Transcript Workers* compensation claims

WORKERS’ COMPENSATION
CLAIMS: MINING DATA TO
IMPROVE SAFETY OUTCOMES
IN THE GRAIN ELEVATOR
Sai K. Ramaswamy
&
Dr. Gretchen A. Mosher
June 27th 2016
Iowa State University
GRAIN ELEVATOR HAZARDS
• Grain elevator operations involve numerous safety hazards
• Increased storage capacities, larger and faster handling capacities and
increased automation some contributing factors
• Some of the safety hazards include:
•
•
•
•
•
•
Entrapment & crushing
Suffocation
Grain dust explosions
Dealing with heavy machinery
Electrical and fire hazards
Working at heights and confined spaces
Iowa State University
AGRICULTURAL INJURY DATA
• The Survey of Occupational Injuries and Illness (SOII), published annually by
the Bureau of Labor Statistics (BLS) most widely used source
• Other sources include:
• Occupational Injury Surveillance of Production Agriculture (OISPA) survey a NIOSH
and USDA collaboration
• Several state-level surveillance programs such as NY state farm fatality and injury
tracking
• Trauma registries
• Most existing injury data sources focused on production agriculture
• Surveillance of injuries especially non-fatal injuries are less structured and
hence difficult to obtain
• Very little injury information available for agribusiness such as grain elevators
Iowa State University
LIMITATIONS OF SOII DATA
• Takes 3 years to produce SOII for the given year
• Under-reporting of agricultural injuries
• Several states such as Colorado, South and North Dakota
do not participate in the BLS survey
• Uses the North American Industry Classification System
(NAICS) code so direct connections to specific agribusiness
can be difficult
• Injuries of contingent workers are underrepresented in
SOII
Iowa State University
LIMITATIONS OF SOII DATA CONT.
• The SOII collects data from a sample of workplaces and
not all workplace in the U.S
• Data for SOII is collected from OSHA logs maintained
by the employers and hence is prone to data entry errors
• Collected from an entire firm or only for certain facilities
within the firm
• Many groups exempted for example agribusiness with
less than 11 employees
Iowa State University
WORKERS COMPENSATION(WC) CLAIMS
DATA
• Alternative source of data to study work-related
injuries
• Highly preferred in epidemiological studies due to
their size and volume of data
• Used to study work-related injuries in industries such
as construction, mining, saw milling and logging
• Very little application of this data source in the grain
elevator industry
Iowa State University
WC CLAIMS DATA ADVANTAGES
• Mandatory for employers in all states except Texas
• No exclusions , as long as the firm has one hired worker
WC must be provided
• Majority of workers in the U.S. are covered by WC
• Cost efficient as this information is already being
collected by state and insurance companies
• Contains detailed injury information
• Is not a sample of injuries like the SOII
Iowa State University
SOII VS. WC DATA
• Wuellner et al. (2016) study found WC data captured 96% of SOII
data, but SOII only captured 70% of WC data
• Joe et al. (2014) study found about 2/3rd records matched between
SOII data and WC data for amputations and carpal tunnel
syndrome cases
• Friedman et al. (2013) found wide fluctuations in matching %
between SOII and WC data, up to 90% in one particular year
• Boden & Ozonoff, (2008) & Rosenman et al. (2006) compared SOII
and WC claims across six states and found a 25-78% underreporting
in SOII
Iowa State University
DATA FOR THIS STUDY
• The claims dataset in this study was obtained from a Midwest based
insurance company with operations across U.S.
