NSQIP reveals significant incidence of death following discharge

The rates of post-discharge deaths after surgical procedures are unknown and may represent areas of quality improvement. The NSQIP database captures 30-d outcomes not included within normal administrative databases, and can thus differentiate between in-hospital and post-discharge deaths.

Retrospective analysis of NSQIP from 2005 through 2007. Inclusion criteria were procedures whose median length of stay was greater than 1 d (to exclude outpatient procedures), and whose overall death rate was greater than 2% (to include only procedures where mortality was a significant issue). Procedures where less than 25 deaths occurred were excluded (for sample size concerns).

There were 363,897 patients with 2236 different CPT codes captured in NSQIP. There were 6395 deaths; among them, 1486 (23.2%) occurred after discharge. Thirty-eight CPT codes met the analysis threshold. In two of the CPT codes, there were no post-discharge deaths (repair of ruptured abdominal aortic aneurysm [AAA], repair of ruptured AAA involving iliacs). In the other 36 CPT codes, the proportion of deaths occurring after discharge ranged from 6.3% (repair of thoracoabdominal aneurysm) to 50.0% (femoral-distal bypass with vein). The highest percentage of post-discharge mortality occurs on d 1 after discharge. Fifty percent of post-discharge mortality occurs by d 7; 95% occurs by d 21.

Approximately one-fourth of postoperative deaths occur after hospital discharge. There is significant variation across surgical procedures in the likelihood of postoperative deaths occurring after discharge. These data indicate a need for closer and more frequent monitoring of post-surgical patients. These data also call into question conclusions drawn from hospital-based outcomes analyses for at least some key diseases/procedures. This analysis demonstrates the power of the risk-adjusted 30-d follow-up NSQIP data, but perhaps more importantly, the responsibility of surgeons to monitor and optimize the discharge process.