I’m at a pay for performance meeting in Los Angeles…but no wi-fi in the room so only occasional postings..
The best known name in clinical process improvement is Intermountain’s Brent James. He discovered Deming in 1986; ended up at a 4 day conference that Deming taught and had to translate the words out of manufacturing into health care. Deming said track cost outcomes as well as clinical ones. Luckily Intermountain had built-in activity based cost activity. And they used it for a study on post-operative infection. They measured using antibiotics pre-operative at optimal time, and found that they were there at 40% at right time,. with an infection rate way better than their peers. But they were only good copared to everyone else. By 2001 they were at 96% given at optimal time infection rates down from 1.6% to 0.4%, and they saved $714,000 per year in health care costs they never had.
The criticism of IHC is that for 10 years they knew about how to do it, but they didn’t implement it system wide till 2001. Quality improvement is just process management and improvement. A typical hospital has around 1,000 different processes. They all have physical (clinical), satisfaction and cost outcomes — all of which are a result of those processes. Analytically Brent can’t tell the three apart.
Currently reprising Anderson’s study on the % of “quality waste” in US healthcare — was thought to be 25–40%. Now they think may be 40%-50%. Then there’s “inefficiency waste” (the use of more resources to create the same outcome) which they think may be more like another 50%.
In 1995 he could identify more than 65 processes where they were saving money. He found $30m in savings. But how to spread that? Building on some work of family docs on pheumonia, implementing the correct compliance of anit-biotics for pneumonia. Again went from 22% to 90%. Complication rate down, mortality rates down 26% (around 100 deaths per year), and costs down about 10–12% (depending on how you measure it).
But the hospital adminstrators wanted to know where these showed up in their budgets. So he tracked the per-case revenues, and found that their costs had fallen, but their revenues had fallen more — because their DRG did not “creep”. And the more expensive DRGs were more profitable, and they were still losing money on the less complicated DRG. So there was actual incentive to do it wrong.
(For InterMountain about 85% was either FFS or DRG/per case based). So they were going to lose out every time they moved patients to better DRGs.
What do they do now? They use these in contract negotiations to try to keep more of the share by improvements and reduce costs too. Final strategy is to try to move everyone into shared risk. Brent’s reason that California groups got pounded because they didnt have the mechanism to measure it (no IT systems). Now they try to get their partners to split the savings three way (hosp, docs, insurers). Of course they can’t yet do this with the Federal government. He hopes for P4P from the Feds.
There are 2 main models A) paying bonus on ranking—bonus doesnt cover the added revenue drop, B) Use shared savings
BUT several issues — 1) P4P needs sophisticated data, 2) needs to look at all care (to avoid shifting from one category to another) 3) Lead time: major costsavings from many programs are 2–3 years down the track (example is mental health counselling)
Brent James says that P4P is here to stay and that we know how to close that gap. Getting 30% savings is realistic.
Jack Lewin asks if they are getting anywhere to get Medicare to pay on admission not discharge diagnosis.
He was asked if there is possible change without major changes to the payment system—the AMGA-outpatient payment demo is interesting because it’s a shared savings model. But the bigger issue is getting the information systems into the field to do this. Last year in InterMountain a $3m investment in management systems and data created a $15m variable cost on the bottom line.
ABC News reporter Joy Victory found a really juicy angle on the problems of being uninsured. Her article is called 
