Employee Baseline Characteristics
Of the 2302 employees at the work sites, 265 (12%) agreed to take part in the program. The numbers for each site and the characteristics of these participants are shown in . The mean age across the sites was 40.9 years (SD 8.1), and the mean BMI at the start of the program was 27.1 kg/m2
(SD 4.8). Ethnicity was also recorded to determine appropriate risk of metabolic disease for a given waist circumference and BMI [30
]. Of the 265 employees, 13 (5%) were Asian, 3 (1%) were Black, 244 (92%) were White, and the remaining 5 (2%) classified themselves as other or mixed.
Baseline demographic data for participants at each site
Of the 265 participants who agreed to take part, 233 started the program (); 32 participants were excluded or withdrew in the 3-month period between initial recruitment and program start for the following reasons: did not attend screening/training session (n = 5), ineligible employment status (not permanent or leaving employment, n = 8), illness (n = 1), extended holiday (n = 1), or no physician approval received (n = 17).
Figure 1 Flow chart of enrollment, withdrawal, and follow-up (*For LOCF analysis of weight change, the 5 subjects withdrawing after the program start were also included; noncompliant participants were not included in the weight LOCF analysis as no follow-up weight (more ...)
The office sites had a higher total number of employees compared to the factory sites, hence the larger number of participants from these office sites. There were significant differences between the sites at baseline for age, BMI, BP, and heart rate (Pvalues for testing equivalence of the sites: age, P = .009; BMI, P < .001; systolic BP, P = .004; diastolic BP, P = .001; pulse, P < .001), with employees from the factory sites tending to be older and having higher baseline BMI and BP (see ). BP medication was recorded at baseline and at the end of the 12-week study. The use of hypertensive medication was more prevalent in employees at the Factory North site at baseline (6/43, 14%) compared to the other sites combined (5/221, 2%); 3/264 (1%) screened participants started BP medication following the baseline screening.
We compared the baseline health profiles of the 264 employees participating in the MiLife Web program to HRA data collected previously from a larger group of employees at the same sites (n = 992 total). Employees who participated in the Web program were of a similar age as those attending the HRA but had a higher mean baseline BMI (HRA: 25.0 kg/m2[SD 4.1]; MiLife: 27.1 kg/m2[SD 4.8]; P < .001) and higher mean baseline diastolic BP (HRA: 78.9 mmHg [SD 10.0]; MiLife: 86.0 mmHg [SD 11.0]; P < .001). The difference in BMI between the two populations was most noticeable at the factory sites: the mean BMI for Factory North HRA was 26.6 kg/m2(SD 4.2, n = 174) compared to 29.7 kg/m2for Factory North MiLife (SD 5.6; P < .001), and the mean BMI for Factory South HRA was 26.1 kg/m2(SD 4.2, n = 122), compared to 28.4 kg/m2for Factory South MiLife (SD 4.7; P < .001).
Of the 233 participants starting the program, 6 withdrew and 5 were noncompliant (no data upload or log-in throughout the 12-week period). In the remaining 222 subjects, website use remained high, with 78% (173/222) of the participants still using the website at the end of the 12-week study and 69% (153/222) continuing to use the website after the 12 weeks ().
Employee website use during and following the 12-week study, by work site
Nonusage attrition data (the proportion of participants who stopped using the program and the proportion who remain) are presented in in comparison to reported nonusage attrition data from other Internet eHealth interventions [31
]. Nonusage attrition rates were significantly different between the sites, with the highest use at Office South (72/75 using the Web program at 12 weeks) and the lowest use at Factory North (19/36 using the Web program at 12 weeks; P
< .001). There was no difference in nonusage attrition rates between Factory South and Office North. Logistic regression including the baseline independent variables with site as a covariate showed that nonusage attrition was lower in both Factory North and Factory South as age increased (OR 1.07, P
= .03). To illustrate this, 30/58 (52%) of participants 48 years old and under were using the program at 12 weeks, compared to 21/23 (91%) of participants over 48 years old. Inclusion of the dependent outcome variables in this model showed that age was no longer significant but that nonusage attrition was lower in those subjects with a greater weight change over the 12-week period, independent of site (OR 1.38, P
Figure 2 Nonusage attrition curves  for MiLife and Farvolden et al , Linke et al , and van Straten et al  eHealth interventions
Log-in data are shown in ). Spaces between weeks can be clearly seen, indicating that most participants were using the website on weekdays and not weekends, with 7381 (92%) of the 8067 log-ins recorded during the 12-week study occurring between Monday and Friday. Data recorded over the Christmas holiday period also suggest that most users did not use the program on non-workdays, although 69 (31%) of the 222 website users did log-in to the program at least once during this 2-week break. Continued use of the program outside of the study period can also be seen on this graph.
Figure 3 Participant’s log-in data throughout and following the study period (Each dot represents a user’s visit and log-in to the website. Green boxes are the website visits recorded during the 12-week study period. Data within the red box were (more ...)
During the first 2 weeks of the study, participants were spending more time on the website per log-in compared to the subsequent weeks (mean week 1: 11.6 minutes; mean week 2: 8.6 minutes; ). After this initial period, the mean time per website log-in (weeks 3-12) was approximately 7 minutes. The total website interaction time per week was also collected for each user. As both the frequency of use and the time per visit dropped with ongoing program use, the total interaction time with the website () was higher in the first 4 weeks, dropping to a mean value between 10 and 20 minutes per week for the remainder of the study. Multivariate linear regression showed that the log total interaction time over the 12 weeks was higher in participants with the greatest weight loss (slope = .082; P< .001). To illustrate the magnitude of this slope (mean log duration 2.472), each additional kilogram of weight loss was approximately equivalent to an extra hour of program use over the 12-week period. Gender was also associated with total interaction time over the 12 weeks, with women spending, on average, 200 more minutes interacting with the Web program over the 12-week period (P= .002).
