We used the Washington State Workers’ Compensation Disability Risk Identification Study Cohort (D-RISC)⁹ data to examine the prevalence of overweight and obesity at the time of injury, the incidence of clinically significant weight gain in the year after injury, and early predictors of weight gain 1 year after occupational back injury. In D-RISC, potential risk factors for chronic disability were assessed in domains of interest that were used previously for occupational injury research.^8-11 Eight domains (sociodemographic, employment-related, pain and function, clinical, health care, administrative/legal, health behavior, and psychological)⁸ were assessed in baseline telephone interviews with workers with recent back injuries.

D-RISC was a prospective, population-based study that recruited Washington State Workers Compensation State Fund workers from June 2002 through April 2004 with accepted and provisional claims for occupational back injuries. Weekly claims review identified workers who missed at least 4 days from work and received wage replacement benefits (temporary total disability). Approximately two-thirds of the non-federal Washington workforce is covered by the State Fund. The remaining third are covered by large, self-insured companies and were not included due to insufficient administrative data.

In D-RISC, from the claims database, 4,354 workers were identified. Of those, 1178 (27.1%) could not be contacted, 909 (20.9%) declined enrollment, and 120 (2.8%) were ineligible.⁹ The remaining 2,147 (49.3%) were enrolled in D-RISC and completed baseline interviews. Persons were later excluded from the analysis sample if they were not eligible for wage-replacement benefits in the first year after claim submission (n=240), were hospitalized for the injury (n=16), were missing information on age (n=3) or were not confirmed to have a back injury upon medical review (n=3).⁹ Hence, 1,885 (43.3%) were included in D-RISC. Of the 1,885, 1,319 participants completed the follow-up interview approximately 1 year after claim receipt and 1,269 (96.2%) participants reported their weight during both interviews.

Upon inspection of the data, 16 participants had very large weight changes after 1 year (≥ 50 lbs). From additional administrative records, we were able to obtain other data on weight for 3 of the 16 participants, and used these data in the analyses. We excluded 6 of the 16 participants from analysis due to inconsistencies between self-report and clinical data that could not be reconciled. The self-reported and clinical data of the remaining 7 participants, among the 16, were very similar and the original self-reported weights were retained in the data, creating a final analysis sample of 1,263 participants.

The analysis sample was slightly older [mean age (SD) 40.3 (11.1) vs. 37.5 (11.2) years, P < 0.001], had fewer workers of Hispanic ethnicity (14% vs. 21%, P = 0.008), was more educated (less than high school 11% vs. 19%, P < 0.001), was more likely to be married or living with a partner (68% vs. 57%, P < 0.001), and contained more workers with general health insurance (72% vs. 60%, P < 0.001), as compared to the 622 persons who did not complete the follow-up survey or were excluded due to problematic weight data.

Study participants completed structured telephone interviews at baseline and at 1 year. Workers were asked their current weight in both interviews. The baseline interview also asked for participant height, which was used to determine baseline BMI (weight in pounds divided by (height in inches² x 703)).²⁵ Baseline measures for the current study were a subset of those obtained in the larger study, with selection based upon prior research pertaining to occupational injury, BMI, and weight and weight change. Additional data were obtained from the Washington State Department of Labor and Industries (DLI) claims database, including the region of the worker's residence, the worker's type of industry, the specialty of the first provider seen for the injury, and the number of days between the injury and the first medical visit for the injury. Additionally, medical record review by trained occupational nurses, with substantial inter-rater reliability, was used to determine injury severity.¹¹ (See Table 1 and the Appendix for more information about the measures.)

A weight gain at one year of at least 7% of baseline weight was used as a measure of clinically significant weight change.^{14, 16 – 18} Definitions of clinically significant weight change are not consistent in the literature. Weight changes of any,⁷ 3%,²⁷ and 5% have also been used,^{12, 13, 28} but we chose the more conservative measure of a 7% gain.

