Using Data Mining to Predict Injuries:

Transcription

1 Using Data Mining to Predict Injuries: A new model for injury prediction and prevention 2054 Steve Thompson, ARM, COSS sthompson@aspenrmg.com Aspen Risk Management Group ERGOhealthy ASSE Member since 1989 Prediction The way things will happen in the future Large corporations invest heavily on possible events, risks and business opportunities Prediction in business "... is at least two things important and hard 2

2 Supernatural prediction Paranormal or supernatural Astrology and occult sciences Not substantiated by controlled experiments One in 10 people in the world are concerned that the world could end in 2012 (Reuters poll) 3 Paranormal or supernatural Astrology and occult sciences Not substantiated by controlled experiments One in 10 people in the world are concerned that the world could end in 2012 (Reuters poll) 4

3 Temple of Apollo, Delphi 5 Priestess of Delphi 6

4 Precog 7 Nostradamus 16 th =century French seer Studied astrology and occult sciences Best known for The Prophecies

5 Nostradamus predictions Louis Pasteur Nostradamus predictions Atomic bomb

6 Nostradamus predictions Hitler and World War II Nostradamus predictions Sept. 11, 2011

7 Nostradamus predictions The end of the world Edward Cayce (the Sleeping Prophet ) Two World Wars (start and end dates) The economic crisis in 1929 Fall of communism and the collapse of the Soviet Union Revival of Soviet Union to begin in 2010 Growth of political power in China Global warming 14

8 Scientific prediction In science a prediction is rigorous and quantitative Computer models are frequently used to describe the past and future behavior of something Accurate workplace injury prediction and forecasting is very difficult In insurance an actuary uses a life table to predict life expectancy Life table

9 Scientific prediction Neptune was predicted by Sir Isaac Newton Radio waves were predicted by James Clerk Maxwell Anti-matter was predicted by Paul Dirac Old Farmer s Almanac Injury prediction studies Early prognostic factors for duration on temporary total benefits in the first year among workers with compensated occupational soft tissue injuries Chronic musculoskeletal pain predicted hospitalization due to serious medical conditions in a 10 year follow up study

10 Injury prediction studies Prediction of chronic disability in work-related musculoskeletal disorders: a prospective, population-based study Early predictors of chronic work disability: a prospective, population-based study of workers with back injuries (ISSLS prize winner: Injury prediction studies Listening to injured workers: how recovery expectations predict outcomes--a prospective study Worker recovery expectations and fear-avoidance predict work disability in a population-based workers' compensation back pain sample

11 Injury prediction studies Eight predictive factors associated with response patterns during physiotherapy for soft tissue shoulder disorders were identified High pain ratings predict treatment failure in chronic occupational musculoskeletal disorders Injury prediction studies Biopsychosocial multivariate predictive model of occupational low back disability

12 Injury prediction studies Life values as predictors of pain, disability and sick leave among Swedish registered nurses: a longitudinal study Injury prediction studies A model for occupational injury risk assessment of musculoskeletal discomfort and their frequencies in computer users

13 Injury prediction studies Symptoms of stress predict musculoskeletal disorders Prediction today Baby Gender Future Height Tornado, Hurricane, Weather Earnings of a company (earnings per share) Your Future Predict Your Social Security Payout How to predict Mitt Romney s running mate Computer predicts sports injuries (experimental)

14 Common safety and health predictors Gas detection devices Health risk survey Safety culture, and other safety and ergo surveys Severity index (predictive analytics in workers comp) Underwriters use predictive modeling to price risk Peer behavior, and environmental observations Pre and new hire orientation Casualty prediction Strains, sprains, and discomfort 40% of WC costs = musculoskeletal disorders Cisco and Accenture pilot studies (10,000 EEs) Persistent discomfort - pulled off task 5 minutes out every 15 minutes Lost time/day/employee: 2.6 hours Lost time/week/employee: 13 hours 62% decrease in reported discomfort 39% decrease in claims for ergonomic injuries

15 Minutes add up 13.3 hours/week x 4 weeks = 53.3 hours of lost production/employee/month Cost of lost production for 1 employee: $15/hr x 53.3 hours = $799.50/month $25/hr x 53.3 hours = $1,332.50/month $50/hr x 53.3 hours = $2,665/month *(20 min. per hour) x (8 hours) x (5 days) x (4 weeks) Pain and discomfort predicts productivity loss Injury risk assessment why not add Safety inspections Safety audits Incident investigations Near misses Review of preventative maintenance records and on and on...

16 Why not Fear of what s lurking beneath the surface Fear of cost to fix Fear What to do with the information What bucket does it fit in Now what No action, or poor planning post data mining Would you Introduce process to determine culture and risk levels Explore with employees discomfort and risk levels to improve productivity and predict injuries Data-mine existing records health, safety, workers compensation Be prepared to do something about your findings

17 If you were wondering Scientific date for the destruction of the planet Unknown Precisely because of human fallibility, extraordinary claims require extraordinary evidence ~ Carl Sagan Using Data Mining to Predict Injuries: A new model for injury prediction and prevention Steve Thompson, ARM, COSS 2054 sthompson@aspenrmg.com Aspen Risk Management Group ERGOhealthy Summary and QA ASSE Member since 1989