|
Back to Annual Meeting Program
National Database Research Beyond ICD-9 Codes: Evaluating Post-Pancreatectomy Diabetes Using Pharmacy Claims
Elan R. Witkowski*1, Elizaveta Ragulin-Coyne1, Zeling Chau1, Sing Chau NG1, Heena P. Santry1, Shimul a. Shah1, Jennifer F. Tseng1,2
1Surgical Outcomes Analysis & Research (SOAR), University of Massachusetts Medical School, Worcester, MA; 2Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA
Background: Large databases are central tools in surgical research, but the utility of certain diagnosis and procedure codes may be limited. Diabetes can occur after pancreatic cancer (PC) resection, but being insulin-dependent vs. non-insulin-dependent may impact quality of life more than the simple diabetes/no diabetes dichotomy would suggest. We demonstrate novel use of the newly available Medicare Part D pharmacy claims as an improved method of describing the presence and clinical impact of post-pancreatectomy diabetes mellitus.
Methods: Patients (pts) who underwent PC resection between 5/1/07 and 5/1/08 were identified in the SEER-Medicare linked database, including Part D. Neuroendocrine pts, pts who died during their index hospitalization, and pts without continuous Part D were excluded. Claims for insulin and oral hypoglycemic medications were identified during three intervals: (1) four months prior to operation until the day prior to operation, (2) postoperative day 0-60, and (3) >60 days after operation. Based on preoperative diabetes status, probabilities of developing oral-med-dependent or insulin-dependent DM were calculated. These results were compared to results determined by ICD-9 diagnosis code.
Results: After screening 455 patients, a cohort of 123 patients met the inclusion criteria. Using ICD9 codes, 53 patients (43.1%) were identified as having preoperative diabetes, of whom 37(30.1%) had preop claims for diabetic medications. All patients who received treatment had an ICD9 diabetes diagnosis coded. Among these patients, 29(23.6%) took oral hypoglycemics and 19(15.5%) took insulin preoperatively, with some overlap between groups. Postoperative diabetes (>60 days) was examined in 114 patients with survival sufficient for analysis. Using ICD9 codes, 75 patients (65.8%) were identified as diabetic. Part D claims indicated that 67(58.8%) never required medication, 19(16.7%) required only oral hypoglycemics, and 28(24.6%) insulin. Only 13 patients (11.4%) developed a new insulin requirement.
Conclusions: Administrative database research is limited by the quality and clinical relevance of available data. Our pilot study demonstrates a novel use of outpatient pharmacy claims to define medication-dependent diabetes after pancreatic cancer resection. The utilization of pharmacy claims may augment researchers’ ability to detect the presence of various diseases and ascertain both their clinical relevance and potential quality of life impact.
Back to Annual Meeting Program
|
