Skip to Content
View site list


Pre-Bid Projects

Pre-Bid Projects

Click here to see Canada’s most comprehensive listing of projects in conceptual and planning stages

OH&S, Technology, US News

ISN's machine learning model crunches safety data

Russell Hixson
ISN's machine learning model crunches safety data
ISN — ISN, a global contractor and supplier information management company, recently released the latest version of its white paper on serious incidents and fatalities. The group developed a machine learning model to pore over tens of thousands of incident reports to help predict when and how incidents happen.

ISN, a global contractor and supplier information management company based in Texas, is using machine learning to predict how and when serious injury and fatality incidents (SIF) occur at work.

ISN’s Data Science Team developed a novel deep learning text classification model to classify contractor incidents. The model was trained on the severe injury dataset from the Occupational Safety and Health Association (OSHA) and incorporates input from ISN’s HSE professionals to ensure accurate classification.

The model then analyzed over 60,000 recordable incidents to predict which incidents reported by ISN contractors over the past three years were likely to have been serious injuries or fatalities based on the incident description and outcome.

“Lots of these forms are handwritten so there was a lot of cleansing of the data, looking for common words or things that help relate to an incident,” explained Rick Dorsett, ISN’s director of HSE review and verification. “We looked at all that data and started pulling info off the OSHA logs to look for trends.”

Director of HSE Duane Duhamel added that the data offered a unique opportunity.

“These OSHA 300 logs are consistent in nature, the same template for all organizations and they are an annual exercise,” he said. “It’s really a plethora of unstructured data. The fact that ISN was forward thinking enough to build some machine learning to scrape and scrounge through it to make it structured so it can help drive action is very profound.”

The analysis found more than 18,181 SIF cases, 16,775 hospitalizations, 1,304 amputations and 537 fatalities. ISN’s analysis found while the total number of SIFs decreased in 2020, the rate of their occurrence increased.

“The decrease in the number of SIFs may be, in part, due to the COVID-19 pandemic,” stated the report. “As the world grappled with the COVID-19 pandemic in 2020, work for many organizations slowed thus decreasing the chance for work-related injuries and illnesses. However, the increased rate highlights there is still work to be done when it comes to SIF prevention.”

By using the incident description provided on the contractors’ submitted OSHA logs, ISN’s analysis found the most common injuries sustained in a SIF event were sprains, strains and tears in the lower extremities, and were most often caused by contact with an object or equipment.

ISN noted these results are almost identical to what was reported in their 2020 SIFs White Paper, with the exception that the lower extremities were affected more often in 2020 with the upper extremities being affected more in 2019. ISN also found a higher number of SIFs during the summer months, particularly July and August.

“This trend may be a result of extended hours of sunlight in the United States that permit longer working hours that could result in fatigue among workers,” reads the report. “Organizations should consider this in their risk assessments during these times of the year and implement additional safeguards, such as a fatigue management program, to mitigate future events.”

When looking at all SIFs that occurred in 2020, and also separating out fatalities specifically, almost half occurred with contractors who reported they perform work in construction. The leading type of events that caused a fatality in 2020 for those performing work in the construction industry was contact with an object or equipment, followed closely by falls, slips and trips.

Duhamel noted that one surprising trend the data showed was that while serious injuries have been significantly trending down, fatalities have plateaued and even gone up in some years.

“There is not a direct correlation between SIF reduction and fatality reduction,” he said, noting that many forward-thinking companies have been implementing parallel strategies to deal with both issues.

The analysis also looked what causes lower or higher SIF rates.

ISN found operating commercial vehicles, recent citations, bringing chemicals to a site, working with rigging and working higher than six feet off the ground all significantly increased incident rates. Submitting programs for a safe return to work, working in a Process Safety Management facility, having electrical safety programs or environmental programs all cut down SIF rates.

ISN believes the analysis it is gathering has value to help contractors prevent SIFs but wants to continue expanding its study.

The authors added there are valuable insights to be gathered from near misses and incidents that may not have resulted in an actual SIF but had the potential to do so.

Duhamel said ISN plans to include insights around these “potential SIFs” in the next iteration of its SIF white paper. He added that creating a safe space where employees aren’t afraid to report incidents and are supported by leadership is critical to the future of health and safety in the workplace.

ISN said it plans to include insights around these “potential SIFs” in the next iteration of its SIF white paper.


Follow the author on Twitter @RussellReports.

Recent Comments

comments for this post are closed

You might also like