Development of a Risk Assessment Model

1
Development of a Risk Assessment Model for
Carpal Tunnel Syndrome (CTS)
Heecheon You , Andris Freivalds. Ph.D. ,
Zachary Simmons, M.D. , Milind J. Kothari, D.O.
, and Sanjiv H. Naidu, D.O. Department of
Industrial Manufacturing Engineering Wichita
State University Department of Industrial
Manufacturing Engineering Division of
Neurology Orthopedics Rehabilitation The
Milton S. Hershey Medical Center The Pennsylvania
State University
2
Summary
This study developed - a risk assessment
instrument and - a risk assessment
model directed toward CTS including occupational
and non-occupational risk factors. The proposed
risk assessment system may contribute to reducing
the incidence of CTS at work by providing
acceptable exposure information.
3
Agenda

• Introduction
• - Carpal Tunnel Syndrome
• - Problem Statement
• - Objectives
• Study Design Materials
• - Case-Control Design
• - Risk Assessment Instrument
• Model Development
• Conclusions
• Discussion

4
Introduction
Carpal Tunnel Syndrome
Peripheral neuropathy due to localized
compression to the median nerve within the carpal
tunnel at the wrist.
5
Introduction
Limitations of Previous CTS Research

• Incomprehensiveness Study a partial set of CTS
  risk factors.
• ? Not sufficient to understand relative
  contributions of risk factors to the development
  of CTS.
• Difference in research methodology
• Qualitative findings

6
Problem Statement
Limitations of Previous CTS Research (contd)

• Differences in research protocol

• Study population
• Case definition criteria
• Exposure measurement methods
• ? Make the study results difficult to be compared
  and integrated.

• Qualitative findings

(e.g.) females and/or heavy individuals are more
susceptible to CTS Awkward postures, excessive
grip forces, and/or repetitive hand/wrist motions
increase the risk of CTS. ? Need research for
quantitative models explaining relationship
between exposure and CTS development.
7
Problem Statement
CTS Risk Assessment Models

• Few CTS risk assessment models

• Strain Index model for distal upper extremity
  disorders (Moore and Garg, 1995)
• ? Theory-based, incomprehensive model, NOT
  validated.
• Fuzzy linguistic model for CTS (McCauley-Bell and
  Crumpton, 1997)
• ? Expert-judgement-based, comprehensive model,
  NOT validated.
• Discriminant model for CTS among female VDT
  operators (Matias et al., 1998)
• ? Empirical data-based, gender- and task-specific
  model, cross-validated

8
Objectives

• Develop a risk assessment instrument Survey
  reliable risk exposure information in an
  individual within a reasonable time (1 to 1.5
  hrs/subject) in a retrospective manner.
• Examine relative contributions of occupational
  and non-occupational risk factors to the
  development of CTS Contrast risk exposures of
  case group with those of control group.
• Develop a risk assessment model for CTS Estimate
  the likelihood of developing CTS for an
  individual exposed to certain occupational risks.

9
Study Design
Study Design

• Case-Control Design
• 2 case groups
• - 22 work-related CTS patients (W-CTS),
• - 25 non-work related CTS patients (NW-CTS).
• - Classification type of insurance covering
  their
• medical costs (workers compensation
  insurance for
• W-CTS other healthy insurance for
  NW-CTS).
• 1 control group 50 healthy workers.
• Selection Criteria
• Symptomatic patients diagnosed with CTS.
• No CTS symptom history for healthy workers.
• Working at the current job for at least one year.

10
Study Design
Hypotheses

• Assume additivity of the effects of risk
  exposures to the development of CTS.
• Depending on the type of CTS (W-CTS, NW-CTS),
  relative contributions of occupational and
  non-occupational factors to the CTS risk are
  different.

11
Study Design
Hypothetical Features of Study Groups

• Personal susceptibility

• NW-CTS high
• W-CTS moderate
• Healthy low

• Occupational exposure

• W-CTS high ? moderate
• Healthy NW-CTS moderate ? low

• Contrast the distinctive features of the case and
  control groups to identify relative contributions
  of occupational and non-occupational factors to
  CTS.

12
Problem Statement
Ideal Features of Risk Assessment System

• Practicality Practical to use.
• Specificity Specific to injury and/or task.
• Comprehensiveness important risk factors
  Included.
• Use of data Based on observed or measured data.
• Exposure measurement Reliable and accurate
  measurements.
• Quantitativeness Quantitative model for
  exposure-event or exposure-severity relationship.
• Validation Agreed with previous findings and
  cross-validated.

13
Relationship between CTS Scales
Patient Recruitment

• Patients diagnosed with unilateral or bilateral
  CTS at EMG lab, Hershey Medical Center
  participated in the study immediately after their
  nerve conduction studies.
• Selection Criteria
• Clinical symptoms in one or both upper
  extremities,
• No surgery for CTS on the involved limb(s),
• Age ? 18 years,
• Currently employed,
• Working at the current job for at least one year.

14
Relationship between CTS Scales
Participant Composition

• 64 hands with CTS from 45 patients
• Gender 11 males, 34 females.
• Age average 46.7 years (s.d. 10.2, R 24 to
  65).
• Body mass index (BMI) average 30.1 (s.d.
  6.4, R 19.0 to 46.9) obese level BMI gt 30.0
  (Werner et al., 1994).

