PS4-39 DEVELOPMENT OF NEW HARM-BASED WEIGHTING APPROACH TO COMPOSITE INDICATORS: THE AGENCY FOR HEALTHCARE RESEARCH AND QUALITY PATIENT SAFETY FOR SELECTED INDICATORS (AHRQ PSI-90)

Wednesday, October 21, 2015
Grand Ballroom EH (Hyatt Regency St. Louis at the Arch)
Poster Board # PS4-39

Kathryn M. McDonald, MM1, Halcyon Skinner, PhD, MPH2, Patricia Zrelak, PhD3, Garth Utter, MD3, Sheryl Davies, MA1, Hawre Jalal, PhD4, Robert Houchens, PhD5 and Patrick Romano, MD, MPH3, (1)Stanford University Center for Health Policy and Center for Primary Care and Outcomes Research, Stanford, CA, (2)Truven Health, Ann Arbor, MI, (3)UC Davis, Sacramento, CA, (4)Department of Health Policy and Management, University of Pittsburgh, Graduate School of Public Health, Pittsburgh, PA, (5)Truven Health, Sacramento, CA
Purpose: Quality measures support many uses. Although targeted measures address quality improvement, composite measures have appeal for consumer choice and payment incentives. Various approaches to weighting composite components are used, but none are patient-centered harm-weighted. Our project aims to develop a composite patient safety indicator (PSI) aggregating different types of harms on a single scale. 

Methods: We developed a composite of 11 AHRQ PSIs combining the volume of each (e.g. post-operative iatrogenic pneumothorax, pressure ulcer), the excess risk of harms associated with each (e.g., mortality, readmission, debridement procedure), and a disutility for each harm. Data sources include the 2012 HCUP SID  for volume, and linked data from the 2012-2013 CMS inpatient and outpatient files for risk-adjusted modeling of excess harms with propensity weights. We multiply excess harms by a utility weight assigned to each harm. When not available in the literature, utility values for harms, were estimated in two steps: 1) elicit relative rankings of all 37 harms from clinicians, 2) fit these rankings to established literature-based utilities reflecting patient preferences.

Results: The table shows final weights, harm weights (excess harm and disutility) and volume weights. Post-operative sepsis has higher mortality rates, and consequently a higher harm weight. The final weight of an indicator is reduced if the volume of that PSI is low (e.g., PSI-08).

Component

Harm Weight

Volume Weight

Final Weight

Pressure Ulcer (PSI-03)

0.0382

0.0294

0.0142

Iatrogenic Pneumothorax (PSI-06)

0.1388

0.0669

0.1175

Central Venous Catheter Related Blood Stream Infection (PSI-07)

0.0627

0.0336

0.0267

Postoperative Hip Fracture (PSI-08)

0.1214

0.0016

0.0025

Perioperative Hemorrhage or Hematoma (PSI-09)

0.0821

0.2868

0.2978

Postoperative Physiologic/Metabolic Derangement (PSI-10)

0.0836

0.0201

0.0213

Postoperative Respiratory Failure  (PSI-11)

0.0479

0.2379

0.1441

Perioperative Pulmonary Embolism/Deep Vein Thrombosis  (PSI-12)

0.0670

0.2479

0.2100

Postoperative Sepsis (PSI-13)

0.2323

0.0511

0.1501

Postoperative Wound Dehiscence (PSI-14)

0.0234

0.0162

0.0048

Accidental Puncture/Laceration (PSI-15)

0.1025

0.0085

0.011

Conclusions: Weighting PSIs in a composite by their associated relative burden or quality of life summarizes total iatrogenic patient-experienced harm, and will compare hospitals accordingly. This method is consistent with a view that decision-making by providers, consumers, and other stakeholders should be driven by an objective of reducing net harm and increasing utility.