Obesity Crisis in America

• An estimated 30% of Americans are considered obese

Obesity Crisis

• Effective, low-cost interventions are needed to increase physical activity to prevent and treat obesity
• The rate of obesity has dramatically increased in the U.S. over the last two decades
• In 1980, ~10% of the U.S. adult population met the criteria for obesity, having a BMI greater than 30
• CDC reported a 35.7% obesity rate more recently in 2013
• Obesity rate increases have been observed in many other countries, representing a worldwide problem
• A global call has been made by health organizations, governmental research funding institutes, businesses, and economists for solutions to reverse the obesity trend
• Strategies for reducing obesity include increasing caloric expenditure, decreasing caloric intake, or both
• Physical activity is operant in nature and malleable, making behavior analysts uniquely qualified to develop behavioral treatment strategies to address suboptimal activity levels and monitoring frequency and duration of specific behaviors then provision of reinforcers when some criterion is met

Contingency Management

• Contingency management(CM) has had success increasing physical activity
• Those benefiting has been shown in healthy adults, older adults, children, obese adults, individuals with muscle or arthritic pain or brain injury, and developmentally disabled individuals
• Reinforcers include money, privileges, tokens or points, and prize draws
• Studies using weight loss alone as an endpoint produced mixed results
• Physical activity is sensitive to the consequences that follow it, demonstrating the potential of operant techniques to alter these behaviors

Accelerometer Tech

• Recent advances in accelerometer technology have largely eliminated the challenge of continually monitoring ongoing behavior outside the laboratory setting
• Accelerometers monitor behavior in the X, Y, and Z planes, are small, and can be worn in discrete locations on the body
• Temporal resolution with which data are recorded has improved
• Activity can be measured in short-duration bins rather than simply aggregating steps across the entire measurement window to permit structural analysis of ongoing activity
• Stepping occurs in bouts of high activity without appreciable pausing interspersed with relatively longer periods of inactivity
• Grouping steps into bouts provides for the derivation of within- and between-bout response rates
• Structural analyses may allow more tailored intervention strategies
• Accelerometers provide a convenient means for transmitting and accessing data making them ideal for monitoring, recording, and analyzing physical activity as it occurs in real time
• A competitive market for these fitness devices has given rise to the Fitbit, the Jawbone Up, and various iPhone and Android applications
• These devices often include access to software that constructs graphical representations of behavioral data with the purchase of the device or application

CM Considerations

• A second challenge for CM interventions centers around designing a procedure in which subjects can contact reinforcement early in the intervention
• Early failures to earn reinforcers is associated with a higher likelihood of disengagement in treatment
• Ensure contact with the reinforcer by shaping behavior toward a desired endpoint gradually using percentile schedules
• Percentile schedules have been shown to be efficacious in shaping health-related behavior
• Smoking abstinence can be shaped by basing reinforcement criteria on current smoking levels, as indicated by exhaled CO readings
• Instead of requiring perfect behavior (i.e., complete abstinence) immediately, gradual changes are more effective
• Titration Procedure involved in reinforcing behavior to encourage improvement over time
• Accomplished by measuring baseline rates of behavior over a period of time and setting reinforcement criteria based on the rank-ordered distribution of obtained rates of behavior
• Criteria are set to require that behavior exceeds values obtained on the third best day in the last week, ensuring participants are meeting criteria
• Galbicka suggests approximate percentile (70th) as a good criterion to improve behavior

Methodology

• CM procedures used in the current study feature techniques developed in the substance abuse literature
• Directed towards accomplishing the goals of maximizing efficacy while minimizing costs
• Using an intermittent schedule of reinforcement based on the “prize bowl” interventions developed by Petry et al. (2004) for reducing cocaine consumption
• Consequences (the draw) are delivered for desired behavior, but costs are minimized by only providing actual wins under a random-ratio schedule
• The study used a Fitbit accelerometer ($99) to measure subjects' steps for 3 weeks that automatically records total steps taken during 24-hr periods (beginning at midnight) in 1-min bins
• Evaluate effects of intermittent prize reinforcement on meeting step criteria and to characterize the structure of behavior change using bout analysis