• Nearly 40,000 records of which 8,000 claims from grain elevators and
grain milling facilities
• Claims from 2008 to 2016
• Dataset includes
Demographic information - such as gender, DOB and Hire date
Injury information such as – cause and nature of injury, body part
injured
Cost information such as total cost of claim, medical and indemnity
costs
Iowa State University
DATA ANALYSIS METHODOLOGY
• All variables were treated as categorical variables
• The cost information were categorized as “No cost”, “Less
than $1,000”, “$1,000-$10,000”, “$10,000-$100,000 and
“$100,000+”
• Descriptive analysis included
• Frequency counts and percentages
• Contingency tables
• Chi-square analysis and Classification and Regression Trees
(CART) were used to mine and investigate the relationship
among the variables
Iowa State University
CLASSIFICATION AND REGRESSION TREE
(CART)
• A decision support tool used for building tree-like prediction
models
• If the outcome variable (dependent variable) is categorical then the
model building process is called a classification problem
• If numerical then the model building process is called a regression
problem
• The goal of CART is to identify patterns and build a concise model
of the outcome variables in terms of the predictor variables
• Neural networks, genetic algorithms, Bayesian methods, log-linear
model are all different types of classification tree models
Iowa State University
#
Car
Age Children Subscription
1
Sedan
23
0
Yes
2
Sports
31
1
No
3
Sedan
36
1
No
4
Truck
25
2
No
5
Sports
30
0
No
6
Sedan
36
0
No
7
Sedan
25
0
Yes
8
Truck
36
1
No
9
Sedan
30
2
Yes
10
Sedan
31
1
Yes
11
Sports
25
0
No
12
Sedan
45
1
Yes
13
Sports
23
2
No
14
Truck
45
0
Yes
Age
<=30
>30
Car type
Children
sports, truck
sedan
Yes
No
>0
0
Car type
sedan
Car type
sports, truck
sedan
No
Yes
Yes
Iowa State University
sports, truck
No
REASONS FOR USING CART
• Well-suited for both categorical as well as continuous variables
• To use statistical models such as probit/logit models underlying
assumptions must be satisfied
• No pre-defined or underlying relationship requirements
between the outcome and predictor variables
• No risk of erroneous predictions when assumptions violated
• Widely used in various fields yet few applications on agricultural
injury data till date
Iowa State University
PRELIMINARY RESULTS
• The total cost of claim has a significant association with age and
tenure of the employee
• Total cost of claim also has a significant association with nature of
injury, cause of injury and body part
• Strain injury and slips falls and trips were the two most common
causes of injuries
• Upper extremities such as hands, fingers were the most impacted
body part
• Strain and lacerations were the two most common types of injuries
Iowa State University
PRELIMINARY CART ANALYSIS
RESULTS
Rule#
Splitting Condition
Leading cause of injury
1
Disability (Y/N)
Fall, slips or trip injuries
2
Disability (Y/N) and Tenure ( <=5yrs., 5+yrs.)
Fall, slips or trip injuries ;
Strain injury
3
Disability (Y/N) and Age group ( <=40yrs.,
40+yrs.)
Fall, slips or trip injuries ;
Strain injury
Disability (Y/N) and Nature of injury( specific,
multiple)
Fall, slips or trip injuries for
both types
Disability (Y/N) and Body part group ( Head,
trunk, upper extremities, lower extremities,
multiple)
Heat or cold exposures,
strain, strain, fall, slips or
trip, heat or cold exposures
respectively
4
5
Iowa State University
NEXT STEPS
• Classify the data using several variables simultaneously to
identify patterns that can help safety decision making
• Building multivariate prediction models using CART
• Develop a benchmark for agricultural cooperatives to
use as a barometer for continuous improvement in
worker safety
Iowa State University
CONCLUSIONS
• WC claims data analysis supplements the existing research using
SOII data
• All claims are thoroughly vetted and hence WC data are highly
reliable
• Training employees who are below age 40 years and five or less
years experience can help reduce the number of severe injuries
• Fall protection equipment and better housekeeping can help
eliminate the most common cause of work-related injuries in
the grain elevator
Iowa State University
LIMITATIONS
• Data were from only one insurance company and results cannot
be generalized for entire industry
• WC claims data are also susceptible to underreporting due to:
• Lack of awareness
• Non-serious nature of injury
• Fear of employer retribution
• The research team no control on the data being collected for
each WC claim
Iowa State University
REFERENCES
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
Baker, D., E. (1993). Safe Storage and Handling of Grain. MU Extension Publications. Retrieved from http://extension.missouri.edu/p/G1969
BLS. (2015). Industry Injury and Illness Data. Retrieved from Washington, D.C.: http://www.bls.gov/iif/
Boden, L. I., & Ozonoff, A. (2008). Capture-recapture estimates of nonfatal workplace injuries and illnesses. Annals of Epidemiology, 18(6), 500-506.
doi:10.1016/j.annepidem.2007.11.003
Chang, L.-Y., & Wang, H.-W. (2006). Analysis of traffic injury severity: An application of non-parametric classification tree techniques. Accident Analysis
and Prevention, 38(5), 1019-1027. doi:10.1016/j.aap.2006.04.009
Coleman, P. J., & Kerkering, J. C. (2007). Measuring mining safety with injury statistics: Lost workdays as indicators of risk. Journal of Safety Research, 38(5),
523-533. doi:10.1016/j.jsr.2007.06.005
Dembe, A. E. (2001). The social consequences of occupational injuries and illnesses. American Journal of Industrial Medicine, 40(4), 403-417.
doi:10.1002/ajim.1113
Douphrate, D. I., Rosecrance, J. C., Stallones, L., Reynolds, S. J., & Gilkey, D. P. (2009). Livestock-Handling Injuries in Agriculture: An Analysis of
Colorado Workers' Compensation Data. American Journal of Industrial Medicine, 52(5), 391-407. doi:10.1002/ajim.20686
Douphrate, D. I., Rosecrance, J. C., & Wahl, G. (2006). Workers' compensation experience of Colorado agriculture workers, 2000-2004. American Journal of
Industrial Medicine, 49(11), 900-910. doi:10.1002/ajim.20387
Foley, M., Ruser, J., Shor, G., Shuford, H., & Sygnatur, E. (2014). Contingent workers: Workers' compensation data analysis strategies and limitations.
