An Office Ergonomics Quasi-Experimental Field Study: Interim Results

1
An Office Ergonomics Quasi-Experimental Field
StudyInterim Results
Presentation Department of Industrial
Engineering University of Michigan Ann Arbor,
Michigan Oct. 8, 2002 Please Do Not Cite, Quote
or Reproduce
2
Sources of Support
Steelcase, Inc.
Liberty Mutual, Inc.
HWO
Health And Work Outcomes
The W.E. Upjohn Institute
3
Research Team
Ben Amick III, PhD University of Texas School of
Public Health Kelly DeRango, PhD WE Upjohn
Research Institute Michelle Robertson, PhD
Liberty Mutual Health and Safety Research
Center Anne Moore York University Ted Rooney, RN,
MPH Lianna Bazzani, MS Health and Work
Outcomes Noe Palacios, MS Paul Allie,
MS Steelcase, Inc
4
The Interventions

• Highly-adjustable ergonomic chair
• Benefits come from chair adjustability and design
  features
• State-of-the-art ergonomic training
• Helps users get the most out of their chairs
• Chair Training and Training-only groups equally
  maximize the ergonomics of their overall work
  station

5
The Model of Change
Training
Knowledge
Productivity
Health
Highly Adjustable Chair
Postures Behaviors
Functional Health
Satisfaction
6
Study Timeline
Month -2
Month -1
Month 2
Month 12
Month 6
Month 0

• Baseline Data
• Employee surveys
• Observations
• Measurements
• Productivity
• Performance Data

• Testing Intervention Effectiveness
• Employee surveys
• Observations
• Measurements
• Productivity Performance Data

• Intervention
• Group 1
• Group 2

• Intervention
• Group 3

3 Groups 1. Receives Chair and
Training 2. Receives Training 3. Receives
Training at End
7
Primary Health Outcome Research Hypotheses
Control Group

• Over the work week and the work day...
• participants receiving office ergonomics training
  will have reduced MS symptom growth relative to
  control group
• participants receiving an ergonomic chair as well
  as ergonomic training will have reduced MS
  symptom growth relative to the training-only
  group and control group

Training-Only Group
Chair Training Group
Post-Intervention Symptom Level
Symptom Growth
8
Construction of the Bodily Pain Measure
No Pain or Discomfort
Extremely Severe
BPsc S (row scores) varies from 0 -90
9
Bodily Pain Symptom Data Collection

• At 2 and 1 months pre-intervention, and at 2, 6
  and 12 months post-intervention, participating
  workers were asked to report symptoms 3x per day,
  over the course of 1 business week.
• If a worker completed gt80 of the diaries they
  were not asked to complete a second week.
• 80 was used as a criterion for examining trends
  over the week.
• We include all data completed during this two
  week period.
• Maximum of 30 measures per data collection period
  or a total of 150 measures/individual during the
  study.

10
Summary of Bodily Pain Data
Counts of Individuals Submitting Bodily Pain
Scores
Month of Administration
Counts of Bodily Pain Scores
Month of Administration
11
No MS Symptom Increase Over the Week Only The Day

• In each of the two baseline administrations of
  the diary, symptoms increased over the workday
  but not the workweek.

(The letters B, E, and M refer to the beginning,
middle, and end of the workday, respectively.)
12
Key Design and Statistical Issues

• Because random assignment was not done there is a
  need to statistically adjust for important group
  differences that may explain observed findings.
• Because pain can result from many different
  sources we need to adjust for important
  predictors of pain.
• Need to adjust for multiple measurements of
  bodily pain for the same person.
• Need to develop an efficient statistical test of
  the primary hypothesis.

13
Approach toStatistical Analysis

• Multilevel model constructed with individuals
  specified as level 2 and time points of
  measurement specified as level 1.
• Chair and training groups are compared to the
  control group.
• Test of primary hypothesis involves a three-way
  interaction between group, time of day, and
  phase of the study (pre- or post-intervention).
• Models are conditioned on significant pre- and
  post-intervention group differences and important
  predictors of pain.
• Analysts blinded to groups in all but final
  analysis.

