Construction site

Members

Matthew Hallowell

CU Boulder

Mike Quashne

BGE

Rico Salas

UC Boulder

Matt Jones

UC Boulder

Brad MacLean

Wolfcreek Group

Ellen Quinn

Otis Elevator

Jenny Bailey

Xcel Energy

Bruno Cagnart

TechnipFMC

Len Colvin

TVA

Matt Compher

Quanta Services

Mike Court

Graham Group Ltd.

Nick DiMartino

MasTec

Ester Brawley

CRC

Phillip Hodge

Chevron

Greg Kelly

Enbridge

Gregg Slintak

Consolidated Edison Comp.

The Tyranny of TRIR

Total recordable injury rate (TRIR) has been used as the primary measure safety of performance for nearly 50 years. Since organizations conform to the same definition, the TRIR metric has been used to compare industries, sectors, companies, and even projects. TRIR is used in many ways and dominates discussions about safety performance from the worksite to the board room.

With over 3.2 trillion worker-hours of data supplied by the Construction Safety Research Alliance members, we explored the question: given how it is used, to what extent is TRIR a statistically valid measure of safety performance? Our team of researchers analyzed the data using generalized linear modeling, Monte Carlo simulations, and parametric analyses, which yielded the following conclusions:

  1. There is no discernible association between TRIR and fatalities;

  2. The occurrence of recordable injuries is almost entirely random;

  3. TRIR is not precise and should not be communicated to multiple decimal points of precision; and

  4. In nearly every practical circumstance, it is statistically invalid to use TRIR to compare companies, business units, projects, or teams.

Besides the myriad of reasons why it is flawed (underreporting, case management, etc.), the work conclusively shows that TRIR is statistically meaningless given the way it is currently used. For us to transform safety into an evidence-based discipline, we need to scientifically dismantle such unsubstantiated institutions. With TRIR debunked, we are addressing the natural next question, “If not TRIR, then what?”

Reference the final published results here: