Prevention in Prison: The Diabetes Prevention Program in a Correctional Setting Open Access

This study was set up as a two-group quasi-experimental, staggered-start control design. The second group served as a control group for the first 6 months, at which time they also were enrolled in the intervention. Data were collected at baseline and at 6, 12, and 18 months. The staggered-start design was selected in response to ethical concerns of the institutional review board, who wanted all study participants to have the opportunity to participate in the intervention.

Participants were federal inmates in a female minimum-security institution or a male low-security institution. Both institutions were located in the same geographical location in the northeast United States. Because these were adult institutions, all individuals in the study population were ≥18 years of age.

To enroll in the study, participants had to have a starting BMI ≥24 kg/m² (or ≥22 kg/m² if Asian) and have the ability to ambulate. They also either had to have an A1C of 5.7–6.4%, indicating prediabetes, or have a high risk of developing diabetes according to the CDC screening test (17). Individuals were disqualified from participating if they had an expected release date before study completion, were previously diagnosed with diabetes, or had had a myocardial infarction diagnosed within the past 6 months. In addition, all recruited participants completed a modified Physical Activity Readiness Questionnaire (PAR-Q). Any positive responses on the PAR-Q triggered a review of their medical file by their primary care provider (PCP). If the PCP deemed it was unsafe for the participant to engage in moderate-intensity physical activity, the individual was not allowed to participate.

This study obtained ethical approval through the Federal Bureau of Prisons’ institutional review board. All participants granted informed consent; freedom to withdraw from the program at any time without repercussion was stressed throughout the study.

Participants were recruited using posters placed in the medical and housing units throughout the two institutions. Messages were also sent out electronically through the prison’s Intranet system. Individuals in the prison’s medical department were also asked to inform their patients about the research program. Individuals who expressed interest and met all qualification criteria were randomly assigned to either group 1 (the intervention group) or group 2 (the control group, starting the intervention at month 6). Simple randomization using a random number generator was used to make these assignments.

As shown in Table 1, loss to follow-up at 6 (21%), 12 (38%), and 18 (64%) months mostly occurred as a result of attrition (e.g., transfers and releases) and, to a much lesser extent, because of withdrawal from the program. Loss to follow-up occurred more frequently in group 2.

We used the 2011 edition of the University of Pittsburgh Diabetes Prevention Support Center’s CDC-approved DPP curriculum titled Group Lifestyle Balance (18). Four trained lifestyle coaches (three registered nurses and one medical assistant) facilitated the program. The program was 12 months long, with classes meeting weekly for the first 3 months, biweekly for the next 2 months, and monthly thereafter. Through the curriculum content, participants were encouraged to track and limit their calorie and fat gram consumption to obtain a weight loss of 7–10%. The program also advised slowly working up to at least 150 minutes/week of moderate-intensity physical activity. Classes covered such topics as healthy eating, strategies to increase physical activity, managing slips and self-defeating thoughts, and other topics associated with developing and maintaining health behaviors.

No alterations were made to the curriculum, but additions of information relevant to the incarcerated environment (i.e., the commissary list and dining hall menu) were used as needed. The University of Pittsburgh has produced DVD versions of the first 12 Group Lifestyle Balance classes, and these were used in some of the classes; however, a live facilitator was always present to expand on information and answer questions.

Participants in the program were provided with liquid and dry measuring cups, pedometers, and food journals. Because of security concerns, participants did not have access to food scales. Instruction was therefore provided on alternative methods for determining portion sizes. Food journals were collected at each class, reviewed by facilitators, and returned at the following session with comments. These journals were used to track and encourage proper dietary changes. Both the male and female inmates had access to gyms, exercise equipment, and various physical activity classes that are routinely offered to the general population in both institutions.

Four separate intervention classes were held, two for each group. Classes for men and women were held separately due to the nature of the prison setting but were scheduled simultaneously, and class sizes did not exceed 13 participants.

The main outcome variables included weight, blood pressure, A1C, fasting lipid panel (total cholesterol, LDL cholesterol, HDL cholesterol, and triglycerides), and diagnosis of diabetes. Weight was checked on the same scale throughout the study, and scales were calibrated and safety-checked every 6 months. Blood pressures were taken manually by two individuals and compared. If the readings were within 6 points of each other, the average was recorded as the final data point. If the discrepancy was wider, both individuals re-checked until sufficient agreement was obtained. Laboratory values were obtained using venous samples, and only one Bureau of Prisons laboratory was used.

Additional measurements tracked included class attendance rates and reasons for nonvoluntary attrition (i.e., transfers and releases). Qualitative data were obtained using a written survey at intervention completion.

SAS Enterprise Guide version 7.11 (SAS Institute, Cary, N.C.) was used to complete all analyses. The baseline demographic characteristics and program completion were examined overall and between groups. χ² and t distribution test statistics were calculated to test for significant differences across relevant categorical strata and continuous variables, respectively. A P value threshold of 0.05 was used as the standard for significant differences. To account for the correlational nature of the data (i.e., several measurements taken on the same person) repeated-measures analysis of variance was performed using the software’s PROC GLM procedure to obtain both univariate and multivariate tests for repeated measures for each response (19–21).

Repeated measures were analyzed between groups (0–6 months) and within groups over time (0–12 months). The 6-month between-group analyses compared 1) the study period for group 1 starting at study baseline and ending at 6 months when they were actively engaged in the study intervention to 2) the study period for group 2 during the exact same time period when they were engaged in the control protocol only. In the below graphical representations of this analytical approach, “X” indicates engagement in the study intervention and “O” indicates engagement in the control protocol.

The 12-month overall analyses combined the 12 months of active engagement for both groups, such that we examined group 1 (study baseline to study 12 months) and group 2 (study 6 months to study 18 months) results.

All unadjusted models determined to be significant were assessed for confounding by sex, age, race, prediabetes status, program completion, and number of classes attended. The final adjusted models presented include only variables identified as confounders.

About half (55%) of this sample of 47 incarcerated people was male, and a little more than half (62%) were 45–64 years of age (Table 2). This population was racially/ethnically diverse, including mostly black (43%), white (30%), and Hispanic (26%) participants. A majority of the population (81%) screened positive for prediabetes. The two study groups were not significantly different with regard to sex, race/ethnicity, age, or prediabetes status. Overall, about half of participants (51%) completed the program, and participants attended an average of 14 classes. Participants in group 1 were significantly more likely (P <0.05) to complete the program (71 vs. 29%) and to attend more classes (mean of 17 vs. 11) compared to group 2.

After participating in this study for 6 months, between-group analyses revealed that group 1 (active intervention) experienced significantly greater decreases in weight and BMI compared to group 2 (active control) (Table 3). The average weight loss for group 1 was more than double that of group 2 (12 vs. 5 lb) (P <0.001). However, both groups (active intervention and control) experienced significant decreases in weight, BMI, triglycerides, and A1C after 6 months.

In combined analysis of both groups who received the active intervention for 12 months, significant decreases were observed in weight, BMI, and A1C (Table 4). The average overall weight loss over 12 months was ∼10 lb. Sustained losses between baseline and 6 months were either maintained (weight, BMI) or continued to decrease (A1C) at 12 months.

Participants who completed the 12-month lifestyle change program showed significant improvement in weight and A1C, which indicates the DPP may be an effective option for diabetes prevention in correctional settings. Average weight loss for program completers was 4.6% of starting body weight, a result that meets or exceeds those found in the community, where weight loss tends to average ∼4% but is often much lower, especially for groups with a high percentage of black participants (14,22,23). The reduction in A1C (mean reduction of 0.4% after 12 months) was also significantly higher than that found in most other studies, with a recent meta-analysis demonstrating a mean change of –0.21 (22,24). Although our active control was only in place for 6 months, the differences in the two groups at that point indicate that the positive results are most likely attributable to the intervention rather than to other variables.

Our study demonstrated no significant changes in blood pressure and lipid levels in the intervention group versus the control group at 6 months or as a pre-/post-intervention analysis at 12 months; however, improvements were still noted. In fact, the changes in blood pressure, HDL cholesterol, and triglycerides achieved by participants in our study after the 12-month program exceeded the average results reported in a meta-analysis of outcomes from U.S.-based DPP programs (24). For example, systolic and diastolic blood pressure dropped by 7 and 2.6 mmHg, respectively, in our study, compared to average reductions of 4.29 and 2.56 mmHg, respectively, computed in the meta-analysis (24). The original DPP study found long-term improvements in LDL cholesterol, triglycerides, and blood pressure for participants in all study arms, with participants in the lifestyle intervention group achieving similar results with less medication (25). It is therefore still possible that our results, although not significant, may indicate improved cardiometabolic outcomes for our participants in the future.

Our version of the DPP was similar to most of the CDC-recognized, year-long group programs implemented in the community. One of the biggest differences between the prison-based program and those in the general population was that classes consisted of single-sex participants. Although the facilitators reported that this affected the group dynamics of the classes, small sample sizes meant that analyzing outcomes by sex was impossible. Future studies may benefit from determining what, if any, differences exist in results for male versus female incarcerated participants.

Other differences in the program related to access to resources and adaptation of some of the program content. For example, inmates were not able to have food scales and often had their measuring cups and pedometers, provided by the program facilitators, confiscated by correctional officers because of miscommunications. This meant participants could not always accurately track their calories, fat grams, or steps. In addition, the curriculum content on eating out at restaurants or shopping at grocery stores was modified to reflect the choices available at the prison commissary. Participants were also taught other methods for controlling calories, such as using their hand to determine portion sizes, that can be used in any environment without regard to the breadth of food choices or tools available. The positive study results, which occurred without any negative security or administrative issues, indicate that these adaptations and differences allowed for a successful program within the prison. Adaptations such as those described here have become common for the DPP, as these modifications allow for increased relevance for the intended population and suitability for various settings, and they do not typically compromise outcomes (12,15).

Fewer than half of participants who began the lifestyle change program were present for the full 12 months and were considered program completers. Significantly more individuals from group 1 finished, which is likely due to the fact that the intervention stage of their program ended 12 months from baseline, whereas for group 2, the intervention ran from month 6 to month 18. The attrition rate that has been found in community settings is highly variable, ranging anywhere from 3 to 59%, with higher numbers typically found in urban or primary health care settings (22,23). Although our rate tends toward the higher side of that range, less than half of our attrition was attributable to voluntary exit of participants, most of whom expressed a lack of interest once their program was set to commence. Fifty-four percent of those who did not complete the program were either transferred to another institution or released from prison mid-program. It may be beneficial for future implementers to coordinate with custody staff to enroll participants likely to remain at a single institution for the entire 12-month program period. Residential drug treatment programs, for which inmates are often designated at specific institutions for the entire program course, may serve as a model for how to address this issue.

Reports of other types of health promotion interventions in the correctional environment are extremely sparse, with most studies featuring small sample sizes and few outcome measures. The limited data available, however, demonstrate some positive results that are similar to our own. A pilot study involving 15 inmates with diabetes who participated in a pharmacist-led education workshop demonstrated a significant decrease in A1C compared to the control group (26). The average A1C of inmates with diabetes also decreased significantly more for a group of female inmates (n = 24) who participated in nutrition education and received a reduced-calorie diet compared to a group of 39 control subjects who retained the higher-calorie diet that was provided to both groups at baseline (27). In the United Kingdom, an evaluation of health trainers who provide health information and support to patients on a variety of wellness topics found that both the incarcerated trainers and the incarcerated participants self-reported improvements in health behavior, knowledge, and skills, including healthier eating and improved fitness level (28). Similarly, inmates from a state prison who participated in four to five group classes on a range of nutrition topics also reported improved nutrition and general health compared to baseline (29). None of these interventions were as intensive as the DPP, and many relied on self-report, but they do help demonstrate that health promotion can be successful in prison settings. Investigating multiple approaches to improving the well-being of the incarcerated population is probably worthwhile.

This study had several limitations. Loss to follow-up was large due to unforeseen and uncontrollable movement of participants inside and outside of the prison system. The small sample size affected the extent to which we were able to explore overall tests of significance and affected our ability to assess for confounding in within- and between-groups differences. Thus, the final adjusted models only included variables for which confounding was identified. Because all participants eventually were enrolled in the intervention, only 6 months of control data were available, making determination of program effectiveness challenging due to possible confounding variables.

Except for class attendance, researchers did not track program compliance. There are therefore no data available on physical activity minutes completed by participants or on how that goal was met. The lengths of incarceration and sentences were also not collected, but doing so may allow for a greater depth of analysis in the future.

As the inmate population ages, health care costs continue to rise, and finding solutions to improving older prisoners’ health and well-being becomes increasingly important. These results indicate that the DPP may be one aspect of the solution to this emerging issue. Further research should be done with enhanced administrative and upper-management support that allows for reduced attrition. Correctional systems and communities would most likely benefit from identifying ways to implement the DPP in a way that satisfies the unique security and logistical needs of prisons. This study and its results make it clear that diabetes prevention is feasible and realistic in correctional settings and can serve as a springboard for future studies and programming in this unique environment.

No potential conflicts of interest relevant to this article were reported.

A.F. researched data and wrote the manuscript. M.S.G. conducted data analysis and wrote the manuscript. A.D. researched data. A.F. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The views expressed in this article are those of the authors and do not necessarily represent the views of the Indian Health Service, Cowlitz Indian Tribe, the Department of Health and Human Services, the Federal Bureau of Prisons, or the U.S. government.