October 25, 2011

In the state of Colorado, there are efforts underway to reduce the number of accidents occurring on the I-70 mountain corridor. If you have driven this route in the last ten years, it was probably during the tourist season, and you saw the sources of the problem first-hand. Specifically, there are major mountain ascents, long tunnels burrowing under mountains, "merge zones" (lanes changing from 3 to 2 before entering tunnels), posted speed changes, too many cars, and too much snow (and many people not used to driving on it). This is a bad mix of conditions and result in one of the most congested stretches in the U.S. interstate highway system.

Civil and traffic engineers have known for years that the major problem to congestion and accidents is the traffic speed differential. The American Association of Surface Highway Transportation Officials points out that crashes are not related so much to speed as they are to "the range in speeds". Also, more studies have shown that when the speed varies, there is a greater chance of a crash. To fight this, Colorado is engaged in a program called "active traffic management". This involves (in part) putting out pace cars on the highway to regulate the speed of all the vehicles, thereby improving the average speed. Though some (actually many) might complain about this, the fact is that it does work: the average time per car over the entire population of cars is improved.

In the business of process engineering, we are fond of saying that variation "kills" (the example above makes a more literal case). In the lab environment, when we say variation kills we mean that this is the thing to concentrate on more than any other. For example, it's not so much the average turnaround time of a cardiac marker that is the problem: it's the fact that most of the tests are completed in 30 minutes and some are taking 2 hours! The problem is the variation, or more specifically, the unreliable processes that are creating variable outcomes. Although the data range might be 1 ½ hours, the average will look artificially good at under one hour. This is why many labs (and other organizations) rely on a compliance number instead of an average number (i.e., turnaround times are under one hour, 80 percent of the time). This provides a more accurate picture of what is going on in the lab. With this in mind, our quick approach to dealing with variation includes the following:

Read the Full Article