14
Dealing With Confounders and Covariates

• Over 30 variables examined.
• Step 1. Examined covariates for significant pre
  -intervention group differences.
• Step 2. Examined covariates for significant
  post-intervention group differences.
• Step 3. Examined covariates for significant
  prediction of bodily pain.
• Step 4. Examined correlation matrix of
  covariates to identify redundant variables.
• Step 5. Examined confounders satisfying above
  criteria in a multivariate model to assess
  significant contribution to model fit.

15
Final Confounders and Covariates

• Self-reported repetitiveness of work for
  hands/wrists (0-6, 6 more repetitive)
• Self-reported social support from co-workers and
  supervisors (0-8, 8 more support)
• Self-reported freedom of distractions from noise
  in the office (1-6, 6 less distraction)
• Self-reported health (1-5, 5 poor health)
• Job level from administrative data (1-5, 5
  highest level)
• Disability status from administrative data (0-1,
  disability report filed)

16
Sample Exclusion Criteria

• Part-time employees (N11).
• Employees who had filed a workers compensation
  claim within the six-months prior to study
  initiation (N0).
• Employees not sitting in their office chair for
  at least 4 hours per day, on average (N0).

17
Total SamplePre-and-Post Intervention
Total Population Averages
Variables placed in hypotheses-testing model
Note for Time in Office Chair/At Office
Computer 00 hrs. 11-2 hrs. 23-4 hrs. 35-6
hrs. 47-8 hrs. 59 hrs.
18
Total Sample vs. Study Groups
Total Group Population Averages
Variables placed in hypotheses-testing model
Note for Time in Office Chair/At Office
Computer 00 hrs. 11-2 hrs. 23-4 hrs. 35-6
hrs. 47-8 hrs. 59 hrs.
19
Multilevel Model Predictive Variables And Their
Coefficients
20
Chair With Training Reduces Pain Symptom Levels
Over the Day

• Post-intervention, the chair and training group
  had a lower level of symptoms throughout the day
  compared to either the training-only or the
  control groups after adjusting for significant
  group differences and independent predictors of
  pain.

21
Recap

• The lack of symptom growth over the work week was
  an unexpected finding.
• The chair-and-training group showed a marked
  decrease in symptom level and symptom growth
  post-intervention this decrease was
  statistically significant relative to the control
  group.

22
Preliminary Conclusions

• The Leap chair with training significantly
  reduces musculoskeletal symptom growth over the
  work day compared to workers not sitting in a
  Leap chair and workers who did not receive any
  training.
• The Leap chairs effect persists twelve months
  after being introduced.
• If musculoskeletal injuries are the result of
  cumulative traumas, and pain is an indicator of
  acute trauma, the Leap chair should reduce the
  incidence of musculoskeletal injuries in office
  workers
• There is a suggestion of a training effect, but
  the small sample size prevents any conclusions
  from being made at this time.

23
Next Steps

• We will confirm the assumed underlying model of
  any intervention effect by demonstrating reduced
  muscle loading with the ergonomic chair.
• Examine other health outcomes to explore
  consistency of effects for a range of health
  outcomes.
• Include RULA and OEA as pathway effects.

24
Potential Criticisms

• Lack of randomization
• Solved with larger sample and more robust
  statistical control of pre-intervention
  differences.
• Lack of clinically meaningful differences.
• Solved with larger sample and body-site specific
  changes over time associated with the chair
  design. Also addressed with nested studies
  showing changes in muscle loads.
• The control group had worse health at baseline
• We statistically control for pre-intervention
  differences. Additionally, with other sites
  these differences will diminish.

25
Potential CriticismsContinued

• Effects not reproducible across employers
• Solved with additional companies where observe
  same effect.
• You cannot rule out a Hawthorne Effect.
• True, but we can demonstrate the impact of the
  chair on biomechanical load changes.
• Differences due to training not chair (absence of
  chair only group)
• Only solved with fourth group if goal to specify
  chair only effect IS THIS REAL!
• Also biomechanical loads study will demonstrate
  chair effects.

