Health Care: The Possible Link Between Information and Quality Innovation

1
Health Care The (Possible) Link Between
Information and Quality Innovation

• Bernard Black
• University of Texas
• Law School and McCombs School of Business

2
Very preliminary

• No paper
• Data partly collected
• but not analyzed
• getting data is a struggle, even when
  theoretically public
• Some ideas on where to get data and what to look
  for
• Want your feedback -- whats interesting?

3
My excuse
4
My other excuse
5
Background observation 1

• We know remarkably little about the safety and
  quality (call this quality) of medical
  procedures
• Some aggregate data on common procedures
• Hospital-level data is scarce
• Physician-level data is almost nonexistent

6
Background observation 2

• Hospitals and docs often dont want quality
  disclosure
• (1) They oppose collecting data
• (2) If collected, they oppose letting anyone see
  it on a disaggregated basis (e.g., Werner and
  Asch, 2005).
• (3) Failing that, they try to make the data
  uninformative
• e.g., HHS Hospital Compare on cardiac outcomes
• (4) Or really painful to collect
• no time series (Hospital Compare for infection
  prevention)
• cant download a spreadsheet (Hospital Compare,
  NY, CFF, etc.)
• Often, they succeed at one of these stages

7
Background observation 3

• When disaggregated information exists
• its hard to find
• its hard to use
• most patients ignore it
• doctors often ignore it in making referrals
  (Schneider and Epstein, 1996)

8
Background observation 4

• Hospitals and physicians have weak incentives to
  provide quality care
• Hospitals
• paid for procedures, not for success
• Often, paid more for low-quality procedures,
  which require (paid) followup
• Physicians
• paid much more for procedures than for thinking,
  slow medicine, or watchful waiting
• oppose treatment guidelines which limit
  aggressive treatment, often successfully

9
Background observation 5

• Large quality variations across physicians,
  hospitals, and physician-hospital pairs
• Overall quality levels are appalling
• 100,000 deaths per year (Inst. of Medicine,
  2000)
• no evidence this number is dropping
• compliance with NIH and other guidelines far
  below 100, everywhere you look
• Little relation between medical intensity and
  outcomes (Dartmouth Atlas of Health Care)
• US is 1st in the world in medical spending per
  capita, 1st in spending/GDP (both by large
  margins)
• but 37th in life expectancy

10
Hypothesis You manage what you measure

• Predict If we measure hospital quality, it will
  improve
• the worst will get a lot better
• some of the worst will drop out
• the middle will get somewhat better
• the best will compete, at least a little bit
• the best-to-worst range will decline
• the average will improve
• Supporting evidence
• NY cardiac bypass reporting (Hannan et al., 1994)

11
Testable?

• Need hospital level, time series data
• For specific processes and outcomes
• Outcome data is tricky
• Death easy to measure, but uncommon for most
  procedures
• Hence weak statistical power
• Morbidity often harder to measure
• Outcome depends on patient initial health, etc.
• Hence gaming potential, need to risk-adjust
• Cardiac bypass data is risk-adjusted
• unclear how good the risk adjustment is

12
Focus on hospitals, not docs

• Better data
• More control levers
• Reduced gaming risk
• One reading of cardiac bypass studies
• hospital disclosure is likely net good
• but physician disclosure is likely net bad
• Less political opposition (?)

13
Why Plausible for Hospitals?

• Internal control levers
• hospitals are managed by professionals
• even if (few) patients are watching, they care
  about their reputations
• some infrastructure exists
• safety officers are common
• internal error review in place (JCAHO standards)
• External control levers
• Many markets are policed by a minority of careful
  shoppers, why not health care?
• Shoppers can be docs and insurers, as much as
  patients
• hospitals will also care more about physician
  quality
• accreditation (JCAHO, CMS)
• medical malpractice

14
Benchmarking alone?

• Could pure benchmarking project work?
• Tell hospitals about each others performance
• maybe about specific physicians
• Dont tell the public
• at least until we get the risk adjustment honed
• This room is full of reputation-driven people ?

15
Some possible research areas

• Might be enough data, in a few nooks and crannies
• infection rates
• central line catheters in ICUs
• maybe some other procedures
• limited gaming risk
• infection prevention
• low gaming risk
• trying to get the data from CMS . . .
• cystic fibrosis
• good data on
• survival
• lung function (morbidity measure)
• location-based, so limited gaming risk
• but sicker patients might be more likely to move
• trying to get the data in usable form from CFF .
  . .
• cardiac bypass
• lots of operations, risky, so some statistical
  power from mortality data
• but gaming risk (especially if surgeon-level
  reporting)

16
Example central line catheter infections

• 80,000 infections per year
• 20,000 deaths
• Almost all preventable, in 5 easy steps
• wash hands
• use chlorhexidine soap
• use full surgical drape
• remove unnecessary catheter lines
• avoid femoral artery where possible

17
Pronovost, NEJM (2006)
Median central line infection rates (per 1,000
catheter-line days. 96 Michigan hospitals with
ICUs National average rate 5.3
18
Ascension Health study
Berriel-Cass, et al. (2006)
19
ICU infection rate data
Also some small voluntary disclosure
programs FL reports only risk-adjusted rate
20
Pennsylvania outliers

• Hahneman Univ. Hospital
• 2005 157 infections (9.7 rate)
• 2006 59 infections (3.5 rate)
• Hershey Medical Center (Penn State)
• 2005 166 infections (7.7 rate)
• 2006 208 infections (9.3 rate)
• Infer Hahneman learned Hershey didnt
• My hypothesis They will.

21
On bad law and zealous bureaucrats

• HHS reaction to Pronovost study
• Shut it down, for violating HIPAA regs
• No kidding
• 1,000 lives/year in Michigan alone, and HHS shut
  it down
• Initial HHS view
• ok to require docs to wash hands
• But keeping track of whether they do is
  research on human subjects ? requires informed
  consent
• Tells you HHS priority on health care quality

22
Whats my comparative advantage?

• Find state and federal laws
• nontrivial job, now underway
• extract the data from govt websites
• Also, inexcusably, a nontrivial job
• Exploit state variation, time variation
• natural experiment
• prior research is limited
• none on infections, infection control prcess, or
  CF
• limited on cardiac bypass
• not that many others are looking
• docs are not good econometricians
• economists dont know the law, and this issue has
  not been a priority for them

23
Antibiotics before surgery

• See www.hospitalcompare.hhs.gov
• Check hospitals in your area
• Limit 3 used to be 12, but now 3 after a recent
  quality downgrade
• Any guesses on who wanted the 3-hospital limit?
• Time series data not available
• Might need a FOIA request, well see

24
Cardiac bypass mortality

• Raw and risk-adjusted mortality rates

Prior research Dranove et al. (JPE 2004, NY
PA, selection effects) Hannan et al. (JAMA 1994,
NY) Peterson et al. (JAmCollCard 1998, NY)
25
Cystic fibrosis

• Source Cystic Fibrosis Foundation
• Data for 2002-2006 on their website
• but almost unusable
• asked them to help, theyre not interested

26
Cystic Fibrosis
Lung function in children (versus non-CF
children, adjusted for age, weight,
gender) Brackenridge (Austin TX) vs goal and
national average
27
CF over time

• The very good news
• 50 years ago CF kids used to die by age 3
• Today The oldest survivors are in their 50s
• The uncomfortable side story
• Odds of living to 30, or 40, or 50, depend on
  where you live