defini- nadian correctional facilities,90 this likely
over time.84,86,87 If the prevalence has not tions and measures of childhood abuse and its contributed to underestimation of child abuse
changed, limiting the search may have un- demonstrated effects on children.88 We also prevalence and may limit the direct applica-
necessarily affected the summary estimates’ did not include attending residential school as bility of our findings in terms of informing
values and precision. We categorized child- a specific type of child abuse; however, abuse interventions implemented in the Canadian
hood abuse as a binary variable, although the in residential schools may have been captured correctional system.
experience and impact of child abuse vary in our prevalence estimates because we did We used only English terms in our search,
significantly according to factors such as se- not specify any limits regarding the setting of and we identified only one French-language
verity, duration, and timing of exposure.8 abuse or the perpetrator of abuse. Given the study in our review.60 We believe it is
e8 Systematic Review Peer Reviewed Bodkin et al. AJPH March 2019, Vol 109, No. 3
AJPH OPEN-THEMED RESEARCH
unlikely that we missed any relevant violence, and the state’s obligation to protect full texts, and C. Bodkin and F. Kouyoumdjian reviewed
full texts. S. J. Bondy conducted the meta-analysis.
French-language studies, however, because children from abuse,94 our findings support 3 C. Bodkin, L. Pivnick, S. J. Bondy, R. Elwood Martin,
we reviewed reference lists of studies from the clear imperatives. First, policymakers and C. Jernigan, and F. Kouyoumdjian contributed to in-
province of Quebec and contacted experts. researchers should advance research to elu- terpretation of data. C. Bodkin, L. Pivnick, S. J. Bondy,
and F. Kouyoumdjian drafted the article, and all of the
Finally, there was substantial heterogeneity cidate the severity and nature of exposure to authors contributed to revisions of the article.
between studies in the prevalence of all types childhood abuse and the mechanisms linking
of childhood abuse, which likely reflects a such abuse to imprisonment among people in CONFLICTS OF INTEREST
combination of diversity in included pop- prisons in Canada; attention should be paid to The authors have no conflicts of interest to report.
ulations, methodological issues such as bias, populations with a particularly high preva- HUMAN PARTICIPANT PROTECTION
and statistical issues such as lack of pre- lence of childhood abuse, including women, No protocol approval was needed for this study because
cision.20,91,92 We chose to present and ex- youths, and Indigenous peoples. Given that only publicly available data were used.
plore heterogeneity substantively.31 We used most people who experience abuse in
REFERENCES
random effects in analyses, examined sub- childhood do not go on to perpetuate abuse 1. Malakieh J. Youth correctional statistics in Canada,
groups that we defined a priori, and consid- or experience imprisonment,7,9 research 2015/2016. Available at: http://www.statcan.gc.ca/pub/
ered the distribution of results within strata 85-002-x/2017001/article/14702-eng.htm. Accessed
should also explore protective factors that December 31, 2018.
and the effects of outliers.20,91,93 Our data mitigate criminal justice system involvement.
2. Reitano J. Adult correctional statistics in Canada, 2015/
plots show which individual study estimates Second, correctional authorities should 2016. Available at: http://www.statcan.gc.ca/pub/85-
fall outside the confidence intervals for promote trauma-informed services that seek 002-x/2017001/article/14700-eng.htm. Accessed De-
summary estimates, highlighting evidence of cember 31, 2018.
to create a safe, transparent, and empowering
potential clinical heterogeneity. Note that the environment while avoiding retraumatiza- 3. Kouyoumdjian F, Schuler A, Matheson FI, Hwang
SW. Health status of prisoners in Canada: narrative re-
summary prevalence was close to or greater tion.95 Imprisonment may serve as a unique view. Can Fam Physician. 2016;62(3):215–222.
than 50% for each of the following categories: opportunity to improve health and interrupt 4. Wolff N, Shi J. Childhood and adult trauma experi-
any childhood abuse among women, sexual the cycle of abuse and transmission of violence ences of incarcerated persons and their relationship to
abuse among women, neglect among to future generations. Such an approach could adult behavioral health problems and treatment. Int J
Environ Res Public Health. 2012;9(5):1908–1926.
