Friday, June 20, 2008
Mining Patient Data
Tracking drug safety
Health insurers have been amassing huge volumes of data charting medical claims and prescription drug use among the people they cover. Now they are putting the information to a surprising use: Tracking drug safety for the public. Insurers plan to announce one of the first real time drug surveillance systems in collaboration with the Food and Drug Administration. One insurer will systematically scan the medical information of more than half of its 35 million members to look for hidden patterns or spikes in medical problems that might be linked to certain medications or combination of drugs as they happen.
Seems the drug safety monitoring system used by the FDA is slow and passive and relies largely on harried doctors and drug companies to report problems they see crop up with patients, and captures less than 10% of bad reactions. The big wake-up call was the 2004 withdrawal of Vioxx that was taken by 20 million people over 5 years before enough evidence revealed its link to heart attacks. Congress has passed a law mandating the FDA to set up a new more comprehensive computerized system with records of tens of million patients to scan for problems with medications. Rather than re-create its own, the FDA intends to contract with insurers and other health systems to rely on their databases. The agency will bring together different groups in a network and partner with a lot of groups. For consumers, the hope is that problem drugs will be detected much faster and more reliably than is possible with the current system.
The network of databases will be used in several ways. One will be to follow up on signals of potential problems that cropped up during the pre-approval testing of a new drug but weren't enough to prove a risk. Other signals or alerts will continue to come from doctor reports or academic studies, and insurers aim to build artificial intelligence into the system so that it will identify spikes or signals of potential problems. In addition to medical and pharmacy claims data, they will also crunch lab test results and medical charting information of their members. In the Vioxx case, the insurer could have spotted heart related hazards within 4 months after the drug's debut, if they had mined their members data.
Drug industry representatives seem to be watching such initiatives with a lot of interest and say that it would be important that the database network not just rely on medical claims, whose coding doesn't always reflect what's happened to a patient, but more comprehensive patient information, but right now there is a sparsity of electronic medical records.
Health insurers have been amassing huge volumes of data charting medical claims and prescription drug use among the people they cover. Now they are putting the information to a surprising use: Tracking drug safety for the public. Insurers plan to announce one of the first real time drug surveillance systems in collaboration with the Food and Drug Administration. One insurer will systematically scan the medical information of more than half of its 35 million members to look for hidden patterns or spikes in medical problems that might be linked to certain medications or combination of drugs as they happen.
Seems the drug safety monitoring system used by the FDA is slow and passive and relies largely on harried doctors and drug companies to report problems they see crop up with patients, and captures less than 10% of bad reactions. The big wake-up call was the 2004 withdrawal of Vioxx that was taken by 20 million people over 5 years before enough evidence revealed its link to heart attacks. Congress has passed a law mandating the FDA to set up a new more comprehensive computerized system with records of tens of million patients to scan for problems with medications. Rather than re-create its own, the FDA intends to contract with insurers and other health systems to rely on their databases. The agency will bring together different groups in a network and partner with a lot of groups. For consumers, the hope is that problem drugs will be detected much faster and more reliably than is possible with the current system.
The network of databases will be used in several ways. One will be to follow up on signals of potential problems that cropped up during the pre-approval testing of a new drug but weren't enough to prove a risk. Other signals or alerts will continue to come from doctor reports or academic studies, and insurers aim to build artificial intelligence into the system so that it will identify spikes or signals of potential problems. In addition to medical and pharmacy claims data, they will also crunch lab test results and medical charting information of their members. In the Vioxx case, the insurer could have spotted heart related hazards within 4 months after the drug's debut, if they had mined their members data.
Drug industry representatives seem to be watching such initiatives with a lot of interest and say that it would be important that the database network not just rely on medical claims, whose coding doesn't always reflect what's happened to a patient, but more comprehensive patient information, but right now there is a sparsity of electronic medical records.
# posted by Srivathsan @ 9:12 PM 
