Scoring Severity of Injury – Hidden Probabilities

Last updated on May 25th, 2023 at 04:11 pm

I’ve been thinking a lot about risk-scoring tools and our algorithms. One of the key elements of risk is the Severity of Injury. There are hidden probabilities attached to the Severity of Injury scores that are assigned that are not discussed clearly in any of the risk assessment standards that are commonly in use.

This all started when I was challenged to write an analysis of the problems with the CSA Risk Scoring Tool that you can find in the 2014 version of CSA Z434. That tool is deeply flawed, in my opinion, but that is not the topic of this post. If you want to read my analysis, you can download the white paper and the presentation notes for my analysis from the Compliance inSight Publications page [1].

Scoring risk can be a tricky thing, especially in the machinery sector. We rarely have real-world data to use in the analysis, so we are left with the opinions of those building the machine as the basis for our evaluation. Severity is usually the first risk parameter to be estimated because it’s seen as the “easy” one – but only if the characteristics of the hazard are well known. One aspect of severity often missed is the probability of a specific injury severity. We’re NOT talking about how likely it is for someone to be injured here; we’re talking about the most likely degree of injury when the person interacts with the hazard. Let me illustrate this idea another way: Let’s call Severity “Se,” any specific injury “I,” and the probability of any particular injury “P_s.” We can then write a short equation to describe this relationship.

Since we want there to be a possibility of no injury, we should probably relate these parameters as a product:

Ok, so what? This equation says that the Severity (S_e) of any given injury (I) is the product of the specific type of injury and the probability of that injury. More simply yet, you could say that you should be considering the most likely type of injury that you think will occur when a person interacts with the hazard. Consider this example: A worker enters a robotic work cell to change the welding tips on the robot’s welding gun. This task has to be done about once every two days. The entry gate is interlocked, and the robot was locked out before entry. The floor of the work cell has wireways, conduits and piping running across it from the edges of the cell to the various pieces of equipment inside the cell, creating uneven footing and lots of slip and trip hazards. The worker misses his footing and falls. What can you expect for Se in this case?

We know that falls on the same level can lead to fatalities, about 600/year in the USA [2], but that these are mainly in the construction and mining sectors rather than general manufacturing. We also know that broken bones are more likely than fatalities in falls to the same level. About a million slips and falls yearly result in an emergency room visit; of these, about 5%, or 50,000, result in fractures. Ok, so what do we do with this information? Look at the typical severity scale from IEC 62061 [3].

Table 1 – Severity (Se) classification [2, Table A.1]

Consequences	Severity (Se)
Irreversible: death, losing an eye or arm	4
Irreversible: broken limb(s), losing a finger(s)	3
Reversible: requiring attention from a medical practitioner	2
Reversible: requiring first aid	1

Using Table 1, we might come up with the following list of possible injury severities. This list is not exhaustive, so feel free to add more.

Table 2 – Potential Injury Severities

Possible Injury	Severity (Se)
Fall on same level – Fatality	4
Fall on same level – Broken wrist	3
Fall on same level – Broken collarbone	3
Fall on same level – Torn rotator cuff	2
Fall on same level – Bruises	1
Fall on same level – Head Injury	3
Fall on same level – Head Injury	4
Fall to a lower level, > 3 m – Fatality	4

How do we score this using a typical scoring tool? We could add each of these as line items in the risk register and then assess the probability of each, but that will tend to create very large risk registers with many line items with very low risk scores. In practice, we decide on what we think is the most likely degree of injury BEFORE we score the risk. This results in a single line item for the hazard rather than seven, as would be the case if we scored each potential injury individually.

We need a probability scale to use in assessing the likelihood of injuries. No published scoring tool I know of currently has a scale for this, so let’s do the simple thing: Probability (Ps) will be scored from 0-100%, with 100% being a certainty.

Going back to the second equation, what we are doing is assigning a probability to each of the severities that we think exist, something like this:

Table 3 – Potential Injuries and their Probabilities

Possible Injury (I)	Severity (Se)	Probability (Ps)
Fall on same level – Fatality	4	0.0075%
Fall on same level – Broken wrist	3	5%
Fall on same level – Broken collarbone	3	5%
Fall on same level – Torn rotator cuff	2	5%
Fall on same level – Bruises	1	90%
Fall on same level – Head Injury	3	1%
Fall on same level – Head Injury	4	0.0075%
Fall on same level – Lacerations to hands	2	90%

The percentages for fatalities and fractures were taken roughly from [1]. If you were to construct a frequency distribution, you would find that these injuries tend to fall on a bell curve like the one shown in Fig. 1.

We can look at a table like this and say that cuts and bruises are the most likely types of injury in this case. We can either group them for the overall risk score or score each individually, adding two separate line items to the risk register. I will use the other parameters from [2] for this example and develop an example risk register, Table 4. In Table 4,

S_e = Severity

P_r = Probability of the Hazardous Event

F_r = Frequency and Duration of Exposure

A_v = Possibility to Avoid or Limit Harm

The algorithm I am using to evaluate the risk is from [1]

Note that I have selected the highest severity of the combined injuries when I have combined the two potential injuries into one line item (Item 1 in the register). The less likely severities, particularly the fatalities, have been ignored. You can click on  Table 4 to see a larger, more readable version.

I did not reduce the Se scores in the Final Risk Score because I have not changed the slip/trip and fall hazards, only to the likelihood of the injury occurring. In all cases, we can show a significant risk reduction after mitigation. I’m not going to get into risk evaluation (i.e., Is the risk effectively controlled?) in this particular article, but the fact that you can show a significant risk reduction is essential. There are many considerations in determining if the risk has been effectively controlled.

Conclusions

The probability of specific injuries occurring must be considered when estimating risk. This process is mainly undocumented but occurs. When risk analysts are considering the severity of injury from any given hazard, this article provides the reader with one possible approach that could be used to select the types of injuries most likely to occur before scoring the rest of the risk parameters.

---

References

[1] D. Nix, ‘Evaluation of Problems and Challenges in CSA Z434-14 Annex DVA Task-Based Risk Assessment Methodology‘, Kitchener: Compliance inSight Consulting Inc. 2015.

[2] National Floor Safety Institute (NFSI), ‘Quick Facts – Slips, Trips, and Falls’, 2015. [Online]. Available: http://nfsi.org/nfsi-research/quick-facts/. [Accessed: 21- Jul- 2015].

[3] Safety of machinery — Functional safety of safety-related electrical, electronic and programmable electronic control systems, IEC 62061. International Electrotechnical Commission (IEC), Geneva, 2005.

[4] Bell Curve – Normal Distribution. pngkey. [Online]. Available: https://www.pngkey.com/png/detail/116-1165950_bell-curve-normal-distribution.png. [Accessed: 2023-05-25].

© 2015 – 2023, Compliance inSight Consulting Inc.
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.