women, physical abuse among both men and address the consequences of trauma, facili-
5. Baglivio MT, Epps N, Swartz K, Huq MS, Sheer A,
women, and emotional abuse among both tate healing, and prevent retraumatization Hardt NS. The prevalence of adverse childhood expe-
men and women. These findings increase our through criminal justice system involvement. riences (ACE) in the lives of juvenile offenders. J Juv Justice.
confidence that at least half of people who Third, all sectors should recognize childhood 2014;3(2):1.
experience imprisonment have experienced as a pivotal developmental stage and redouble 6. Fox BH, Perez N, Cass E, Baglivio MT, Epps N.
some form of childhood abuse. Trauma changes everything: examining the relationship
efforts to support healthy families and prevent between adverse childhood experiences and serious, vi-
We have presented heterogeneity statistics child abuse, including as a strategy to po- olent and chronic juvenile offenders. Child Abuse Negl.
and information about studies with high in- tentially prevent criminal behavior and 2015;46:163–173.
fluence as exploratory and potentially in- imprisonment. 7. English DJ, Widom CS, Brandford C. Childhood
formative given the limited and emerging victimization and delinquency, adult criminality, and
Future work needs to involve people with violent criminal behavior: A replication and extension.
nature of evidence in this area to date. lived experiences of childhood abuse and Available at: https://www.ncjrs.gov/pdffiles1/nij/
However, given the wide confidence in- imprisonment in Canada, policymakers, grants/192291.pdf. Accessed December 31, 2018.
tervals for all summary estimates and the ac- health care and social service leaders, pop- 8. Gilbert R, Widom CS, Browne K, Fergusson D,
knowledged unexplained variability, we Webb E, Janson S. Burden and consequences of child
ulations overrepresented in corrections fa- maltreatment in high-income countries. Lancet.
believe that it would be inappropriate to cilities, and researchers in further defining the 2009;373(9657):68–81.
directly apply these estimates (e.g., to calcu- issues and planning and implementing ef- 9. Fitton L, Yu R, Fazel S. Childhood maltreatment and
late service requirements or to conduct cost fective strategies. These efforts should include violent outcomes: a systematic review and meta-analysis
analyses). Further research is needed to de- of prospective studies. Trauma Violence Abuse. 2018
qualitative methods exploring pathways from [Epub ahead of print].
termine whether the studies with the highest childhood abuse to imprisonment and iden- 10. Widom CS, Wilson HW. Intergenerational Transmission
or lowest prevalence rates are indicative of tifying opportunities for intervention, as well of Violence. Dordrecht, the Netherlands: Springer Neth-
true case-mix differences across facilities, as quantitative research incorporating sensi- erlands; 2014.
which would require expanded surveillance tive and acceptable tools to better characterize 11. Afifi TO, MacMillan HL, Boyle M, Taillieu T,
with representative sampling and consistent this public health problem. We also recom-
Cheung K, Sareen J. Child abuse and mental disorders in
Canada. CMAJ. 2014;186(9):E324–E332.
data collection methods. mend reviews of international data, which 12. Dube SR, Felitti VJ, Dong M, Chapman DP, Giles
could address some of the limitations of our WH, Anda RF. Childhood abuse, neglect, and household
review. dysfunction and the risk of illicit drug use: the Adverse
Conclusions Childhood Experiences Study. Pediatrics. 2003;111(3):
Given the state’s responsibility to provide CONTRIBUTORS 564–572.
health care and social services for people in C. Bodkin, L. Pivnick, S. J. Bondy, C. Ziegler, R. Elwood 13. Friestad C, Ase-Bente R, Kjelsberg E. Adverse
Martin, and F. Kouyoumdjian contributed to study childhood experiences among women prisoners: re-
custody, the cyclical nature of childhood conceptualization and design. C. Bodkin, L. Pivnick, and lationships to suicide attempts and drug abuse. Int J Soc
abuse and intergenerational transmission of F. Kouyoumdjian screened abstracts and retrieved Psychiatry. 2014;60(1):40–46.
