Pharmaceutical marketers go to great lengths to find the doctors who aren't prescribing their drugs—and to devise methods to reach them. But MedNetworks, a startup that grew out of the Harvard lab of sociologist and physician Nicholas Christakis, is offering pharmaceutical companies a shortcut. By mining anonymized medical-claims data, the company says, it can identify which doctors may be the strongest influencers of their colleagues.

MedNetworks uses computational tools developed at Christakis's lab to look at the prescribing patterns of large groups of doctors, build maps of professional ties, and track how the popularity of a new drug grows. The company has found that certain doctors are particularly strong influencers: when these doctors write prescriptions for a newly released drug, colleagues within three degrees of separation soon follow suit. With such historical insights in hand, "we've shown that we can predict adoption of pharmaceuticals among doctors," says MedNetworks cofounder Larry Miller.

This approach could go far beyond aiding drug companies. By gleaning patterns from the seeming chaos of physician partnerships and referrals, patient health records, and prescription-writing records, computational social-analysis tools could also identify which people in a community would be most influential in spreading a public-health message.

To that end, MedNetworks is working on identifying such influencers among citizens in Hermosa Beach, Redondo Beach, and Manhattan Beach, California. Starting with public information such as census data, club rosters, and PTA lists, the company is trying to determine which individuals act as community and neighborhood influencers. The client in that case—Healthways, a consultant for employers trying to drive down health costs—wants to find people who would be most effective in spreading messages about reducing smoking and obesity.