Saturday, August 05, 2006


Was drinking yesterday with Tommi and his philosopher buddies. One of them was a girl named Essi, who worked in some EU-funded information society assessment project.

She told that in Sweden, the recipe system has been digitalized. The doctor sends the recipe information to a database, and at the drug store the drug clerk fetches the information from the database.

One obvious advantage of this kind of system is that it produces a big data set that allows data mining - supposing that privacy concerns don't prevent experts from mining it. Half Sigma demonstrates in his writings just how much information you can mine from a good data set. He also compares his findings to real research, and often finds that the results are the same. I'd guess that good data sets and proper data mining could considerably increase the productivity of researchers in some areas.

One obvious candidate for digitalization are the patient records. Earlier, I thought that privacy concerns were the main concern here, but Essi told that not only doctors but also the staff has access to relevant records. Therefore, the patient records can't be very confidential. So, it's just data; why can't they digitalize it? What's the problem, damnit?

If I were a dictator, all health-related information would be put to a big database. This would include not only patient records, but also all research data sets. Then, a public interface would be created to the database. This interface would allow anyone to make simple database queries to find out about numbers and correlations between various factors.

Privacy concerns would be addressed by putting a lower limit to the amount of cases. For example, if some query returns less than 20 matches, the results would not be shown for privacy reasons. If the database contained 5 million people, the restriction wouldn't be that harsh.

This kind of database would help researchers in quick testing of hypotheses. For example, this research tells about a parasite T. gondii that spreads from cats to humans and may cause schizophrenia. If some earlier research had tested the presence and absence of T. gondii for whatever reasons and this information was saved to the database, then a simple correlation between T. gondii and schizophrenia could be computed. This wouldn't replace the actual research - feeding antipsychotic drugs to rats that have T. gondii - but it would enable better justifications before any expensive animal tests are done. Database queries are damn cheap.

More bang for buck in medical research would speed up the achievement of SENS, so anyone appreciating longer life should agree.

No comments: