Kaiser Permanente Colorado (KPCO) is an integrated health care organization that primarily serves the greater Denver metropolitan area. Members who had newly diagnosed prediabetes were potentially eligible to participate in this matched-cohort longitudinal study. In 2006 we began to access existing electronic medical records dated from February 2004 through March 2005 to create a cohort of 14,379 people with impaired fasting glucose (IFG) measurements of 100 to 125 mg/dL. IFG values were chosen instead of a diagnosis of prediabetes because of high variability in the use of the prediabetes code and changes in its definition over time. The inclusion criteria for this study was having an IFG measurement of 100 to 125 mg/dL, being aged 18 years or older, and being a member of the health care organization for at least 6 months before the study start date of February 2004. We excluded people with 1) an IFG measurement of 126 mg/dL or higher, 2) a diabetes diagnosis in the first 30 days after the IFG measurement, and 3) a dietitian contact in the 6 months before the study period ().
Figure. Flow of study participants in an intervention for patients with prediabetes, Kaiser Permanente Colorado, Colorado, 2004-2005. Dietitian contact refers to a potential participant having had an individual consultation with a dietitian and the reason coded (more ...)
There were 14,379 potential participants newly diagnosed with prediabetes eligible for the study. Of this group, 1,911 were ineligible (1,082 had diabetes diagnosed within 30 days after the first impaired fasting glucose [IFG], 807 had dietitian contact 6 months before study start date [February 2004]; and 22 had IFG measurement ≥126 mg/dL). Of the 12,468 eligible participants included in the matching program, 1,030 attended diabetes class, and 820 of them were matched to controls from the 11,438 potential controls. Sixty potential controls were excluded because they did not have 12-month follow-up weight measurement. A total of 760 matched participant and control pairs were included in final analysis.
Of the 12,468 eligible participants, 1,030 attended a single 90-minute small-group session that targeted personal action planning for healthful eating, physical activity, and weight management. This intervention, which was developed in partnership with KPCO, and its development process have been described in detail elsewhere (8
). Briefly, each class of approximately 10 to 20 participants began with a presentation by a dietitian or weight loss specialist that included information about prediabetes and diabetes, recommendations for a healthful diet and regular physical activity, and information on how diet, physical activity, and weight loss delay the onset of diabetes. The 90-minute session, which was designed to incorporate social cognitive factors such as increasing self-efficacy, reducing barriers to physical activity, and identifying rewards for a healthful lifestyle, involved a question-and-answer period and small-group problem solving. At the conclusion of the session, participants created a personal action plan for preventing diabetes. The personal action plans were structured to include individual physical activity, healthful eating, weight loss goals, personal reasons for wanting to avoid diabetes (eg, maintain health in order to spend more enjoyable time with grandchildren), and strategies for decreasing barriers to the physical activity and dietary goals. Four to 6 classes were offered monthly from March 2004 through February 2005. These classes were advertised through KPCO's magazine with all health promotion offerings, KPCO's Web site, and doctor referrals.
We used electronic medical records to select 1 or 2 controls for each member who attended a small-group session. Matching was based on the following: exact match on month and year of IFG measurement, exact match on sex, within 5 years of age, within 2 body mass index (BMI) units, and within 5 mg/dL for initial IFG measurement. Each matched control was assigned a dummy date for participation in a small-group session; the date was identical to the date for its matched case. Matching was completed by using the first IFG result within the study time frame. People with a second IFG measurement before attending class or before their dummy small-group participation date had to have results in the IFG measurement range to ensure eligibility; otherwise, they were excluded from this study.
Weight, as recorded in the electronic medical records from February 2004 through April 2006, was used as the primary outcome to investigate the effectiveness of the intervention. Weight was measured by clinical staff using a weight scale during routine medical visits at that time. The weight measurement obtained closest to the date of participation in the small-group session (recorded up to 30 days before the session) was used as the baseline value for small-group participants; for matched controls, the weight recorded closest to the dummy participation date was used. Follow-up values were measurements made 12 months after initial class attendance (recorded up to 30 days after the end of the 12-month period). We eliminated 60 matched controls because of the lack of 12-month follow-up weight measurements that were within the 30-day window. This gave us 760 matched pairs for the final analyses (n = 1,520; mean age [SD], 63 ; 53% were women) (). Participants in the small-group sessions (n = 760) were a mean age of 62 years. Racial and ethnic characteristic data were not available from the medical records used in this study. To determine the potential reach of the intervention, the proportional participation in the small-group sessions was calculated.
Although conditional analyses are typically used for studies that match on outcomes, it is less critical for studies that match on "exposure," as in this study (9
). Mixed models analyses were used to adjust for matching variables and covariates and to account for individual random effects over time. Additionally, the percentage of participants losing a clinically significant amount of weight (≥5%) was calculated. To test for group differences between small-group participants and matched controls, the nonparametric χ2
test of independence was used. All tests were 2-sided with significance set at α = .05. The SAS software program version 9.1.3 (SAS Institute, Inc, Cary, North Carolina) was used to analyze the data. This study was approved by KPCO's institutional review board.