March 2019, Vol 109, No. 3 AJPH Bodkin et al. Peer Reviewed Systematic Review e9
AJPH OPEN-THEMED RESEARCH
14. Government of Canada. Criminal code. Available at: 33. Sterne JA, Gavaghan D, Egger M. Publication and 51. Heney J, Kristiansen C. Dying on the Inside: Suicide and
https://laws-lois.justice.gc.ca/eng/acts/C-46. Accessed related bias in meta-analysis: power of statistical tests and Suicidal Feelings Among Federally Incarcerated Women.
December 31, 2018. prevalence in the literature. J Clin Epidemiol. 2000;53(11): Ottawa, Ontario, Canada: Carleton University; 1996.
1119–1129. 52. Heney J. Dying on the inside: suicide and suicidal
15. Government of Canada. Controlled Drugs and
Substances Act. Available at: https://laws-lois.justice.gc. 34. Pivnick L, Kouyoumdjian F, Ziegler C, Elwood feelings among federally incarcerated women. Diss Abstr
ca/eng/acts/C-38.8. Accessed December 31, 2018. Martin R, Bodkin C. A systematic review of the prev- Int B Sci Eng. 1997;57(10-B):6574.
alence of child abuse in people who experience in- 53. Johnston JC. Northern Aboriginal offenders in federal
16. Patterson GR, DeBaryshe B, Ramsey E. A de-
carceration in Canada. Available at: https://www. custody: a profile. Available at: http://www.csc-scc.gc.
velopmental perspective on antisocial behavior. Am
crd.york.ac.uk/PROSPERO/display_record.asp? ca/research/r36e-eng.shtml. Accessed December 31,
Psychol. 19890;44(2):329–335. ID=CRD42017056192. Accessed December 31, 2018. 2018.
17. Messina N, Grella C. Childhood trauma and women’s
35. Karim N. Post-Traumatic Stress Disorder and the Psy- 54. Johnston JC. Aboriginal Offender Survey: case files
health outcomes in a California prison population. Am J chological Effects of Long Term Imprisonment: A Cross-Sectional
Public Health. 2006;96(10):1842–1848. and interview sample. Available at: http://www.csc-scc.
Sample of Long Term Offenders in the Pacific Region of Canada. gc.ca/research/092/r061-er61-eng.pdf. Accessed De-
18. World Health Organization. Child maltreatment. Cambridge, England: University of Cambridge Press; cember 31, 2018.
Available at: http://www.who.int/mediacentre/ 2001.
factsheets/fs150/en. Accessed December 31, 2018. 55. Joubert D, Archambault K, Brown G. Cycle of co-
36. Loucks AD. Criminal Behavior, Violent Behavior, and ercion: experiences of maltreatment and disciplinary
19. Sanderson S, Tatt ID, Higgins JP. Tools for assessing Prison Maladjustment in Federal Female Offenders. Kingston, measures in Canadian inmates. Int J Prison Health. 2014;
quality and susceptibility to bias in observational studies in Ontario, Canada: Queen’s University; 1996. 10(2):79–93.
epidemiology: a systematic review and annotated bibli- 37. Allenby K, Taylor K, Cossette M, Fortin D. A Profile of 56. Keown L, Gobeil R, Biro SM, Ritchie MB. Ethno-
ography. Int J Epidemiol. 2007;36(3):666–676. Women Who Sexually Offend. Ottawa, Ontario, Canada: cultural Offenders: An Initial Investigation of Social History
20. Higgins JPT, Green S. Cochrane handbook for sys- Correctional Service of Canada; 2012. Variables at Intake. Ottawa, Ontario, Canada: Correctional
tematic reviews of interventions. Available at: http:// 38. Beauregard E, Stone MR, Proulx J, Michaud P. Service of Canada; 2015.
handbook.cochrane.org. Accessed December 31, 2018. Sexual murderers of children: developmental, precrime, 57. Marshall WL, Serran GA, Cortoni FA. Childhood
21. National Institute for Health and Care Excellence. crime, and postcrime factors. Int J Offender Ther Comp attachments, sexual abuse, and their relationship to adult
Methods for the development of NICE public health Criminol. 2008;52(3):253–269. coping in child molesters. Sex Abuse. 2000;12(1):17–26.
