Wednesday, July 11, 2007

Drug recall

According to news reports on 6 June 2007, including this one from the BBC, Roche have had to recall all batches of the anti-HIV drug Viracept (generic name nelfinavir) because it was contaminated with potentially cancer causing chemicals. The contamination is being blamed on human error.

According to the Roche website the company are to establish Viracept Patient Registries in order to register and closely follow patients who may have been exposed to a chemical impurity in their Viracept HIV formulations.

Andy Brazier

Inexplicable errors

This is something I have come across a couple of times recently. Someone makes a completely bizarre error and no one can explain exactly why it occurred. I guess this is one of the things with humans, we are not logical creatures and instead more emotional. Sometimes things can't be explained.

This article called "What next for crane safety?" by Phil Bishop, 27 June 2007 describes an incident at Canary Wharf in May 2000 where the top of a tower crane fell whilst being raised into position. In this task it is important to attach the top of the crane to the 'climbing frame.' Because this was forgotten the crane was simply balancing and able to fall. According to the article this is such an obvious and well known issue that it seems incredible that it was forgotten.

Advice in the article is given in the article to reduce the likelihood of errors, but it does not actually address the accident in question. I think the reality is that some bizarre things will happen and we should not spend lots of effort preventing them in the future because we can be fairly sure they are one off incidents. However, we can use these incidents to delve into systems as a whole and find weaknesses and hence opportunities to improve.

Andy Brazier

Quantified human reliability analysis

Available here. Academic paper by Marzio Marseguerra, Enrico Zio, and Massimo Librizzi1 entitled Human Reliability Analysis by Fuzzy “CREAM”

The work uses the the Cognitive Reliability and Error Analysis Method (CREAM) model, which assumes that the human failure probability depends on the level of control a person has over the contextual scenario in which requested to perform. Four modes of control are identified
* Scrambled,
* Opportunistic,
* Tactical,
* Strategic.

The fuzzy approach allows for ambiguity and uncertainty in the calculations.

I must take a closer look at the paper some time.

Andy Brazier