Camden found its costliest patients by crime-mapping hospital billing data

Camden, New Jersey was one of the poorest, most crime-ridden cities in the nation, its school system in receivership and the state running the government. Family physician Jeffrey Brenner, serving as a citizen member of a police reform commission there, absorbed the Compstat approach of policing — map incidents, focus resources on the hot spots. When the police refused to build crime maps, Brenner made his own from hospital billing records; beaten back by police unions, he kept playing with the hospital data anyway.

The billing maps showed what no single hospital could see. One percent of patients accounted for a third of the city's medical costs. Beyond assaults, Brenner found that a single building in central Camden had sent more people to the hospital with serious falls than any other in the city — 57 elderly people in two years, almost three million dollars in health-care bills — the kind of environmental finding that no per-patient view would surface.

In May 2009 Brenner closed his medical practice to work on the problem full time, operating as the Camden Coalition of Healthcare Providers from desk space at Cooper Hospital, studying medical-home programs for the chronically ill and building teams around the heaviest users. The coalition measured its first 36 super-utilizers: 62 hospital and E.R. visits per month before joining, 37 after — a forty-percent cut — and bills of $1.2 million per month before, just over half a million after, a fifty-six-percent reduction.

Treat the data like crime: merge three hospitals' billing records into one searchable map, rank by cost, and send care to the addresses and people at the top.

Cost concentration was invisible to every participant: each hospital saw its own slice of a revolving patient, so nobody could rank or even name the top users.

The mapping transferred a proven method — police hot-spot targeting — to a dataset no one had thought to rank that way.

Addresses, not just names, turned out to be the actionable unit: a single building could be a bigger cost driver than a diagnosis.

Concentrating care teams on a few dozen people was affordable precisely because their bills were so concentrated.

Borrow the analytics of the nearest solved problem: police had already proven hot-spot targeting works — medicine only needed the same map drawn over billing data instead of crime pins.

The New Yorker published the findings in January 2011, noting the caveats — the sample was small, the measurements excluded personnel and medication costs, and some patients might have improved anyway — but calling the results, net, 'almost certainly, revolutionary.' The article also flagged that the program's funding model remained unclear.