guidance. Available at: https://www.nice.org.uk/ 39. Chubaty DE. Victimization, fear, and coping in 58. Marshall L, Marshall W. Sexual addiction in in-
process/pmg4/chapter/appendix-g-quality-appraisal- prison. Diss Abstr Int B Sci Eng. 2001;62(2-B):1071. carcerated sexual offenders. Sex Addict Compulsivity. 2006;
checklist-quantitative-studies-reporting-correlations- 40. Chubaty DE. Victimization, fear, and coping in 13(4):377–390.
and. Accessed December 31, 2018. prison. Forum Correct Res. 2002;14(1):13–15. 59. Martin MS, Dorken SK, Colman I, McKenzie K,
22. Munn Z, Moola S, Lisy K, Riitano D, Tufanaru C. 41. Colantonio A, Kim H, Allen S, Asbridge M, Petgrave Simpson AI. The incidence and prediction of self-injury
Methodological guidance for systematic reviews of ob- J, Brochu S. Traumatic brain injury and early life expe- among sentenced prisoners. Can J Psychiatry. 2014;59(5):
servational epidemiological studies reporting prevalence riences among men and women in a prison population. 259–267.
and cumulative incidence data. Int J Evid-Based Healthc. J Correct Health Care. 2014;20(4):271–279. 60. Milcent M-P, Granger L. Etude du Rapport à la Paternité
2015;13(3):147–153.
42. Day DM, Hart TA, Wanklyn SG, McCay E, Mac- et des Perceptions des Figures Parentales des Pères Incestueux.
23. Munn Z, Moola S, Lisy K, Riitano D, Tufanaru C. pherson A, Burnier N. Potential mediators between child Montreal, Quebec, Canada: University of Montreal;
Systematic reviews of prevalence and incidence. Available abuse and both violence and victimization in juvenile 2001.
at: https://reviewersmanual.joannabriggs.org. Accessed offenders. Psychol Serv. 2013;10(1):1–11. 61. Motiuk LL. The Validity of Offender Needs Identification
December 31, 2018. 43. Wanklyn SG, Day DM, Hart TA, Girard TA. Cu- and Analysis in Community Corrections. Ottawa, Ontario,
24. Nyaga VN, Arbyn M, Aerts M. Metaprop: a Stata mulative childhood maltreatment and depression among Canada: Correctional Service of Canada; 1994.
command to perform meta-analysis of binomial data. Arch incarcerated youth: impulsivity and hopelessness as po- 62. Mowat-Leger V. Risk Factors for Violence: A Comparison
Public Health. 2014;72(1):39. tential intervening variables. Child Maltreat. 2012;17(4): of Domestic Batterers and Other Violent and Non-Violent
306–317. Offenders. Ottawa, Ontario, Canada: Carleton University;
25. Peters JL, Sutton AJ, Jones DR, Abrams KR, Rushton
L. Contour-enhanced meta-analysis funnel plots help 44. Dhawan S, Marshall W. Sexual abuse histories of 2001.
distinguish publication bias from other causes of asym- sexual offenders. Sex Abuse. 1996;8(1):7–15. 63. Nunes KL, Hermann CA, Renee Malcom J, Lavoie K.
metry. J Clin Epidemiol. 2008;61(10):991–996. 45. Dutton DG, Hart SD. Risk markers for family vio- Childhood sexual victimization, pedophilic interest, and
26. Palmer TM, Sterne JAC, eds. Meta-Analysis in Stata: lence in a federally incarcerated population. Int J Law sexual recidivism. Child Abuse Negl. 2013;37(9):703–711.
