My Class

1

IE 552 Mechanics of Musculoskeletal System Dr.
Andris Freivalds Class 33
2
Reliability of Tool - 1

• Consistency or stability of measurement over time
• Coefficient of variability
• CV s/x
• Test/retest reliability Correlation coefficient
  

3
Reliability of Tool - 2

• Split-half procedure (Spearman-Brown prophecy)
  cant repeat, sensitization
• rtot 2r½/(1 r½)
• Kuder-Richardson internal consistency
• r n/(n-1) 1 x(1-x/n)/s2
• Cronbachs alpha variation of 4
• a n/(n-1) 1 - ?si2/s2

4
Reliability of Analysts (2 analysts)(Inter-rater
agreement)

• Proportion of agreement
• p0 a d
• Kappa statistic
• ? (p0 pc)/(1-pc)

gt0.8 v. good 0.6-0.8 good 0.4-0.6 fair
lt0.4 poor
5
Reliability of Analysts ( 3 analysts)

• Intra class correlations (ICC - Case 1) each
  item analyzed by a different analyst

6
Reliability of Analysts ( 3 analysts)

• Intra class correlations (ICC Case 2) each item
  analyzed by same analysts

7
Summary of Sources of Variation and Appropriate
Procedures and Tests
8
Accuracy and Precision
Accuracy small deviations from true
value Precision small variations about mean
9
Signal-Detection Theory
10
Classification Table for Test Results and the
Truth
11
Generalizations on Self Reports

• Basic survey data (job tile, work history, job
  duration) ? ? 0.6 0.7
• Physical activity (general) ? r 0.7 - 0.9
• Physical activity (detailed postures vs.
  videotape analysis) ? r 0.1 0.3
• Symptom surveys (retest) ? r 0.9
• Symptom surveys (vs. NCV) ? r 0.4 - 0.5
• Tendency to overestimate exposure

12
Generalizations on Analysts

• Physical activity ? ICC 0.2-0.7
• Ratings of postures ? ? 0.3
• Postures in angles ? ? 0.0 - 0.5
• Force (retest) ? ? 0.8
• Force vs. direct ? ? 0.5
• Training effect, memorize sequences
• Greater variability with ? complexity

13
Self Reports gt Analyst Rating
14
Initial Identification of Problem -1

• Medical data (OSHA 300 logs) analysis
• Calculate incidence (or severity) rates (IR)
• IR injuries x 200,000
• employee hours
• Compare to industry averages
• Meatpacking 7.1
• Auto assembly 6.9
• Mens slacks 4.5

15
Initial Identification of Problem -2
Pareto analysis 80/20 rule
16
Initial Identification of Problem - 3

• Chi-square analysis
• Find critical one of m departments
• ?2m-1 Smi1 (ei-oi)2 ei hi x oT
• ei hT
• ei expected injuries for area i
• hi hours worked for area i
• hT total hours worked
• oi observed injuries for area i
• oT total injuries

17
Initial Identification of Problem - 4

• Example of Chi-square analysis
• e1 900/2,900 x 36 11.2

18
Initial Identification of Problem - 5

• Chi-square analysis significance
• ?2 (11.2-22)2/11.2 (7.4-4)2/7.4
• (17.4-10)2/17.4 15.1
• ?2 at a0.05 for 2 df yields 5.99
• Since 15.1 gt 5.99, Area 1 is significantly higher
  in injuries, needs to be studied further

19
Percentages of the ?2 distribution
20
Initial Identification of Problem - 6Symptom
Surveys (OSHA, 1990)
Symptoms Survey
Date
______________ Department ________________________
___ Job Title _______________________________ Num
ber of years on this job ______________ Number
of years on previous job ___________ 1. Have you
had any pain and discomfort during the last year?
Yes No (Circle answer) If yes, check area
Neck ___ Shoulder ___ Elbow/forearm ___
Hand/wrist ___ Fingers ___ Upper back ___ Low
back ___ Thigh/knee ___ Lower leg ___ Ankle/feet
___ 2. When did you first notice the problem?
_________________________________________ 3. How
long does each episode last? (Mark an X along the
dotted line) 1 hour 1 day.1
week.1 month..6 months 4. What do
you think caused the problem? ____________________
_____________________ 5. How would you rate this
problem? (Mark an X along the dotted line)
None Unbearable 6.
How much time have you lost in the last year
because of this problem? _______________ 7. How
many days in the last year were you on light duty
because of this problem? ________
21
Initial Identification of Problem - 7Body
discomfort maps (Corlett,1976) Category ratio
scales (Borg, 1982)
22
Initial Identification of Problem -8Nordic
Questionnaire (Kuorinka, 1987)
23
Comments on Surveys

• Should interpret with caution
• Workers hesitant to be truthful
• Anonymity may be critical
• Pain tolerances vary
• Proper design, validation critical
• I prefer objective data