It’s time to do away with the accident investigation (AI version 1.0) as you know it. That’s right, do away with your current version of the accident investigation form. Yes, I said it and I will say it again, do away with it. And no, I am not crazy. I am just tired of not seeing results.
Over the years, I have managed thousands upon thousands of worker’s compensation claims and seen plenty of near misses (that phrase also stinks). Unfortunately, I rarely see an accident investigation form completed and if I do it’s not worth looking at anyway. The AI just doesn’t help prevent future accidents from occurring and until we change how we collect data it won’t.
400 stolen from one source
During a recent presentation on Big Data for the CalState University system (Fitting the Pieces Together) I spoke about my research on accident investigations forms. Here’s the skinny – after pulling down some 400 AI forms, they all look the same. Simple, boring and so broad and general that nothing can be gleaned from the information collected. They all seem to be replicas of a single source, kind of like all the safety professionals had the same idea – “I’ll borrow that one!” If the AI was a Gucci bag, LAPD would have a special task force looking for these counterfeits.
The problem is that when we all borrow the wrong source file we all end up with the same piece of worthless waste of time.
Why don’t we do the accident investigation
Here’s a little proof that your AI 1.0 is worthless. You don’t do the accident investigations. Time and time again, as I mentioned above, AI’s are not performed. Why not, because you already know that it is not going to produce meaningful results and impact your bottom line. So, how do you create an AI 2.0 that will?
Variety is key
AI’s are tools. Just like you wouldn’t use a screwdriver to hammer in a nail, you shouldn’t use the same AI to investigate auto accidents as you do falls from ladders. For AI 2.0, you need to have a variety of AI’s for the types of accidents you are investigating. One for slip and falls, one for ergonomics, one for autos, and one for the other major categories. That is the only way you are going to get the information you need to address the accidents at hand.
Add a dash of big data
Big data is better data. I know that you can’t instantly turn your small risk management department into a big data warehouse, but if you don’t start today you never will be able to. When I conducted my research on the 400 AI’s, I found that a majority of the data sought on the forms was simply demographic data. And as we know demographic data doesn’t really tell the story. I have never been sure how being single or married, 40 or 30, or being a certain race had anything to do with falling off a ladder. It doesn’t and still doesn’t.
What we need to start collecting are the finer details (big data) of the accident. For instance, let’s examine a fall from a ladder, what big data should we walk away with? Here’s some ideas:
- brand of ladder
- height of ladder
- weight rating of ladder
- age of ladder
- wear rating of ladder
- width of ladder
- ladder material
- rung shape – square, round, etc
- rung depth
- rung wear
- ladder stability rating
- ladder feet style
- ladder tied off
- anyone holding ladder
- how many have fallen from this ladder previously
- ladder used inside or outside
- ladder bottom on what type of surface
- ladder resting upon what surface
- slip resistance of each surfacing
- weight of injured
- weight of objects carried up ladder by injured
- brand of shoe
- shoe tread wear
- shoe size
- laces or buckles
- pant, dress or other clothing length – over the shoe or not
- light value in foot-candles at time of incident
- and much, much more
Of course we need to have some of the basic elements from AI 1.0 to bring the AI full circle and understand what happened. By adding a little bit of big data to the AI we can start to collect meaningful information and hopefully get to a place where we are preventing accidents and better yet, predicting when and where these will happen. That is the goal of AI 2.0 – collect better data.