An Updated Collection from the Stata Journal. 2nd ed. College Psychiatry. 1992;15(1):101–112. 64. Reckdenwald A, Mancini C, Beauregard E. The cycle
Station, TX: Stata Press; 2016. 46. Dutton DG, Hart SD. Evidence for long-term, of violence: examining the impact of maltreatment early
27. Harris R, Bradburn M, Deeks J, Harbord R, Altman specific effects of childhood abuse and neglect on criminal in life on adult offending. Violence Vict. 2013;28(3):
D, Sterne J. Metan: fixed- and random-effects meta- behavior in men. Int J Offender Ther Comp Criminol. 1992; 466–482.
analysis. Stata J. 2008;8(1):3–28. 36(2):129–137. 65. Reckdenwald A, Mancini C, Beauregard E. Ado-
28. Higgins JPT, Thompson SG, Spiegelhalter DJ. A 47. Forth A, Tobin F. Psychopathy and young offenders: lescent self-image as a mediator between childhood
re-evaluation of random-effects meta-analysis. J R Stat Soc rates of childhood maltreatment. Forum Correct Res. 1995; maltreatment and adult sexual offending. J Crim Justice.
Ser A Stat Soc. 2009;172(1):137–159. 7(1):20–22. 2014;42(2):85–94.
29. Freeman MF, Tukey JW. Transformations related to the 48. Fraser S, Roesch R. Patterns of Substance Use in Ado- 66. Robeson Barrett M, Allenby K, Taylor K. Twenty
angular and the square root. Ann Math Stat. 1950;21(4): lescent Male Young Offenders: Relationships With Child years later: revisiting the Task Force on Federally Sentenced
Maltreatment Experiences and Their Inculcation of Antisocial Women. Available at: http://www.csc-scc.gc.ca/research/
607–611.
Identities. Vancouver, British Columbia, Canada: Simon 005008-0222-01-eng.shtml. Accessed December 31, 2018.
30. Tukey JW. On the comparative anatomy of trans- Fraser University; 2001. 67. Robinson D, Taylor J-A. The incidence of family
formations. Ann Math Stat. 1957;28(3):602–632.
49. Fraser S. Patterns of substance use in adolescent male violence perpetrated by federal offenders: a file review
31. Higgins JP. Heterogeneity in meta-analysis should be young offenders: relationships with child maltreatment study. Available at: http://www.csc-scc.gc.ca/
expected and appropriately quantified. Int J Epidemiol. experiences and their inculcation of antisocial identities. publications/fv/fv03/toce-eng.shtml. Accessed Decem-
2008;37(5):1158–1160. Diss Abstr Int B Sci Eng. 2002;62(9-B):4205. ber 31, 2018.
32. Higgins JPT, Thompson SG, Deeks JJ, Altman DG. 50. Heney J. Report on Self-Injurious Behaviour in the 68. Smale WT. Understanding the Issue of Dropouts: A Young
Measuring inconsistency in meta-analyses. BMJ. 2003; Kingston Prison for Women. Ann Arbor, MI: ProQuest Offender Perspective. Edmonton, Alberta, Canada: Uni-
327(7414):557–560. Micromedia, Correctional Service Canada; 1990. versity of Alberta; 2001.
e10 Systematic Review Peer Reviewed Bodkin et al. AJPH March 2019, Vol 109, No. 3
AJPH OPEN-THEMED RESEARCH
69. Smith A, Corrado R. Youth Violence and Victimization: 88. Stiles MM. Witnessing domestic violence: the effect
Exploring the Cycle of Violence. Vancouver, British on children. Am Fam Physician 2002;66(11):2052, 2055–
Columbia, Canada: Simon Fraser University; 2010. 2056.
70. Stewart CA, Wormith JS. Risk Assessment of Federal 89. Truth and Reconciliation Commission of Canada.
Female Offenders. Saskatoon, Saskatchewan, Canada: Honouring the truth, reconciling for the future: summary
University of Saskatchewan; 2011. of the final report of the Truth and Reconciliation
Commission of Canada. Available at: http://nctr.ca/
71. Stewart L, Nolan A, Thompson J, Sapers J. Social
assets/reports/Final Reports/Executive_Summary_
Determinants of Physical Health Conditions Among Incoming English_Web.pdf. Accessed December 31, 2017.
Canadian Federal Inmates. Ottawa, Ontario, Canada:
Correctional Service of Canada; 2015. 90. Owusu-Bempah A, Kanters S, Druyts E, et al. Years of
life lost to incarceration: inequities between Aboriginal
72. Stewart LA, Wardrop K, Wilton G, Thompson J, and non-Aboriginal Canadians. BMC Public Health. 2014;
Derkzen D, Motiuk L. Reliability and Validity of the Dy- 14:585.
namic Factors Identification and Analysis. Ottawa, Ontario,
91. Petitti DB. Meta-Analysis, Decision Analysis, and
Canada: Correctional Service of Canada; 2017.
