Study design
This study employed a two-group randomized design in rural upstate New York communities. Medically underserved communities with a Rural–Urban Commuting Area code (RUCA) of 4 or higher (micropolitan or rural) were selected. Community sites were pair-matched based upon population size (394 to 8,836) and RUCA code (4.1 to 10.2) [22]. Communities were geographically distinct and were not involved in the original SHHC trial (SHHC-1.0).
Following baseline assessments, communities were randomly assigned in pairs to intervention (n = 6) and control (n = 5; delayed intervention) groups by a statistician who was not involved further in the study. Complete study details have been published elsewhere and are only briefly recapped here [23, 24].
Intervention
The original SHHC-1.0 intervention was created based on extensive community input, including focus groups, surveys, and community audits. Results from the original trial and the process evaluation led to the creation of the refined SHHC-2.0 intervention [21, 25]. The SHHC-2.0 program consists of twice-a-week 60-min experiential group physical activity and nutrition education classes along with sessions on social and environmental change over 24 weeks; the intervention programming started in Spring/Summer of 2017 (April – June) and finished in Fall 2017 (September –November) [23].
Cooperative Extension health educators (herein Extension educators) led classes at various community locations, including libraries, town halls, churches, and other public buildings. The study team provided exercise equipment (low-weight dumbbells, yoga mats), aerobic exercise videos, a Leader Toolkit, Participant Guides, and Health Journals (for participants). Extension educators attended an intensive one-day training and had weekly support calls during active program implementation. Fidelity was assessed by leader questionnaires completed after each class and community site visits conducted by trained staff.
Physical activity specific intervention description
The general goals for the physical activity portion of the curriculum were based on guidelines from the Office of Disease Prevention and Health Promotion [26]: (1) Minimize sedentary time; (2) Strength train at least two days per week; (3) Engage in moderate to vigorous intensity exercise at least five days per week; and (4) Be as physically active as possible daily (taking the stairs, walking farther in a parking lot, etc.).
Forty-five of the 48 classes included progressive strength training and aerobic exercise, which averaged 13 min of strength training per class and 24 min of aerobic exercise per class. Aerobic exercise included walking and aerobic dance DVDs, which progressed from low intensity to moderate intensity ranging from 15 to 30 min per class (e.g., walking slowly to walking intervals with brisk walking or light jogging). The progressive strength training program included warm-up and cool-down stretches (four exercises) ranging from 10 to 20 min per class. Exercises were done in two sets of ten repetitions for a slow 8-s count. The topics covered in class progress over time from introductory strength training and aerobic exercise concepts to sustaining physical activity strategies. The beginning of the program is focused on behavior change, then the focus turns to examine the difficulties and accomplishments of behavior change, and finally how to sustain behavior change.
Participants were also encouraged to engage in physical activity outside of class time, including low to moderate intensity exercise, moderate to vigorous intensity exercise, strength training, and being as physically active as possible. The curriculum provided strategies and recommendations for exercising in bad weather and included programming for both indoor and outdoor options.
Study sample
Enrollment occurred between January and June of 2017. Extension educators recruited participants using flyers, radio advertisements, newspaper articles, social media, and word of mouth. Inclusion criteria included being female, aged 40 years and older, and either: (1) overweight (BMI = 25–30) and sedentary (no more than one bout of > 30 min of leisure physical activity per week on average, during the past three months) or (2) obese (BMI > 30). Women were excluded if they did not provide informed consent or permission from a healthcare provider, had systolic blood pressure > 160 mmHg or diastolic blood pressure > 100 mmHg, had a resting heart rate < 60 or > 100 bpm, had a cognitive impairment, as determined by a 6-item cognitive screening test [27], were participating or planning to participate in another health behavior change program in the next 12 months or were unwilling to be randomized to either group. Once the paired sites had enough participants with baseline data collection complete, then those sites were randomized and intervention site could begin. The Cornell University and Bassett Healthcare Network Institutional Review Boards approved all study procedures.
In accordance with the design of the first trial [18], this trial was powered at 80% to detect the primary outcome, mean change in weight from baseline to 24 weeks between groups of 1.95 kg, with a two-sided alpha, accounting for intraclass correlation of clusters and 15% attrition. It was not specifically powered for the analyses in this report.
Measures
Participants were asked to wear an accelerometer and complete an online survey via Qualtrics on demographics, physical activity behaviors, and psychosocial physical activity measures just prior to the start of the intervention (baseline), at 12 weeks (halfway through the program), and immediately following the 24-week program (post-intervention).