To test our hypothesis that weight gain was associated with receiving wage replacement benefits at one year after claim submission, we used a measure of wage replacement receipt obtained from administrative records that corresponded to a similar timeline as our weight change measure: whether or not workers were receiving wage replacement benefits 365 days after the date the claim was received by DLI. Additionally, we categorized the accumulated days on wage replacement by 1 year after claim receipt (1 – 29, 30 – 89, 90 – 179, 180+ days) to determine if there was a dose-response relationship with clinically significant weight gain.

We first conducted bivariate logistic regression analyses to examine associations between baseline variables of interest in each domain and clinically significant weight change, adjusted by age and gender. Missing, “don't know,” and refusal responses for each variable were combined into one response and included in the analysis. Variables with the most missing data included time from the date of injury to the first medical visit (n=36), region of worker residence (35), paid bill for an opioid prescription within 6 weeks of the first medical visit for the injury (33), recovery expectations (29), source of blame for the injury (26), days of work missed due to non-back health problems in the previous year (23), days of work missed due to back problems in the previous year (18), worker self-report of whether his/her supervisor listens to work-related problems (17), worker self-report of whether the employer had offered job accommodations to allow him/her to work (16), worker self-report of number of previous worker's compensation claims (11), and worker self-report of change in pain since the injury (11).

Next, we created a multivariate logistic regression model predicting clinically significant weight gain (yes/no). We entered as independent variables all baseline variables with P-values < 0.10 in the bivariate analysis, along with age and gender. A standard 0.05 P-value for determining statistical significance of bivariate associations may exclude variables that might be significant in a multivariate model.²⁹ Analyses were conducted with Stata Version 10.³⁰

The sample of workers (N=1,263) was mostly non-Hispanic White (73%; 14% Hispanic; 14% other) and male (69%). At the baseline interview, 29.7% of the study participants were of normal weight (BMI < 25), 40.0% were overweight (25 ≤ BMI ≤ 30), and 30.3% were obese (BMI > 30). At one year, 174 (13.8%) participants self-reported weight that represented clinically significant (7%) weight gain from baseline and 103 (8.2%) participants gained more than 10% of their baseline weight. Sixty-two participants went from normal to overweight status, 66 went from overweight to obese, and 1 participant went from normal weight to obese, for a total of 129 (10.2%) participants with an increase in BMI category by 1 year.

Table 1 shows the variables associated with clinically significant weight gain in the bivariate analyses. Six of 8 domains contained variables associated (P < 0.10) with weight gain. These included female gender (sociodemographics), having a fast-paced work environment prior to injury and not returning to work by the baseline interview (employment-related). The pain and function domain contained one predictor: activity interference due to pain was associated positively with weight gain. Worse current health, aside from injury, was the only predictor of weight gain in the clinical status domain. In the health care domain, weight gain was associated with the specialty of the first health care provider seen for the injury (occupational medicine specialist relative to primary care provider). Three variables were identified in the psychological domain: greater catastrophizing, poorer SF-36⁴⁰ Mental Health scale scores, and lower recovery expectations for the back injury were associated with weight gain. No factors from the administrative/legal or health behavior domains were associated with weight gain. Variables that were not associated with weight gain are listed in the Appendix; these include baseline BMI, injury severity, physical demands at work, fear-avoidance, education, opioid prescription for the injury, and tobacco use status.

Table 2 shows results from the multivariate model that included age and the 9 variables that were associated (P < 0.10) bivariately with clinically significant weight gain. Gender was the only significant predictor of clinically significant weight gain. Women had approximately twice the odds of weight gain, as compared to men (adjusted OR = 2.17, 95% CI 1.54 – 3.07).

Receipt of wage replacement compensation at one year (189 of 1,263 participants) was associated with clinically significant weight gain after adjustment for age and gender (adjusted OR for receipt of wage replacement versus no wage replacement at one year = 2.24, 95% CI 1.51 – 3.33, P < 0.001; Table 3). Almost 25% of participants on wage replacement at 1 year after the injury had clinically significant weight gain, while 13.4% of those not receiving wage replacement compensation at 1 year gained significant weight. In the analysis examining categories of days on wage replacement, adjusting for age and gender, only wage replacement for more than 180 days was associated with clinically significant weight gain, compared to 1 – 29 days (adjusted OR 2.40, 95% 1.63 – 3.53, overall P < 0.001).