Using AHRQ Composites

AHRQ 2008 Annual Conference

Slide 1

Using Agency for Healthcare Research Quality (AHRQ) Quality Indicator (QI) Composites

• AHRQ Annual Conference 2008.

Slide 2

What is a composite?

• Composite measures are combinations of two or more components measures.
• Component measures may be outcome (mortality, morbidity, proxy), process, structure, patient experience, cost or conjunctive (and/or) combinations of each other.

Slide 3

Purpose of a composite

• Advantages of a composite:
  • Improved statistical precision through increasing the effective sample size.
  • Simultaneous consideration of more than one component measure.
    • Optimal when more then one component is important.
    • Do not know in advance which component is most important.

Slide 4

Purpose of a composite

• Two mechanisms of performance improvement to achieve the goal of the composite (e.g. reduce post-operative mortality, adverse events, etc.):
  • Reward effort.
    • Quality improvement, pay-for-performance.
  • Recognize ability.
    • Comparative reporting, selective contracting.

Slide 5

AHRQ QI Composites

• Four composite measures:
  • Mortality for selected procedures.
  • Mortality for selected conditions.
  • Patient safety for selected indicators.
  • Pediatric patient safety for selected indicators.
• Developed with the composite workgroup:
  • Reports available on AHRQ QI Web site.

Slide 6

AHRQ QI Composites

• Common methodology:
  • The scale is a reliability-adjusted observed-to-expected ratio.
  • The aggregation approach is a weighted average of these ratios.
  • The weights are user defined, but generally the proportion of numerator events.
  • Overall or for specific populations.

Slide 7

National Quality Forum (NQF)

• Composite Measure Evaluation Framework.
  • NQF Member comments due September 11, 2008, by 6:00 PM ET.
• Importance, scientific acceptability, usability, feasibility.
  • Components NQF-endorsed or assessed to have met the individual measure criteria.

Slide 8

Public Reporting and PPV

• Table shows:
  • Hospital—Patients—Rate—Events.
  • A—1,000—0.050—50.0
  • B—1,000—0.025—25.0
  • Total—2,000—(blank)—75.0

Slide 9

Public Reporting and PPV

• Table shows:
  • Hospital—Patients—Rate—PPV—Events.
  • A—1,000—0.050—0.500—25.0—25.0
  • B—1,000—0.025—0.500—12.5—12.5
  • Total—2,000—(blank)—(blank)—37.5

Slide 10

Public Reporting and Positive Predictive Value (PPV)

• Bar graph shows "Variance in PPV" (percentages here are approximate based on where the tops of bars appear to be).
  • 0.20: 4%
  • 0.30: 13%
  • 0.40: 20%
  • 0.50: 26%
  • 0.60: 20%
  • 0.70: 13%
  • 0.80: 3%

Slide 11

Public Reporting and PPV

• Tables show:
  • Hospital—Patients—Rate—E(PPV)*—Events.
  • A—1,000—0.050—0.500—25.0
  • B—1,000—0.025—0.500—12.5
  • Total—2,000—(blank)—(blank)—37.5
  • Hospital—Patients—Rate—E(PPV)*—Events.
  • A—900—0.050—0.500—22.5
  • B—1,100—0.025—0.500—13.8
  • Total—2,000—(blank)—(blank)—36.3
  • With variance

Slide 12

Public Reporting and PPV

• How good is good enough?
• Goal of improving quality as measured by overall population outcomes.
• Demand elasticity:
  • Change in Q for change in signal.
• Provider rate and variance.
• Relationship between PPV and PPV variance.