The measures are detailed below:
-
Light intensity physical activity and moderate to vigorous intensity physical activity (MVPA) were measured as continuous variables (averaged to minutes per day) using the accelerometer ActiGraph Model GT3XE (ActiGraph LLC, Pensacola, FL). Participants were instructed to wear the device at the hip for seven days and only to remove it when sleeping, bathing, or swimming. Data were recorded at 30 Hz and analyzed using an epoch length of 60 s. The low-frequency extension filter was used, and non-wear time was excluded by Choi et al. algorithm [28]. Daily level data were excluded if wear time was less than 10 h (600 min) in a day. Freedson cut points were used for physical activity intensity [29].
-
A dichotomous variable was then created to test if participants met aerobic physical activity recommendations of greater than or equal to 150 min of MVPA per week.
In addition to the accelerometer measures, we collected self-report physical activity data to include such activities as swimming, where the accelerometer could not be worn, or weightlifting, which is strenuous but generates few steps. The International Physical Activity Questionnaire Short Form (IPAQ-SF) was used to self-report on physical activity and scored according to IPAQ’s Guidelines for Data Processing. Metabolic equivalent (MET) minutes per week were compiled according to the IPAQ analysis guidelines [30]. Sedentary behavior was self-reported using the Sedentary Behavior Questionnaire (SBQ) [31]. Self-efficacy for physical activity (Self-efficacy for sticking to exercise habits [scale 1–5]) and for making time for exercise (scale 1–5) [32], social support for physical activity (family participation in exercise [scale 10–50], friend participation in exercise [scale 10–50], and family rewards and punishment for exercise [scale 3–15]) [33] and attitudes toward exercise (combined attitude toward exercise score [scale 1–14]) were also assessed [34].
Basic demographic information was collected at baseline. Demographic questions were derived from national surveys (e.g., U.S. Census). Study participants were instructed to report any adverse events to leaders and/or the research team at any time. Survey questions about adverse events were included at 12 and 24 weeks.
Statistical analyses
For the primary analysis, the differences in change from 0–12 weeks and 0–24 weeks between groups were analyzed using intent-to-treat linear mixed-effects multilevel models, which included random cluster (community) effects to account for the community-level randomization and correlation between participants in the same community as well as a priori covariates age and education. Two secondary analyses were done: (1) with participants 60 years and older and (2) mixed logistic regression with random cluster (community) effects tied to achieving the goal of 150 min per week of MVPA. Participants were categorized by meeting or not meeting that recommendation at the end of the intervention.
Data were missing in the following proportions at each time period for the survey: 5% at baseline, 29% at 12 weeks, and 37% at 24 weeks. For accelerometry, data were missing for the following: 3% at baseline, 31% at 12 weeks, and 35% at 24 weeks. To explore the potential that data may not be missing at random, baseline characteristics of respondents and non-respondents were compared at 24 weeks for each data collection tool (survey and accelerometer). No significant differences in accelerometer data for respondents versus non-respondents were observed for age, income, education, race, BMI, weight, meeting physical activity guidelines, or self-reported perceived overall health. For the survey, non-respondents tended to have a higher BMI compared to respondents, and therefore BMI was used as an auxiliary variable for multiple imputation (MI).
MI was used to estimate missing data and standard errors for the intention-to-treat analyses. Imputations followed standardized, rigorous procedures, including auxiliary variables (random assignment group, community site, age, education, and BMI) and employed hierarchical approaches. Fraction of Missing Information (FMI) was used to measure the level of uncertainty about the values imputed for missing values (median FMI for outcomes was 0.21 [range: 0.03 to 0.57]). We used 70 imputations, which satisfies the recommendations to have the number of imputations (at least) equal to the highest FMI percentage.
Sensitivity analyses
Sensitivity, “tipping point,” analyses were conducted to identify the point at which adjusting imputed values reversed the main findings and determine the plausibility of erroneous conclusions based on data not missing at random and MI modeling [35]. To determine the sensitivity of the analysis to outliers, we ran the same analyses with outliers (1.5 interquartile range above the third quartile or below the first quartile) removed. Complete case analysis was also done using restricted maximum likelihood to incorporate all available data.
Analyses were conducted in 2022 using SAS, version 9.4. The Principal Investigator had full access to all study data and takes responsibility for its integrity and the data analysis.