Predicting Safe Outcomes

A new approach to accident prevention goes beyond the traditional principles.

Anyone involved in safety is familiar with Heinrich’s Law: For every 330 incidents, there will be 300 with no injuries, 29 minor injuries and one major injury. While the accuracy of his theory is questioned, the principle seems sound enough — minor incidents lead to more serious ones.

Heinrich did his work in the 1930s using data from the ‘20s. That’s nearly a century ago, long before we had the power to crunch “big data” to reveal unseen patterns and trends. “In the last 20 years the number of injuries in the U.S. has steadily declined,” says Griffin Schultz, general manager of Predictive Solutions in Pittsburgh. “Fatalities and significant injuries have flatlined.”

He adds that companies are finding the theory that smaller injuries predict larger ones isn’t necessarily true. “They’re also finding that precursors to significant injuries and fatalities are different than precursors to less severe injuries.”

Predictive Solutions, a global safety and predictive analytics company, uses safety observations rather than actual incidents to predict the future. “Injuries are on the far right of the spectrum. That’s what happens at the end; a wrench falls off scaffolding and hits someone in the head. A near miss is one step to the left; a wrench falls and doesn’t hit anyone.”

That is the typical scope of safety analysis based on Heinrich’s Law and leads to answers about what, where and how incidents happen. “Those are lagging indicators — looking backward,” he says.

“Further to the left are safety observations,” he says. “You see a wrench on scaffolding and there’s no toe rail to prevent it from falling. That’s the data we use to predict injuries.”

Predictive analytics, used for years in sales and marketing, finance, maintenance and other business functions, looks ahead to answer questions such as why things are happening and what’s going to happen next; moving from lagging indicators to leading indicators.

The company analyzes data from the entire base of customers using its SafetyNet software, including some leading gas and oil companies. Its 170 million safety observations are growing by 3 million per month.

Using computer analytics, the observations predict workplace injuries at an accuracy rate of between 80 and 97 percent, determined by comparing actual incidents to the number predicted. The company cites these examples of how analytics have helped companies:

  • A Fortune 150 manufacturer reduced its lost work day rate by 97 percent within one year.
  • A Fortune 150 energy company reduced its incident rate by 67 percent within 18 months.
  • A top 20 construction company achieved significant safety improvements including 90 percent of work sites experiencing no lost-time incidents.
  • A top 10 specialty contractor reduced its workers’ compensation fees by 57 percent and 66 percent in two years while the number of hours worked increased.

“It’s no different than quality,” adds Schultz. “Testing a product at the end of the assembly line is one measurement. You want to catch the defect in the steps of the process, not at the end. It’s the same thing in safety. You want to observe every day to make sure the toe rail is installed so that wrench never has the opportunity to fall off the scaffolding.”

The Four Truths Of Safety

Looking at four years of safety data while building its analytical tool, Predictive Solutions found that those projects with the fewest safety incidents had several things in common. Two are what safety professionals have known for many years:

  • Too many unsafe observations predicts an unsafe work site: Commit time and resources to fixing observations before they cause incidents.
  • More inspections predict a safer work site: Make sure your program rewards high levels of inspections.

Predictive Solutions also found two that are more counterintuitive:

  • More inspectors, specifically more inspectors outside the safety function, predict a safer work site: Include as many people as you can in your safety inspection program and ideally, have more nonsafety than safety people doing inspections.
  • Too many “100 percent safe” inspections predict an unsafe work site: Train for and reward the reporting of unsafe observations.

From the white paper Predictive Analytics in Workplace Safety: Four ‘Safety Truths’ That Reduce Workplace Injuries

www.predictivesolutions.com



Discussion

Comments on this site are submitted by users and are not endorsed by nor do they reflect the views or opinions of COLE Publishing, Inc. Comments are moderated before being posted.