Mortality Measurement: High Mortality Rates - Finding the Pony

Slide Presentation for Mortality Measurement Meeting

On November 3, 2008, Steve Jencks and Amy Rosen made this presentation at the Mortality Measurement Meeting. Select to access the slide presentation (PDF File, 96 KB).

Slide 1

High Mortality Rates: Finding the Pony

Slide 2

Charge:

"Deploying and publicly reporting a risk-adjusted measure of mortality as a quality indicator of hospital system-level performance."

• Hospital (not practitioners or diagnoses.)
• System level (single measure for hospital, but perhaps for a service or service line.)
• Risk adjusted (not the raw rate.)

Slide 3

The First National Effort

• Krakauer, Bailey, et al., produced claims-based standardized mortality estimates for all hospitals and distributed toQuality Improvement Organizations (QIOs), asking them to evaluate and use as they saw fit.
• The rates were published by the Centers for Medicare & Medicaid Services (CMS) annually for several years and a list of 50 killer-hospitals as part of the release.
• The project was killed in 1993 by an Administrator sympathetic to public hospitals.

Slide 4

Lessons

• It is very helpful to have articulated the purpose of the publication.
• It is important to have examples of using the data to make choices or improve care.
• Improved goodness of fit seems to be of little interest to the public and never enough for most physicians, and it can be very confusing.
• Selecting random discharges is confusing and hard to explain.

Slide 5

The Medicare Mortality Predictor System

• In 1989, Jencks, Daley, and others developed a system based on chart abstraction that allowed hospitals to abstract data from their own records to predict their mortality rates for heart attack, heart failure, pneumonia, and stroke.
• The purpose of the system was to allow hospitals to understand how much the best available risk adjustment would influence their ranking in the CMS publication.

Slide 6

Lessons

• Unless the articulated purpose matches a felt need of those who must invest resources to use a system, the system will have limited voluntary adoption.
• Improved goodness of fit seems to be of little interest to the public and never enough for most physicians, but it can be confusing and very expensive.

Slide 7

Cardiac Surgery

• Northern New England.
• Hannan et al.
• Epstein and Weisman.
• STS.
• State adoption.

Slide 8

Lessons

• Balancing usefulness against burden and risk.
• Some evidence that outcomes have improved.
• Focused data seem to be more actionable and therefore more accepted.
• Good risk adjustment and professional support have promoted professional acceptance and widespread adoption.

Slide 9

Krumholz-Normand Models

• Medicare, claims-based, 30-day mortality for pneumonia, heart failure, and heart attack.
• Hierarchical regression model results in very few "outliers" and is not easy for most users to understand.
• Posted to Hospital Compare identifying individual hospitals only as high, low, or as expected.

Slide 10

Lessons

• Improved goodness of fit seems to be of little interest to the public and never enough for most physicians, but it can be confusing and very expensive.
• No clear model for use. Even for the small number of outlier hospitals, use has so far been quite limited.

Slide 11

Deeper Modeling Issues

• Chronic disease v. elective surgery
  • Sampling from the dying process v. judging the risks of an intervention.
• Endogenicity:
  • Frequent v. rare rehospitalizations.
• Where and how families and patients want death to occur v. do-not-resuscitate orders (DNRs) .
• Need to model the care system, not just risk factors.

Slide 12

Broader Lessons

• The need for an explicit purpose for the system with an appropriate implementation plan and measures of success.
• The need to see the relationship of mortality to a particular hospitalization within the context of both patient-centered care and a longer patient experience.