Why Reporters (And Hospital Administrators) Should Learn Statistics

Interesting article on amednews.com about the pros and cons of publically posting emergency department waiting times. The pros are that patients can self-triage to the least busy ED, and it might be good for a hospital’s business. The cons are that patients who are really sick might be discouraged from going to any ED if the waiting times are long, and ED doctors might cut corners to speed patient throughput.

One paragraph of the article caught my eye.

“Scottsdale Healthcare began posting wait times in April 2008 at its four EDs, all of which are within about 15 minutes' driving time of one another in the city (two -- a general ED and a pediatrics ED -- are housed at the same center). Its patient satisfaction scores have improved by 2 percentage points [emphasis added], said Nancy Hicks-Arsenault, RN, the organization's systems director of emergency services.”

I can’t be sure but knowing what I do about patient satisfaction scores [a good subject for future blog], I would bet that a 2% increase in patient satisfaction is not statistically significant. In my experience, fluctuations in patient satisfaction scores of 2% are common and well within one standard deviation of the average for these rather crude measures. One of the most popular patient satisfaction survey companies uses a rating scale of 1 through 5 and then converts the responses into percentages. This means that if a patient rates an ED service as a “4” instead of a “3”, that is a 20% increase in satisfaction when the patient may not really have been 20% happier with his experience. The response rate of most patient satisfaction surveys is usually below 10% which further diminishes their validity.

I would have asked to see the raw numbers, performed a statistical test and determined if a 2% increase in patient satisfaction was real or not.