American Journal of Industrial Medicine, 57(7), 764-775. doi:10.1002/ajim.22302
Friedman, L., Krupczak, C., Brandt-Rauf, S., & Forst, L. (2013). Occupational amputations in Illinois 2000-2007: BLS vs. data linkage of trauma registry,
hospital discharge, workers compensation databases and OSHA citations. Injury-International Journal of the Care of the Injured, 44(5), 667-673.
doi:10.1016/j.injury.2012.01.007
Joe, L., Roisman, R., Beckman, S., Jones, M., Beckman, J., Frederick, M., & Harrison, R. (2014). Using Multiple Data Sets for Public Health Tracking of
Work-Related Injuries and Illnesses in California. American Journal of Industrial Medicine, 57(10), 1110-1119. doi:10.1002/ajim.22361
Jones, T., & Kumar, S. (2004). Occupational injuries and illnesses in the sawmill industry of Alberta. International Journal of Industrial Ergonomics, 33(5), 415427. doi:10.1016/j.ergon.2003.11.002
Khanzode, V. V., Maiti, J., & Ray, P. K. (2012). Occupational injury and accident research: A comprehensive review. Safety Science, 50(5), 1355-1367.
doi:10.1016/j.ssci.2011.12.015
Kramer, J., A. (1989). Safety Measures in Handling Stored Grain. Retrieved from http://www.bae.ksu.edu/docs/farm-safety/AF170.pdf
Lehtola, C. J., Brown, C. M., & Becker, W. J. (2008). Grain Handling Facilities--OSHA Standard 1910.2721. ABE208). Gainesville, FL: Institute of Food and
Agricultural Sciences, University of Florida. Downloaded June, 16, 2008.
Iowa State University
REFERENCES CONT..
•
•
•
•
•
•
•
•
•
•
•
•
Leigh, J. P., Du, J., & McCurdy, S. A. (2014). An estimate of the US government's undercount of nonfatal occupational injuries and illnesses in
agriculture. Annals of Epidemiology, 24(4), 254-259. doi:10.1016/j.annepidem.2014.01.006
Lindahl, C., Lundqvist, P., & Norberg, A. L. (2012). Swedish Dairy Farmers' Perceptions of Animal-Related Injuries. Journal of Agromedicine,
17(4), 364-376. doi:10.1080/1059924x.2012.713839
Mujuru, P., Helmkamp, J. C., Mutambudzi, M., Hu, W., & Bell, J. L. (2009). Evaluating the impact of an intervention to reduce injuries among
loggers in West Virginia, 1999-2007. Journal of agricultural safety and health, 15(1), 75-88.
Nicholson, A., & Wong, Y. D. (1993). Are accidents Poisson distributed? – A statistical test. Accident Analysis and Prevention, 25(1), 91-97.
doi:10.1016/0001-4575(93)90100-b
Oleinick, A., & Zaidman, B. (2004). Methodologic issues in the use of workers' compensation databases for the study of work injuries with
days away from work. I. Sensitivity of case ascertainment. American Journal of Industrial Medicine, 45(3), 260-274. doi:10.1002/ajim.10333
Pasman, H. J. (2009). Learning from the past and knowledge management: Are we making progress? Journal of Loss Prevention in the Process
Industries, 22(6), 672-679. doi:10.1016/j.jlp.2008.07.010
Rosenman, K. D., Kalush, A., Reilly, M. J., Gardiner, J. C., Reeves, M., & Luo, Z. W. (2006). How much work-related injury and illness is missed
by the current national surveillance system? Journal of Occupational and Environmental Medicine, 48(4), 357-365.
doi:10.1097/01.jom.0000205864.81970.63
Schwatka, N. V., Butler, L. M., & Rosecrance, J. C. (2013). Age in relation to worker compensation costs in the construction industry. American
Journal of Industrial Medicine, 56(3), 356-366. doi:10.1002/ajim.22093
Utterback, D., Meyers, A., & Wurzelbacher, S. (2014). Workers' Compensation Insurance: A Primer for Public Health. Department of Health and
Human Services: National Institute for Occupational Safety and Health.
Wiatrowski, W. J. (2014). The BLS Survey of Occupational Injuries and Illnesses: A Primer. American Journal of Industrial Medicine, 57(10), 10851089. doi:10.1002/ajim.22312
Wuellner, S. E., Adams, D. A., & Bonauto, D. K. (2016). Unreported workers' compensation claims to the BLS Survey of Occupational
Injuries and Illnesses: Establishment factors. American Journal of Industrial Medicine, 59(4), 274-289. doi:10.1002/ajim.22563
Zhou, C., & Roseman, J. M. (1994). Agricultural injuries among a population based sample of farm operators in Alabama. American Journal of
Industrial Medicine, 25(3), 385-402. doi:10.1002/ajim.4700250307
Iowa State University