Disease-based mortality after percutaneous endoscopic gastrostomy: utility of the enterprise data warehouse

Abstract

Background

Percutaneous endoscopic gastrostomy (PEG) remains a mainstay of enteral access. Thirty-day mortality for PEG has ranged from 16 to 43 %. This study aims to discern patient groups that demonstrate limited survival after PEG placement. The Enterprise Data Warehouse (EDW) concept allows an efficient means of integrating administrative, clinical, and quality-of-life data. On the basis of this concept, we developed the Vanderbilt Procedural Outcomes Database (VPOD) and analyzed these data for evaluation of post-PEG mortality over time.

Methods

Patients were identified using the VPOD from 2008 to 2010 and followed for 1 year after the procedure. Patients were categorized according to common clinical groups for PEG placement: stroke/CNS tumors, neuromuscular disorders, head and neck cancers, other malignancies, trauma, cerebral palsy, gastroparesis, or other indications for PEG. All-cause mortality at 30, 60, 90, 180, and 360 days was determined by linking VPOD information with the Social Security Death Index. Chi-square analysis was used to determine significance across groups.

Results

Nine hundred fifty-three patients underwent PEG placement during the study period. Mortality over time (30-, 60-, 90-, 180-, and 360-day mortality) was greatest for patients with malignancies other than head and neck cancer (29, 45, 57, 66, and 72 %) and least for cerebral palsy or patients with gastroparesis (7 % at all time points). Patients with neuromuscular disorders had a similar mortality curve as head and neck cancer patients. Stroke/CNS tumor patients and patients with other indications had the second highest mortality, while trauma patients had low mortality.

Conclusions

PEG mortality was much higher in patients with malignancies other than head and neck cancer compared to previously published rates. PEG should be used with great caution in this and other high-risk patient groups. This study demonstrates the power of an EDW-based database to evaluate large numbers of patients with clinically meaningful results.