Dialing In: Effect of Telephonic Wellness Coaching on Weight Loss | Page 2

Published Online: February 25, 2014
Min Tao, PhD; Krishna Rangarajan, MS; Michael L. Paustian, PhD, MS; Elizabeth A. Wasilevich, PhD, MPH; and Darline K. El Reda, DrPH, MPH
During the course of the study period, there was a change in the health risk assessment questionnaire administered by the health plan. The University of Michigan Health Risk Appraisal (HRA) was used through May 200928; afterward the StayWell Health Media Health Assessment (HA) was used.29 This change in health assessment questionnaires reflects a business decision of the health plan and not of the authors, who conducted the evaluation retrospectively. The HRA consists of 47 questions while the HA consists of 150 questions. Common health risks assessed by the 2 questionnaires include body weight, existing medical conditions, physical activity, nutrition, alcohol, smoking, stress, safety belt use, skin protection, absenteeism at work, perceived health, and overall life satisfaction. On each instrument, respondents also report their race/ethnicity, gender, level of education, and motivation to lose weight. Although both questionnaires assess similar health risks, the wording of the questions and answers differ markedly. For example, the HRA asks “Do you have heart problems? (Never/Have Currently/In the Past)” while the corresponding HA question is worded as “Have you been diagnosed with coronary heart disease? (Yes/No).” For this evaluation, we only used questions from the 2 questionnaires that could be reconciled to allow consistent measurement over time.


As shown in the Figure, 40,222 individuals responded to 2 questionnaires at least 6 months apart from 2008 to 2010 for our pre-post evaluation. The average time between baseline (T1) and follow-up (T2) questionnaires was 385 days (approximately 1.05 years) with a standard deviation of 70 days. BMI was calculated from responses to height and weight questions on these surveys. Overweight was defined as a BMI between 25 kg/m2 and 29.9 kg/m2 and obesity was defined as a BMI of 30 kg/m2 or greater. Among respondents, 16,178 (40%) were considered overweight or obese at baseline. Of the overweight and obese respondents, 5653 (35%) were referred for a wellness care management (WCM) program during the study period between baseline and follow-up; 8922 (55%) did not receive any referral during the study period, and 1603 (10%) received a referral, but outside of the period, and thus were excluded from the analysis. Two percent (2%) of the adults who reported a more than 2-inch difference in their height across the 2 time periods were excluded from study population due to potential data quality concern. Respondents with complete data on weight, height, age, gender, race/ethnicity, level of education, motivation to lose weight, smoking status, and comorbidities were used for the analysis. The final study population was divided into 3 groups. The intervention group consisted of obese or overweight respondents targeted for the telephonic wellness coaching program (N = 1448, including 1050 participants and 398 nonparticipants). The 2 comparison groups were overweight or obese respondents: (1) not targeted for any telephonic WCM program (N = 7586) and (2) not targeted for telephonic wellness coaching, but targeted for other WCM programs, primarily disease management (N = 1270).

Statistical Analysis

Differences in the distribution of demographic characteristics, self-reported chronic disease status, smoking status, and motivation to lose weight at baseline among the study groups were tested using c2 test, while differences in baseline BMI among the groups were tested using ANOVA. Paired t-tests were used to detect whether significant weight changes were reported for each group between baseline and follow-up.

To illustrate the impact of outliers, we reported the 1-year weight change for each group after removing outliers in 2 ways: a) statistically, by calculating upper and lower fences based on mean and standard deviation of the weight changes between T1 and T2 using “outer fence” formula (mean + 4.72* standard deviation),30 and b) by removing 4% of the study population that had more than an 18-kg (40-lb) difference (increase or decrease) in self-reported weight based on health coach suggestions of what constituted a substantial weight change.

The association between self-reported weight change and the wellness coaching program was determined using multivariable linear regression, adjusted for sex, race, education, motivation to lose weight, and comorbidities. P values less than .05 were considered statistically significant. Analyses were performed in SAS 9.2.31

In addition to our primary analyses among targeted members, we also conducted stratified analysis among wellness coaching program participants. We examined the average weight change stratified by: a) wellness coaching objective; b) the number of phone calls with the wellness coach; and c) the initial stage of change. Since wellness coaching participants could establish multiple goals upon program initiation, we applied a hierarchy to make these categories mutually exclusive. The hierarchy was: weight loss > physical activity > nutrition > smoking > others. Due to sample size concerns, the intensity of intervention (ie, number of phone calls the participant received) was dichotomized into 1 to 3 calls and 4 or more calls. Although the program is designed to consist of 4 calls, on rare occasions participants (<1%) may have had more than 4 calls with a health coach if additional coaching was requested by the participant. Participants who received 4 or more phone calls with health coaches were considered program completers. We used a paired t-test to determine whether weight changes were significant for each group between baseline and follow-up, with 95% confidence intervals (CIs) for weight changes were reported.


Demographic Characteristics for Study Population

Table 1
shows the characteristics of the study population. This population was well educated, with 70% of them having had at least some college education. The majority of the population was Caucasian; 55% of the population was male. The distributions of gender, race/ethnicity, and education were not different among the groups. More than 70% of respondents reported having motivation to lose weight in all study groups.

PDF is available on the last page.

Issue: February 2014
More on AJMC.COM