Complication Rates for Pediatric Hepatectomy and Nephrectomy: A Comparison of NSQIP-P, PHIS, and KID.

BACKGROUND

Three large national data sets are commonly used to assess operative outcomes in pediatric surgery; National Surgical Quality Improvement Program Pediatric (NSQIP-P), Pediatric Health Information System (PHIS), and Kids' Inpatient Data set (KID). Hepatectomy and nephrectomy are rare pediatric surgical procedures, which may benefit from large administrative data sets for the assessment of short-term complications.

MATERIALS AND METHODS

A retrospective review of NSQIP-P (2012-2015), KID (2012), and PHIS (2012-2015) was performed for hepatectomy or nephrectomy cases for children aged 0 to 18 y. Thirty-day perioperative outcomes were collected, analyzed, and compared across data sets and surgical cohorts.

RESULTS

Rates of surgical site infection, wound dehiscence, central line infection, sepsis, and venous thromboembolism were similar across NSQIP-P, PHIS, and KID in both cohorts. Rates of pneumonia and renal insufficiency were higher in PHIS and KID versus NSQIP-P in both cohorts. Blood transfusions in NSQIP-P were higher than PHIS and KID in the hepatectomy group (50.9% versus 43.0% versus 32.4%, P < 0.001), but similar across data sets in the nephrectomy cohorts (12.0% versus 14.0% versus 13.0%, P = 0.15). PHIS reported higher readmission rates than NSQIP-P for both the hepatectomy (56.5% versus 17.9%, P < 0.001) and nephrectomy (32.6% versus 7.6%,P < 0.001) cohorts. Thirty-day mortality rates were similar between NSQIP-P and PHIS, but higher in KID as compared with NSQIP-P for hepatectomy (6.4% versus 0.4%, P < 0.001) and nephrectomy (2.0% versus 0.3%, P < 0.001) cases.

CONCLUSIONS

Administrative data sets provide large sample sizes for the study of low-volume procedures in children, but there are significant variations in the reported rates of perioperative outcomes between NSQIP-P, PHIS, and KID. Therefore, surgical outcomes should be interpreted within the context of the strengths and limitations of each data set.