Injury and Illness Data - Redux

I just can't let this go. On March 9th I posted my thoughts on injury and illness data. On March 10th I got this comment:

"Did you drink the Kool Aid too? I suggest you check out a previous Pump Handle post on this subject and review the statistical analysis published in the technical literature:

I suppose she (Chao)and those at OSHA who drank the Kool-Aid choose to ignore the empirical evidence that suggests that this substantial decline “corresponds directly with changes in OSHA recordkeeping rules.” [Friedman & Forst, 2007]"

Very little pisses me off more than someone accusing me of "drinking the Kool-Aid" on a topic I probably know more about than the commenter. One thing that does piss me off almost as much, are academics who blindly accept another academics work without a critical reading (he says while not so discretely pointing at The Pump Handle). So I read the Friedman & Forst article, and boy did he make one big mistake.

Dr. Friedman's basic argument is that changes to the OSHA Recodkeeping rule have been responsible for an artificial drop in the injury and illness rates, and he cites rule changes in 1995 and 2002 (kind of). So let's look at the 2002 rule change, which Dr. Friedman says is really a 2001 change. (I'm going to set aside the 1995 change since it deals with how the Bureau of Labor Statistics (BLS) collects the data and I don't work for BLS.) In the 5th paragraph Dr. Friedman says:

"Then in 2002, a new OSHA recordkeeping rule was adopted. Although the rule change for OSHA recordkeeping took effect in January 2002, the mandated annual reporting of occupational injuries and illnesses by employers occurs between February and April. Therefore, 2001 injuries and illnesses are recordable under the new 2002 rule. The new regulation dramatically changed which casualties were deemed 'recordable,' and in most cases a more exclusive definition was implemented."

And every CSHO out there is asking: "What the hell is he talking about?" You see, Dr. Friedman demonstrated in that one paragraph that, when he wrote the article, he lacked a fundamental understanding of how the OSHA recordkeeping standard actually works. Let me explain.

At 12:00 midnight, on January 1, 2001, every company that was required to maintain injury/illness logs, should have taken out a brand new, unused OSHA Form 200 (aka OSHA 200 Log). As injuries occurred throughout the year, the employer was required to add the information about the injuries to the OSHA 200 Log, and then update lost/restricted days as the information became available.

At 12:00 midnight, on January 1, 2002, every company that was required to maintain injury/illness logs, should have taken out a brand new, unused OSHA Form 300, notice OSHA 300 Log, and begun the recording process all over for 2002 injuries, but under the new regulation. The employer then had the month of January to finalize the OSHA 200 Log, per the old regulation, for posting in the workplace for the month of February. It is true that the data collection also begins in February, but data collected in 2002 for the 2001 calendar year was recorded under the old recordkeeping rule, not the new one!

Dr. Friedman's entire article is based partially on the assumption that the injuries/illnesses that occurred in 2001 were recorded under the 2002 rule, and therefore there was a significant drop of 12% from 2001 to 2003 (5.7 to 5.0 cases per 100 employees), or about 6% per year. But all Dr. Friedman really has is a drop of less than 6% from 2002 to 2003, and we all know that two data points don't make a good trend. Let's look at the rates for the last 14 years:



Now that we've had some more time under the new recordkeeping rule, let's compare changes under the two versions of the rule. The drop from 2002 to 2007 was 24.5%, the drop from 1996 to 2001 was 22.0%. Not a real big difference. So my question becomes: Do changes to our recordkeeping standard really cause artificial drops in injury and illness rates? The answer: No. Think about it, our current standard hasn't changed since 2002, yet the rates are still dropping. Employers are supposed to be following the same requirements today as they were in 2002, so how can the standard change from 2001 to 2002 still be effecting the rate drops today (the latest data is from 2007)? It makes no sense! Are companies colluding to under report? I doubt it, but even if they were, that is violating the regulation, it isn't a problem with the recordkeeping standard itself.

I could accept an argument that the change made an impact in the rate drop from 2001 (the last year under the old standard) and 2002 (the first year under the new standard), maybe. But the drop from 2006 (4.4) to 2007 (4.0) is the same value drop, and higher percentage drop, than the 2001-2002 (5.7 to 5.3) drop.

I said it in my March 9th post and I'll say it here again: the BLS data is not completely accurate, especially when it comes to recording days lost, but there are real drops in the injury rates. Instead of blaming the standard, let's try to look at what is causing the rate drop. It annoys me to no end that people insist that there has to be only one explanation for the drop. The first step in examining the drop is to accept the premise that it is unlikely that any one factor is causing it, sometimes Occam's Razor doesn't apply.

This is from Dennis Barker, USW Local 1899, Safety and Health Chairman:

"My opinion of why the injury rates have fallen is not because of workplaces being safer I think the main reasons are, 1. The loss of over three million manufacturing jobs since 2000, service type jobs are less hazardous 2. Incentive programs where prizes or money is awarded to employees for not reporting injuries. 3. Dis-incentive programs where discipline or time off is given to employees after every accident or injury.

One term we use for these incentive and dis-incentive programs is the "Bloody Pocket Syndrome" Which means an employee would rather stick his bloody hand in his pocket until they get home and then do their own first aid rather than reporting a work related injury."

Dr. Friedman hinted at several other factors, the reduction in big OSHA cases (most of the recordkeeping egregious cases were in the late 80's and early 90's), fewer CSHOs (1300 in 1990 to 1100 in 2003), and the shift to non-manufacturing sectors. He dismissed the influence of these factors, but I think mistakenly.

When people discuss the service industry as less hazardous, what they often do is dismiss health care. Let's compare health care to manufacturing:



I am not an epidemiologist, I don't play one on TV and I did not sleep at a Holiday Inn Express last night, but here are just a few ideas that need to be explored, with thanks to Dennis and Dr. Friedman (even if his participation was unwitting):

1. The shift of jobs to other industries (note: when a manufacturer down sizes, it is usually the least experienced employees who are released, and we've all seen the studies on the lack of experience versus injury rate).
2. Incentive/disincentive programs.
3. Lack of significant recordkeeping cases.
4. Loss of compliance staff.
5. Increase in the number of safety and health professionals
6. Internet access (I can tell you that many employers absolutely do not like to call OSHA for information, but they might access the website, which, if the rumor I heard was true, gets almost 1,000,000 hits per month).
   

These are only a few off the top of my head, help me out, leave your suggestions under the comments so I can create the list.

Just remember, when we look at these factors, the whole is the sum of all of its parts.