Identifying Patient Safety Indicators from Administrative Data
Amy K. Rosen PhD
VA New England Health Care System
Bedford, MA
Funding Period: October 2002 - September 2005

BACKGROUND/RATIONALE:
Ensuring patient safety has become a high priority for all health care systems. Little is known, however, about the number of patient safety events that typically occur during a hospitalization. Administrative data are known to be a practical method for identifying in-hospital complications, and may therefore also be useful as a method for identifying patient safety events. In this study, we assess the feasibility of using administrative data from the Department of Veterans Affairs to calculate Patient Safety Indicators, known as PSIs, that were developed by the Agency for Healthcare Research and Quality (AHRQ) and modified by the Evidence-based Practice Center at Stanford University. The VA is an ideal setting in which to explore this because it contains rich clinical information on patients as well as documented information on patients' utilization.

OBJECTIVE(S):
Our specific objectives were to assess the ability of VA administrative data to identify potential instances of compromised patient safety in the inpatient acute care setting, to determine the construct validity of PSIs based on VA inpatient data, to validate PSIs using other VA quality data (National Surgical Quality Improvement Program, NSQIP), to identify factors that explain variation in PSI rates, to compare VA and non-VA PSI rates, and to explore variation in safety practices across VA facilities with low vs. high PSI rates.

METHODS:
We developed algorithms and other methods to adapt VA Patient Treatment File (PTF) data for use with PSI software. We obtained NSQIP data and developed methods to match NSQIP and PTF data for assessment of PSI sensitivity and specificity. We used hierarchical Poisson and Bayesian methods to control for statistical artifacts of the inherently low rates of PSI events when comparing event rates across VA facilities. We obtained American Hospital Association and VA NQIS (National Quality Improvement Survey) data and analyzed facility-level predictors of variation in PSI rates. We worked with AHRQ to set up comparable VA and non-VA databases in order to compare PSI rates. We developed an interview protocol, recruited four VA facilities for site visits, and initiated site visits to explore variation in safety practices.

FINDINGS/RESULTS:
A paper on adapting VA administrative data for use with PSIs apperared in an AHRQ compendium on patient safety early in 2005. A paper evaluating use of the PSIs on VA data appeared in Medical Care September 2005. Preliminary study results were presented to the VA National Center for Patient Safety in Fall of 2004. Extensive analysis of the statistical properties of the VA PSI rates and associated variables supported our propositions that VA administrative data are a rich source of information on safety and quality, and that they can be used to identigy potenial instances of compromised patient safety. Hospitalizations with PSI events are consistenly associated with greater lengths of stay, higher estimated costs, and higher mortality rates, compared to hospitalizations without PSI events. We identified 0,974 PSI events in the VA nationwide in FY '01 and 12,165 in FY '04. Observed PSI rates per 1,000 discharges ranged from 0.007 ro "transfusion reaction" to 126.75 for "failure torescue" in FY'04. Comparisons of VA and non-VA risk-adjusted PSI rates showed that VA rates were slightly higher for 10 out of 15 relevant PSIs. Hospitalizations with PSI events had greater lengths of stay, higher mortality, and higher costs than those without PSI events. We found meaningful variation in rates across VA facilities and VISNs for most PSIs. Certain surgical PSIs have moderately high levels of sensitivity when compared with NSQIP.

IMPACT:
These findings may affect the VA’s future patient safety initiatives at the national, VISN, and/or facility levels.

PUBLICATIONS:

Journal Articles

1. Rivard PE, Luther SL, Christiansen CL, Shibei Zhao, Loveland S, Elixhauser A, Romano PS, Rosen AK. Using patient safety indicators to estimate the impact of potential adverse events on outcomes. Medical Care Research and Review. 2008; 65(1): 67-87.
2. Singer S, Meterko M, Baker L, Gaba D, Falwell A, Rosen AK. Workforce perceptions of hospital safety culture: development and validation of the patient safety climate in healthcare organizations survey. Health Services Research. 2007; 42(5): 1999-2021.
3. Rosen AK, Zhao S, Rivard P, Loveland S, Montez-Rath ME, Elixhauser A, Romano PS. Tracking rates of Patient Safety Indicators over time: lessons from the Veterans Administration. Medical Care. 2006; 44(9): 850-61.
4. Rivard PE, Rosen AK, Carroll JS. Enhancing patient safety through organizational learning: Are patient safety indicators a step in the right direction? Health Services Research. 2006; 41(4 Pt 2): 1633-53.
5. Rosen AK, Rivard P, Zhao S, Loveland S, Tsilimingras D, Christiansen CL, Elixhauser A, Romano PS. Evaluating the patient safety indicators: how well do they perform on Veterans Health Administration data? Medical Care. 2005; 43(9): 873-84.
6. Nelson A, Weaver FM. Promoting evidence-based practice in spinal cord injury/disorders health care. Sci Nursing. 2004; 21(3): 129-135.
7. Tsilimingras D, Rosen AK, Berlowitz DR. Patient safety in geriatrics: a call for action. The Journals of Gerontology Series A, Biological Sciences and Medical Sciences. 2003; 58(9): M813-9.
8. Reker DM, Rosen AK, Hoenig H, Berlowitz DR, Laughlin J, Anderson L, Marshall CR, Rittman M. The hazards of stroke case selection using administrative data. Medical Care. 2002; 40(2): 96-104.

DRA: Health Services and Systems
DRE: Quality of Care
Keywords: Adverse events, Organizational issues, Safety
MeSH Terms: none