1992: A software error causes London’s brand new computer-aided ambulance-dispatch system to fail. The ensuing snafu is blamed for anywhere between 30 and 45 deaths.

CAD, as it was known, replaced the London Ambulance Services human dispatchers who had handled things efficiently, if a bit slowly, for years. The new system was supposed to reduce the average emergency-response time to three minutes, as mandated by a recent government decree. That was beyond the capabilities of the human dispatchers, whose cumbersome methods — involving a lot of paper-shuffling and looking things up on maps — meant that most ambulances arrived well after the three-minute window.

Even for a fully functioning computer system this was a daunting task, because the service fielded between 2,000 and 2,500 calls per day, with 60 percent of those calls requiring one of its 200-plus ambulances on the street to be sent to the scene.

CAD had been in operation only a few hours when problems began to arise. In fact, the system went live without load-testing and with at least 81 known software issues.

To make matters worse, computer training had been done 10 months before the system became operational. (Duh, few people can remember a training that far back without getting hands-on experience, and there’s also the issue of 10 months’ worth of changes.) And, oh yeah, there was no backup contingency in place in case something went wrong.

As for the software itself, three primary flaws hampered things from the start: It didn’t function well when given incomplete data, the user interface was problematical and — most damning — there was a memory leak in a portion of the code.

These folks are supposed to respond to disasters, not create them.

But balky software was only one problem. In the subsequent investigation, the London Ambulance Service was pilloried for poor project management and for trying to cut costs by throwing together a system on the cheap: It turns out that the LAS used hardware already on hand instead of purchasing newer, more up-to-date equipment. Among other things, it meant that the system was built using hardware not customized for the specific purpose.

Completing a process that was a cock-up from start to finish, LAS poohbahs committed a sin common to management types the world over: The people who would actually be using the new tools were not consulted during the development phase. It was left to project managers and outside vendors to imagine what might be most useful to the personnel on the ambulances.

The result in those first hours was complete chaos on the streets. As the system crashed, dispatchers failed to send ambulances to some locations while dispatching multiple units to others.

It got worse as people expecting an ambulance and not getting one began to call back, flooding the already-overwhelmed service. In one case, a person who died while awaiting help had already been removed by the mortician before the ambulance arrived.

Heads rolled in the aftermath, as heads usually do. John Wilby, the LAS chief exec, resigned and a number of other managers were either dismissed or reassigned.

A software upgrade in 2006 caused the system to fail again. This time the dispatchers went back to the old pen-and-paper routine. It was slow, but it worked.

Source: ErichMusick.com