Study Parameters

• Participants were 15 students (10 women, five men) between 18-26 years old
• Employed an ABA reversal design where all three phases (Baseline 1, intervention, Baseline 2) lasted 1 week each
• The form of daily prize draws during Baselines 1 and 2 was contingent only on wearing the Fitbit
• Bring Fitbit to lab for charging and data uploading 3 times per week, rather than issuing a docking station is done to prevent access to Fitbit website/data available and to increase the likelihood that subjects would complete their prize draws in a timely fashion
• Prize draws during the intervention was made contingent wearing the Fitbit and meeting experimenter-determined minimum step criteria
• Daily step totals were reported by subjects via text messaging or e-mail and step criteria for following day were relayed from researchers using the same mode of communication
• Prize bowl contained 100 numbered raffle tickets the lottery was explained to the subjects

Measurements

• Data was downloaded through Google Documents with special permission from Fitbit via the Fitbit API developers site
• A programming script run through Google Documents allowed the downloading of minute-by-minute step counts throughout each day
• After downloading, the number of steps taken per minute in 24-hr intervals(total of 1,440 1-min bins) was entered into a Microsoft Excel file for further analysis and was analyzed using Microsoft Excel, Systat, and Sigmaplot
• For each minute, average interresponse times (IRTs) were calculated by dividing 60 s by the total number of steps during that period
• Data collection involved coding according to the percentage of steps throughout the day for which it accounted
• Temporally adjacent empty bins were summed to produce estimations of long IRTs
• Total active minutes reflected the number of bins in which any responding occurred within the day and sleep minutes were calculated by adding together the two longest IRTs
• Inactive minutes were calculated by adding together all the nonsleep bins in which no steps occurred and wear time was calculated by subtracting sleep minutes from the total number of minutes in a day (1,440), then percentage of active minutes was calculated as active minutes divided by wear time
• An exploratory structural analysis of walking was performed as measuring bout lengths, within-bout step rates, and bout-initiation rates for the data

Results

• 13 subjects completed the study
• Visual analysis of daily raw data indicated a clear increase in daily step counts during intervention for several subjects (+23% , range, -15% to +90%)
• Step counts declined after the stepping contingencies were removed (except in subject 14)
• Average step counts varied among the subjects
• Wear time decreased in Baseline 2 in relation to baseline 1 and the intervention
• Statistics were performed to detect whether Fitbit wear time varied between conditions
• Average wear times were 848 min (SD=135) during Baseline 1, 870 min (SD=152) during intervention, and 753 min (SD=181) during Baseline 2
• The change in Baseline 2 could be contributed to decreased wear time
• A RMANOVA with Phase as the factor showed a significant effect of phase on step counts
• Statistics also show the correlations between step counts and four variables: active minutes, bout length, within-bout stepping rate, and pause length
• Results indicated that changes in step count from day to day were highly influenced by the number of minutes the subject was active for each day
• Pearson's correlation coefficients showed all subjects are different and not statistically correlated concerning these variables

Conclusion

• Subjects in this study were all young, healthy college students that were mostly not obese
• Intervention resulted in only two subjects that reported greater than 10,000 steps
• Sleep time, wear time were not significantly different between Baseline 1 and intervention and decreases in Baseline 2 may have been influenced wear time
• Study indicates that the improvements seen in intervention relative to Baseline 1 were likely not due to simply wearing the Fitbit longer during the day
• There are several advantages to increasing stepcounts using CM procedures: Objective measures increasing the probability of reinforcing appropriate behavior
• Procedures provide a relatively frequent reinforcement of desired behavior and the percentile schedule employed in the current study allows the gradual shaping of increased physical activity based on individual performance rather than requiring an abrupt, large shift in behavior
• Limitations, possible future study parameters and recommendations were discussed