The limits of lost time injury frequency rates

An interesting paper from Andrew Hopkins in 1994 discussing the limits of LTIFRs. Likely nothing new for most people now – but it likely was in 1994.

He explores two questions:

·???????? A general discussion on LTIFR and positive performance / process indicators

·???????? What can go wrong when LTIFRs are assumed to be a measure of safety performance

Part 1:

Off the bat, Hopkins states that: “Lost time injury frequency rates are entirely unsatisfactory as measures of safety performance for at least three reasons” (emphasis added).

1. LTIs are “far more sensitive to claims and injury management processes than to real changes in safety performance”

2. Because a given workplace may only have a few of these each year, “variations from year to year will be statistically insignificant, that is, likely to be the result of chance fluctuations, and thus no guide to changing levels of safety” (emphasis added)

3. LTIs tell us “nothing about how well the most serious safety hazards are being managed”.

Hopkins cites an example of a site posting a record of 2 years without an incident and then somebody slipped on a step and fractured their foot. At the same time a container was dropped offshore – but without an injury.

He notes “The longer the period free of injury the greater the level of disappointment and frustration when a lost time injury finally occurs, which statistically speaking is bound to happen” and that this can lead to a demoralising sense of injustice.

He notes that these are some major reasons from “various commentators” about the need to replace or supplement injury rates with other indicators. Some propose using more serious events or even fatalities – but these would be even more ill-suited that LTI given their statistical rarity – and hence “quite useless as tools for measuring changes in safety performance” at particular workplaces (but potentially suitable at an industry level).

MTIs, FAIs, and near misses may be other options – but give little indication of the “potential for serious injury”.

These limits are said to be more issues with outcome measures (injuries and the like). In contrast are process indicators – “indicators which measure safety-relevant processes rather than outcomes such as injury or fatality rates”. These differ from site to site, and measure things which have a reasonable frequency so variations have a chance of being statistically significant of real changes in safety performance. [** Process indicators are not to be confused with process safety indicators]

Different types of process indicators exist :

1. worker behaviour indicators

2. management activity indicators

He discusses some of the pros of behavioural measures – since they focus right at the interface of hazards and involve people in the task of improving safety. But they also have a drawback: “They are focussed on and aimed at changing the behaviour of employees, not managers”, yet managers “are ultimately responsible for health and safety and who are in the best position to take action on such matters”.

He says that the general principle with these types of process indicators is management specifying its safety plans and processes and then constructing measures “which assess how well these are being carried through in practice”.

Moreover, a “deficiency of most process indicators, is that they tend to focus on behaviour, either of employees or of management”. Despite this focus, there is wide agreement that “the most effective way to deal with hazards is not by altering human behaviour but by redesigning machines and systems of work so as to eliminate the hazards”.

Thus, a significant challenge is devising indicators which provide intel on the extent to which a company has succeeded in eliminating safety hazards in this way.

Part 2: Is productivity related to safety?

Next he explores the links between productivity and outcome safety measures using coal mining data.

First he says that while many commentators have said things like “safety pays”, it is misleading. Drawing on coal data (which I’ve mostly skipped), he argues that it’s not accurate to say that higher productivity improves safety doesn’t follow and then there’s “no causal connection between the two variables at all”.

Based again on this data, he says that despite a slight increase in productivity between 1972-82 – there is a doubling in LTIFR. And the LTIFRs in 1990 were nearly the same as 1972 despite a doubling in productivity.

Explaining the shape of the LTIFR curve in coal mining over decades, he argues that the most plausible explanation is that “LTIFR is more a function of claiming practices and injury management and that these changed in various ways during the period under consideration”.

He provides several explanations for why he believes this, including fitting a quadratic function; highlighting that the level of safety has improved slightly but steadily throughout the studied period.

He further argues that “the LTIFR curve is an artefact of other practices” – most particularly, after 1973, miners sustaining injuries were entitled to compensation at full award wages plus production bonuses, whereas prior to this date, compensation was below normal income.

That is, prior to the changes in Jul 73, there was a “substantial disincentive to miners taking time off work when injured; after that date there was no such disincentive”; in his view, this is the most likely explanation for the “dramatic” rise in LTIFR following 1973.

He doesn’t suggest that people were claiming illegitimate injury claims, either, rather those that were genuinely injured had more incentive to keep working and not to report.

It’s said that “This analysis provides a stark warning to anyone who seeks to draw inferences about safety from LTIFR data. These data are far more sensitive to changes in claiming behaviour and claims management practices than they are to safety, and variations in the data are likely to be indicative of changes in these practices rather than changes in safety”. [** This may be a less likely explanation now unless significant claims changes have occurred.]

He also makes some other arguments about the links between productivity and safety (about being more about the use of technology rather than human output), but I’ve skipped this.

In concluding:

·???????? “the assumption that greater productivity in the coal industry is the result of greater management attention to safety is entirely false”

·???????? There is thus no logical reason why managers concerned to improve productivity should turn their attention to improving safety”

·???????? “The role of the inspectorates is vital, particularly in relation to the most serious hazards such as roof falls, gas outbursts and explosions, all of which generate relatively few LTI claims but which can result in loss of life, sometimes more than one life, when they occur”

·???????? “It is these occurrences which make mining a hazardous occupation and it is here that effective government regulation is critical”

?

Author: Hopkins, A. (1994). The limits of lost time injury frequency rates.?Positive Performance Indicators: Beyond Lost Time Injuries: Part 1—Issues, 29-35.