Cost-Effectiveness Analysis. New York, NY: Oxford
73. Swihart G. Female Offenders: Attachment & Parenthood. University Press; 1999.
Vancouver, British Columbia, Canada: University of
92. Petitti DB. Meta-Analysis, Decision Analysis, and
British Columbia; 2002.
Cost-Effectiveness Analysis: Methods for Quantitative Synthesis
74. Ulzen TP, Hamilton H. The nature and characteristics in Medicine. 2nd ed. New York, NY: Oxford University
of psychiatric comorbidity in incarcerated adolescents. Press; 2000.
Can J Psychiatry. 1998;43(1):57–63. 93. Ioannidis JP, Patsopoulos NA, Rothstein HR.
75. Vitelli R. Comparison of early and late start models of Reasons or excuses for avoiding meta-analysis in forest
delinquency in adult offenders. Int J Offender Ther Comp plots. BMJ. 2008;336(7658):1413–1415.
Criminol. 1997;41(4):351–357. 94. United Nations Office of the High Commissioner
76. Statistics Canada. Family violence in Canada: A sta- on Human Rights. Convention on the Rights of
tistical profile. Available at: https://www150.statcan.gc. the Child. Available at: https://www.ohchr.org/en/
ca/n1/pub/85-002-x/2017001/article/14698-eng.htm. professionalinterest/pages/crc.aspx. Accessed December
Accessed December 31, 2018. 31, 2018.
77. Wolf Harlow C. Prior Abuse Reported by Inmates and 95. Substance Abuse and Mental Health Services
Probationers. Washington, DC: Bureau of Justice Statistics; Administration. Trauma-informed approach and
1999. trauma-specific interventions. Available at: https://www.
samhsa.gov/nctic/trauma-interventions. Accessed De-
78. Wolff N, Shi J, Siegel JA. Patterns of victimization cember 31, 2018.
among male and female inmates: evidence of an enduring
legacy. Violence Vict. 2009;24(4):469–484.
79. Weeks R, Widom C. Self-reports of early childhood
victimization among incarcerated male felons. J Interpers
Violence. 1998;13(3):346–361.
80. Lewis DO, Shanok SS, Pincus JH, Glaser GH. Violent
juvenile delinquents: psychiatric, neurological, psycho-
logical, and abuse factors. J Am Acad Child Psychiatry. 1979;
18(2):307–319.
81. Bloom B, Lind MC, Owen B. Women in California
Prisons: Hidden Victims of the War on Drugs. San Francisco,
CA: Center on Juvenile and Criminal Justice; 1994.
82. McClellan DS, Farabee D, Crouch BM. Early vic-
timization, drug use, and criminality. Crim Justice Behav.
1997;24(4):455–476.
83. Goldman JDG, Padayachi UK. Some methodological
problems in estimating incidence and prevalence in child
sexual abuse research. J Sex Res. 2000;37(4):305–314.
84. Laaksonen T, Sariola H, Johansson A, et al. Changes in
the prevalence of child sexual abuse, its risk factors, and
their associations as a function of age cohort in a Finnish
population sample. Child Abuse Negl. 2011;35(7):
480–490.
85. Hamby SL, Finkelhor D. Choosing and Using Child
Victimization Questionnaires. Washington, DC: US Gov-
ernment Printing Office; 2001.
86. Feldman W, Feldman E, Goodman JT, et al. Is
childhood sexual abuse really increasing in prevalence? An
analysis of the evidence. Pediatrics. 1991;88(1):29–33.
87. Shields M, Tonmyr L, Hovdestad W. Is child sexual
abuse declining in Canada? Results from nationally
representative retrospective surveys. Health Promot Chronic
Dis Prev Can. 2016;36(11):252–260.
March 2019, Vol 109, No. 3 AJPH Bodkin et al. Peer Reviewed Systematic Review e11