The Quest for Leading Indicators – The Holy Grail of Safety

In the last few blog posts I’ve written, I have talked about some of the problems we observe with data collection, safety fatigue and engagement. All of these topics relate and culminate in a topic I am passionate about – leading indicators in safety.

Leading indicators are the holy grail in safety. But what are they and how do we get to them?

Perhaps it is important to talk about what leading indicators are not rather than just what they are.

Many individuals mistake a proactive safety activity, such as workplace observations, hazard IDs, toolbox talks etc. as being a “leading indicator”. Leading indicators are not ‘wearing PPE’ and ‘having a safety officer’.  These things in and of themselves are not leading indicators.

Nor are other activities such as training or certification. 

Let us contrast two different definitions.

From Investopedia:

An economic leading indicator is a piece of economic data that corresponds with a future movement or change in some phenomenon of interest. Economic leading indicators can help to predict and forecast future events and trends in business, markets, and the economy.

From a provincial OHS in Canada:

Leading indicators are aspects of workplace activities that can be used to improve OHS outcomes prior to an unwanted outcome occurring – A preventative approach to reducing the risk of workplace injury and illness.

If we look at the difference it is fairly obvious that OHS has used a definition that jumps to conclusions rather than use what a statistical definition of a leading indicator is.

As an example, putting winter tires on your car is a safety activity not a leading indicator. But, if sales of winter tires begin to trend lower, it might be a leading indicator that winter accidents will begin to rise.

Leading indicators must be established specifically to their audience through statistical work that demonstrates a high level of certainty in the correlation and hopeful causation. Even if frequency of toolbox talks or a certification standard is established statistically in one sector, it does not mean it is automatically a leading indicator in another.

“A leading indicator in safety is a piece of data (anything) that corresponds with and can help to predict and forecast future change in risk and unsafe events.“

The result of all of this misunderstanding has been that there are many activities and measures companies follow, thought to be leading indicators because they seem to correlate to safer practices, that are likely ineffectual and are definitely not substantiated.

Leading indicators must be something measurable.

The definition should be: A leading indicator in safety is a piece of data (anything) that corresponds with and can help to predict and forecast future change in risk and unsafe events.

To be a leading indicator, the measure of an attribute, behavior, activity or practice must be able to indicate increasing, or really just changing risk. It is the measurable change, that correlates to risk, which we are looking for to help us predict safety. For instance, the decrease in a positive safety activity could be a leading indicator – the activity itself is not. The safety community has mashed positive/proactive safety activities and leading indicators together.

Leading indicators are found in the domain of large data sets or broad experience (which is a kind of data). That data can be worker behavior, company characteristics, or other attributes (all a form of meta-data) which, to develop leading indicators must have reasonable frequency in their occurrence and are tied to only one thing – abundant incident records.

In fact, nothing is a safety leading indicator until you can trend it to incident data. But how much incident data do you have access to?  Some basic leading indicators can be determined with internal analysis for a few lucky companies with large data sets and meticulous records. For the rest of the companies out there, good luck.

For many things it is simply not possible to tease out an indicator unless it is tied broadly to industry behaviors – very much a macro measure. We can include measures, broadly speaking, such as participation in a certification process or even chaotic corporate growth in a sector. We can take these larger trends and make predictions about companies, and we can take the smaller trends and make predictions about workers.

In all of these, it is essential that the data is present in abundance.

This is why we have seen an urgent trend towards additional sources of incident data such as measuring near misses and potential severity. The reason near miss reporting and classification of potential severity is so critical in modern reporting is because of the rarity of incidents. So, if developing leading indicators requires ties to incident records, then these methods of expanding the data value and frequency are the only ways available to expand this data set.

But what if it is not the only way….. What if there was another way…. Stay tuned!