Tuesday, August 28, 2007

When statistical analysis gets scary

Statistical analysis finds evidence of human-to-human bird flu transmission, reports the Fred Hutchinson Cancer Research Center:
The researchers based their findings on a cluster of eight flu cases within an extended family in northern Sumatra. Using a computerized disease-transmission model that took into account the number of infected cases, the number of people potentially exposed, the viral-incubation period and other parameters, the researchers produced the first statistical confirmation of humans contracting the disease from each other rather than from infected birds.

The cluster contained a chain of infection that involved a 10-year-old boy who probably caught the virus from his 37-year-old aunt, who had been exposed to dead poultry and chicken feces, the presumed source of infection. The boy then probably passed the virus to his father. The possibility that the boy infected his father was supported by genetic sequencing data. Other person-to-person transmissions in the cluster are backed up with statistical data. All but one of the flu victims died, and all had had sustained close contact with other ill family members prior to getting sick -- a factor considered crucial for transmission of this particular flu strain.
The close cousin of this type of research and analysis is predictive analytics -- and I find it somewhat alarming that a Google search for the following terms:

predictive analytics disease

.... turns up 209,000 English pages.

No comments: