Patient Safety Indicators: using administrative data to identify potential patient safety concerns

Health Serv Res. 2001 Dec;36(6 Pt 2):110-32.


Objective: To develop Patient Safety Indicators (PSI) to identify potential in-hospital patient safety problems for the purpose of quality improvement.

Data source/study design: The data source was 2,400,000 discharge records in the 1997 New York State Inpatient Database. PSI algorithms were developed using systematic literature reviews of indicators and hand searches of the ICD-9-CM code book. The prevalence of PSI events and associations between PSI events and patient-level and hospital-level characteristics, length of stay, in-hospital mortality, and hospital charges were examined.

Principal findings: PSIs were developed for 12 distinct clinical situations and an overall summary measure. The 1997 event rates per 10,000 discharges varied from 1.1 for foreign bodies left during procedure to 84.7 for birth traumas. Discharge records with PSI events had twofold to threefold longer hospital stays, twofold to 20-fold higher rates of in-hospital mortality, and twofold to eightfold higher total charges than records without PSI events. Multivariate logistic regression revealed that PSI events were primarily associated with increasing age (p < .001), hospitals performing more inpatient surgery (p < .001), and hospitals with higher percentage of beds in intensive care units (p < .001).

Conclusions: The PSIs provide an efficient and user-friendly tool to identify potential inhospital patient safety problems for targeted institution-level quality improvement efforts. Until better error-reporting systems are developed the PSIs can serve to shed light on the problem of medical errors not limited solely to mortality because of errors.

MeSH terms

  • Adolescent
  • Adult
  • Aged
  • Algorithms
  • Child
  • Child, Preschool
  • Female
  • Hospitals / classification
  • Hospitals / standards*
  • Hospitals / statistics & numerical data
  • Humans
  • Iatrogenic Disease / epidemiology*
  • Infant
  • Infant, Newborn
  • Logistic Models
  • Male
  • Medical Errors / classification*
  • Medical Errors / statistics & numerical data
  • Middle Aged
  • New York / epidemiology
  • Program Development
  • Quality Assurance, Health Care
  • Quality Indicators, Health Care*
  • Safety Management*
  • Sentinel Surveillance*