A. Shmelev1, A. M. Sill1, K. W. Shaw1, G. C. Kowdley1, J. A. Sanchez1, S. C. Cunningham1 1Saint Agnes Hospital Center And Cancer Institute,Surgery,Baltimore, MD, USA
Introduction: Accidental punctures and lacerations (APL) during the surgical care of hepatobiliary (HPB) patients are not infrequent and are often preventable. The Agency for Healthcare Research and Quality (AHRQ) Patient Safety Indicator #15 (PSI-15) was created to determine rates of APL. Unfortunately, using the indicator alone may be insufficient to fully understand the nature of these events in a large dataset. Another approach to APL rates is using diagnostic and procedural codes (ICD-9), which may identify these events indirectly. We hypothesized that these two approaches, PSI-15 and a composite measure of relevant ICD-9 codes, would yield different incidences of APL, and that this information can provide a more accurate picture of the frequency and characteristics of APL events.
Methods: Using the National Inpatient Sample from the Healthcare Cost and Utilization Project for 2000–2012, we analyzed all cholecystectomies plus pancreatic and liver resections. All PSI-15 codes were examined and rates were calculated using AHRQ software. Cases were divided in to two groups, those carrying the PSI-15 marker of APL, and those not. We then determined the frequency of all procedural and diagnostic ICD-9 codes shared by both groups, and selected those codes with the most discordance. Pairwise comparisons (chi-square tests) of each selected code against PSI-15 as a surrogate of APL presence were determined. Diagnostic (n=54) and procedural (n=12) codes that significantly increased the odds of having a PSI-positive finding (with 95% CIs that did not include 1) were selected for inclusion into one composite measure (CM) of APL. Both CM and PSI-15 were plotted along a timeline. Seasonal trends decomposition and log-linear Poisson regression analyses were carried out to test for observed significance of trends over time.
Results:The rate of PSI-15 varied from 1.2% in 2000 to a flat maximum of 1.5% in 2007 and back to 1.1% in 2012. Rates of the composite marker gradually increased from 3.1% in 2000) to plateau at 3.7% in 2011. A Poisson regression analysis of both trends, for PSI-15 and for CM, demonstrated statistical significance (p<0.001).
Conclusion:The ICD-9 codes in our CM occurred more frequently than PSI-15. In general, both measures trend similarly over time with CM exhibiting larger variation. The divergence seen between these two curves in 2007 may be associated with known changes in Diagnosis Related Groups at that juncture and, possibly, reimbursement policy involving adverse events. While arguably less specific, CM may increase sensitivity for detecting APL events during HPB operations in comparison to PSI-15. These results may also inform the interpretation of APL studies using either PSI-15 or ICD-9 approaches.