Mean website log-in duration and frequency throughout the study period (Data points are means with standard error and are presented for all employees as there was no significant difference between the sites based on ANOVA.)
Mean interaction time with the website throughout the study period (Data points are means with standard error and are presented for all employees as there was no significant difference between the sites based on ANOVA.)
Of the 228 employees using the program (222 starting the program plus 6 withdrawals during the 12-week study), 211 (93%) uploaded weight data that could be used to determine weight change during the study period using the LOCF. The mean weight change in this group was −2.6 kg (SD 3.2; P< .001; ).
Weight change from baseline using the LOCF and the mode in which the most time was spent for each employee uploading weight data during the 12-week program
Mean weight loss was higher in those employees who spent most of the 12-week period in the weight loss mode (132/212, 63%; mean weight change −3.5 kg, SD 3.6). No statistical comparison was conducted between modes since subjects could switch modes during the study period. There was no significant difference in weight change between the sites, but there was a significant inverse association between baseline BMI and the amount of weight lost over the 12-week period (−0.284, P < .001), indicating that employees with a higher starting BMI, on average, had greater weight loss during the study period. shows the baseline BMI distribution in each of the modes, suggesting that subjects with a higher baseline BMI spent most of their time during the 12-week study in the weight loss mode.
Baseline BMI and mode in which the most time was spent for each employee uploading weight data during the 12-week program
Physical Activity (PA) Data
The accelerometer-recorded levels of PA were highly variable between individuals in the group, with values ranging from 12 to 714 minutes of moderate or above PA per person per week. The average recorded level for the group was 173 minutes (SE 12.8) of moderate or above PA per week.
Choice of Goals
At all sites, weight loss was the most popular mode. Of the 228 website users (including those who withdrew during the 12-week study), 138 (61%) spent the most time in weight loss mode, 46 (20%) spent the most time in weight maintenance mode, 39 (17%) spent the most time in the PA only mode, and 4 (2%) spent the most time in the nonactive browse mode of monitoring without goal and target setting.
At the start of the program, each participant was asked to select one or more goals that he or she would most want from a list on the website. Research has shown that people with a strongly desirable goal are more likely to enact their intentions to perform a health behavior [34
]. The list consisted of the following goals: improve fitness, increase flexibility, improve health, reduce risk of heart disease, reduce blood pressure, look better, improve mood, improve quality of life, feel slimmer, improve stamina, other. shows the frequency of selection of these items by gender. The most frequently chosen goals for men were “health” and “heart disease,” while the most frequently chosen goal for women was “feel slimmer.” Men were more likely than women to select “heart disease” and “blood pressure” as reasons for participating, while women were more likely to select “look better” and “feel slimmer.”
Frequency of choice of the listed goals prior to beginning the 12-week Web program (F = women, M = men)
Blood Pressure Data
At the 12-week BP assessment, some employees at each site were lost to follow-up (Office South, 28%; Office North, 35%; Factory South, 35%; Factory North, 51%). Logistic regression with site, age, and gender as covariates showed that participants with a higher baseline BP were more likely to attend the follow-up (P= .047).
The high level of participants lost to follow-up was possibly influenced by the proximity of these measures to the Christmas holiday period. As a result, data have been aggregated for all employees who returned for a 12-week BP measure (n = 135, excluding those on hypertensive medication). The mean baseline BP in this group was 129/86 mmHg (SD 15/10, range 94/64 to 181/119), and the mean 12-week BP was 128/80 mmHg (SD 15/10, range 95/59 to 164/100). There was a significant reduction in diastolic BP (−5.9 mmHg, SD 9.9;P< .001).
Data from the Pittsburgh Sleep Quality Index (PSQI) questionnaire collected at baseline and 12 weeks (n = 93 completed) suggested an increase in sleep quality overall, corresponding to a decrease in the global PSQI score (P = .004). This was particularly evident in the following PSQI components: self-assessed overall sleep quality (P < .001), hours of actual sleep achieved (P = .01), ease of both maintaining attention and/or enthusiasm for everyday tasks (P = .006).
Data collected from the exit questionnaires (n = 130) showed that 101 employees (78%) found the website very easy to use, with the most useful tools listed as those providing PA analysis, planning, and information. Many employees liked wearing the PA monitor and found that having it on served as a constant reminder to keep to the program. The site was seen as informative, motivating, and helpful. The PA and weight charts were thought particularly helpful as they enabled participants to monitor their progress and played an important role in providing feedback and motivation. The low response rate (130/222 website users, 59%) to the exit questionnaire may have been influenced by the proximity to the holiday period.
The OH staff at the work sites who responded to the survey (n = 6) agreed that “the study had been a supportive program in the company objective”of vitalizing employee health. Several of the OH staff commented that the study had been a positive initiative that participants had found enjoyable and rewarding and which should be encouraged. Employee participation in the program resulted in some extra work for OH staff in answering participants queries, although OH staff were generally happy with the resources they had received from the study team to support participants.
Feedback from managers at the work sites (n = 6) was overall very positive, and managers received positive feedback from participants. Managers noted that the study appeared to have been beneficial in the workplace, and participation may even have led to increased job satisfaction in some instances. Managers were also in agreement that they would encourage future staff participation in the Web program and would recommend participation in the program to other sites.