26
The Model of Change
Training
Knowledge
Productivity
Health
Highly Adjustable Chair
Postures Behaviors
Functional Health
Satisfaction
27
Key Question
Are Office Ergonomics Interventions Costly or
Beneficial to Employers?
28
Ergonomics and Productivity

• Rarely studied by economists
• Search on EconLit using Keywords ergonomics
  and productivity yields zero hits

29
Ergonomics and Productivity

• Few studies conducted by health researchers
• Very few use experimental design, mostly case
  studies
• Impact of ergonomics interventions limited to
  lost work days, as opposed to on-the-job
  productivity
• Many studies (e.g., National Academy of
  Sciences, 1999) calculate cost to society of
  musculoskeletal disorders (such as carpal tunnel
  syndrome) using data on disability leave and
  workers compensation claims alone

30
Ergonomics and Productivity

• Field is wide open for a study that
• 1) uses experimental design (control group)
• 2) examines impact of ergonomic interventions
  on productivity while people are still on the
  job, not just lost work days
• what is the cost to employers of low-level,
  chronic pain?
• 3) examines impact of ergonomic interventions
  on people with normal pain levels
• is it cost-effective for employers to reduce
  chronic pain using ergonomic interventions?

31
Study Objectives

• To demonstrate the health, quality of work life,
  and economic impacts of two office ergonomic
  interventions
• Seating (highly adjustable ergonomic chairs)
• Cost 1000/participant
• Office Ergonomics Training
• Cost 200/participant
• Firm is a government tax collection agency

32
SF-36 Pain Score

• Gold Standard in health field, widely accepted as
  a valid instrument
• Two questions
• During the past 4 weeks, how much did pain
  interfere with your normal work?
• How much bodily pain have you had during the
  past 4 weeks?
• National norms are available for this pain measure

33
Intervention Effects Bodily Pain (DHD) and
SF-36 Pain Scores

• Controlling for gender, age, disability status,
  education, job characteristics, and
  pre-intervention pain levels
• The chair intervention reduced
• SF-36 pain by 5.16 points
• DHD pain by 3.11 points
• Training-only intervention reduced
• SF-36 pain by 2.79 points
• DHD by 2.45 points
• significant at 5

34
Modeling Productivity Increases Two Effects
1. People become more efficient
(more output per hour worked) 2. People
can work more hours (less sick leave)
Outcomes of Interest 1. Sick leave hours per
month 2. Collections (in dollars) per effective
work day (8 hours of work) a. Collected tax
revenues is the output of the firm b. Allows
us to model a production function using micro
data, with a health intervention as one input
into production
35
Two Approaches to Modeling Production

• Model total effect on productivity controlling
  for pre-existing differences in production
• Health effect
• Chair Satisfaction
• Hawthorne effect
• Short-term and long-term effects
• Model health effect on productivity using pain
  measures as explanatory variables
• a. Health effect only
• b. Long-term effects

36
Total Effects Model of Productivity Changes
Impact on Daily Production
Chair and Training 354.18 (
134.24) Training only 151.01 (
240.00) Significant at 5 level after
controlling for gender, age, tenure, disability,
education, job type and level and
pre-intervention production levels
37
Health Mediated Model of Productivity Changes
Impact of 1-point Decrease in Pain
SF-36 pain score 19.25/point (
5.73) Significant at 5 level after
controlling for gender, age, tenure, disability,
education, job type and level and
pre-intervention production levels.
38
Percentage Increase in Production
Health Effects
Total Effects

• A. Change in production per day per change in
  19.25 --
• SF-36 pain score
• B. Change in pain score per intervention
  5.48 --
• C. Average total benefit per day (A x B)
  105.49 354.18
• Predicted average daily production,
  1,993.98 1,993.98
• pre-intervention
• E. Percentage increase in production (C/D)
  5.3 17.8

39
Absenteeism

• Various models and specifications
  suggest that both interventions have little or no
  effect on absenteeism.
• point estimate implies lt ½ hour saved per month
  per employee
• results not statistically significant
• Note We have shown a strong productivity effect
  in a firm where the effect on absenteeism is
  essentially zero.

40
Conclusions

• Chair and Training intervention substantially
  reduces pain (both Bodily Pain and SF-36) in test
  subjects relative to the control group.
• The Chair and Training intervention costs
  approximately 1000/employee.
• Estimate of the total productivity gain is
  31.3, or 354 per day per employee.
• The likely health effects productivity gains
  imply a 9.3 increase in productivity, or 105
  per day per employee.
• The benefit flows that result from sick leave
  reductions are essentially zero.