15
Relationship between CTS Scales
Participant Composition (contd)

• Comparison of individual characteristics of the
  participants to those of 149 patients with CTS
  for the year 1997 diagnosed at the EMG lab.
• Gender ?2(1) 0.56, p 0.46.
• Age t (73) -0.32, p 0.75.
• Body mass index (BMI) t (69) -0.36, p 0.72.
• No significant difference at ? 0.05.

16
Risk Assessment Questionnaire
Risk Assessment Questionnaire Development

• Survey risk factors (45) associated with CTS and
  corresponding metrics.
• Contents of the survey instrument for CTS
• Survey of personal attributes,
• Assessment of psychosocial work stress,
• Assessment of physical job exposures.

17
Risk Assessment Questionnaire
Risk Scale Definition

• From the 45 risk factors, 106 risk exposure
  scales were defined
• Personal risk scales (factors) 63 (29)
• (e.g.) smoking (1) smoking status
  (never/exsmoker/current smoker),
• (2) smoking experience (no/yes),
• (3) smoking history during last 5 years
  (no/yes),
• (4) current status of smoking (no/yes),
• (5) years of smoking (never
  smoked/1-10/11-20/...),
• (6) years of smoking (years),
• (7) smoking level (never smoked/1-10/..
  cigarettes/day),
• Psychosocial risk scales (factors) 7 (7)
• Physical risk scales (factors) 36 (9)
• (Note) Physical risk exposures of each of the
  left and right hands/ wrists were reshuffled for
  the dominant and non-dominant hands/wrists.

18
Risk Assessment Model
Model Development Algorithm
19
Risk Assessment Model
Pseudo-Univariate Logistic Regression

• Screen candidate variables for CTS model
• Conducted multiple logistic regression for each
  of the 98 reliable risk scales including age,
  gender, and age?gender to eliminate possible
  confounding effects of the two stratification
  variables.
• Screening criteria (1) p lt .25 (Afifi and Clark,
  1990), and (2) estimated OR agree with previous
  findings.
• Screened 27 scales for W-CTS/Healthy
• 21 scales for
  NW-CTS/Healthy
• 24 scales for CTS/Healthy.

20
Risk Assessment Model
Multiple Logistic Regression

• Multiple logistic regression with the candidate
  variables screened for W-CTS/Healthy,
  NW-CTS/Healthy, and CTS/Healthy
• variable selection forward stepwise algorithm.
• test statistic Wald statistic.
• pE .15, pR .20.
• Model Adequacy Checking
• Hosmer-Lemeshow statistic for the goodness-of-fit
  test of each model.
• All the logistic regression models are
  appropriate at ? .05.

21
Risk Assessment Model
Summary of the Models
Note. Bolded are risk scales of which R gt .1
22
Risk Assessment Model
Risk Assessment Models

• While W-CTS/Healthy and CTS/Healthy include
  occupational and non-occupational factors,
  NW-CTS/Healthy does only non-occupational
  factors.
• Support the research hypothesis
• NW-CTS attribute high personal susceptibility.
• W-CTS attribute high physical exposure or
  combined contribution of personal susceptibility
  and physical exposure.
• Imply the necessity of use of rigorous case
  selection criteria in terms of work-relatedness
  in CTS research.

23
Risk Assessment Model
Classification Protocol

• Risk prediction model
• multiple logistic regression
• likelihood of belonging to case group
• If p ? pC, the individual would be classified
  into the case group.
• If not, the individual would be classified into
  the control group.

, where Xi risk scale i.
24
Risk Assessment Model
Determination of pC

• Determine pC based on classification performance.
• sensitivity correct case classification,
  p(case/case)
• specificity correct control classification,
  p(control/control)
• Depending on the location of pC, the sensitivity
  and the specificity of a model varies each other
  in an opposite direction.

25
Risk Assessment Model
Determination of pC (contd)

• Select a value for pC that maximizes both
  sensitivity and specificity in an equal manner.

26
Risk Assessment Model
ROC Curve Analysis

• Construct the ROC curve of each model and compute
  its detectability (d).
• d 0 poor performance d 2.33 excellent
  performance (Proctor and Van Zandt, 1994).
• d 2.51 for W-CTS/Healthy, 2.02 for
  NW-CTS/Healthy, 2.31 for CTS/Healthy.

27
Risk Assessment Model
Model Cross-Validation

• Use the jack-knife method

28
Risk Assessment Model
Cross-Validation Results

• Classification performance variation
• sensitivity decreased.
• specificity increased.
• Overall performance about the same.

29
Risk Assessment Model
Conclusions

• Identified relatively important risk factors to
  be controlled to protect workers from CTS.
• W-CTS gender, wrist ratio, history of
  musculoskeletal disorders at the hands/wrist
  during last 5 years, use of heavy pinch grip
  force, and highly repetitive motions of the
  dominant hand/wrist.
• NW-CTS age, gender, hard driving and competitive
  personality, wrist ratio.
• Developed a valid risk assessment model for
  W-CTS, NW-CTS, and pooled CTS, respectively.
• classification accuracy 84 to 89.
• detectability 2.02 to 2.51.
• cross-validation within ?.25 of classification
  accuracy.

30
Risk Assessment Model
Discussion

• The risk assessment models provide supporting
  evidence of the research hypotheses regarding
  different pattern of contribution of occupational
  and non-occupational risk factors depending on
  work-relatedness of CTS.
• Potential use of the risk assessment models
• Individualized exposure limits Determine
  acceptable exposure limits as a function of
  individual attributes.
• Proper worker placement Avoid a placement of CTS
  susceptible individuals to hand intensive tasks.
• Strategic job improvement plan Prioritize job
  improvement actions based on relative
  contribution of risk factors.