Compared to What? (Pt.2)

In one organisation I have been working with, a number of staff complained that they were suffering from 'sinusitis' following a change in the air-conditioning contractor.

Sick leave statistics were requested from the HR department and sure enough the number of days lost to 'sinusitis' had increased significantly over the number of days lost to sick leave for the same reason in the previous year. So it seems like an open and shut case, right?

Wrong! Sinusitis can easily be confused with the common cold, hay-fever, influenza or certain kinds of headache by the layman, and most of these cases were not medically diagnosed. Even where the person sees a doctor the chances are that if the patient claims that it is down to the air-conditioning, this will influence the diagnosis, especially in the absence of any medical tests.

So what was the actual situation? A business analyst (himself a regular suffer of sinusitis since childhood) took another look at the data and found two interesting anomalies:

• firstly, the total time taken for sick leave from all causes had not increased since the previous year
• secondly, the number of cases of cold and influenza had significantly dropped over the previous year and
• finally, the seasonal pattern of the 'sinusitis' in the current year was remarkably similar to the season pattern of influenza in the previous year.

What this suggested was that the number of actual cases of sinusitis had not increased, but that staff had changed their definition of what constituted 'sinusitis'. In epidemiology, this is known as a classification error. What may have happened is that staff may have seen something on TV about sinusitis or their doctors may have changed what they classed as sinusitis. In any case, what this pointed to was that the apparent increase was an artifact of the change in classification.

The general principle we can derived from this is that you can't look at things in isolation. You need to look at the larger picture and where you notice anomalies see how those anomalies relate to the event or situation which prompted the initial investigation.

In the case above, it seemed reasonable to compare the current year cases of what staff were calling 'sinusitis' with the previous year. In some situations, it does make sense to compare like with like. But where an apparent change bucks a trend then it requires investigation. Where a totality hasn't changed, the only way a part of that totality can increase is for another part to decrease and where the part which has increased is easily confused with the part that has decreased, you may be facing a classification error.

In this instance, accepting the initial data at face value could have led to corrective measures that weren't necessary. Because someone looked beneath the surface, this didn't happen.