Abstract
Background
Existing sources of information on hospital-associated venous thromboembolism (HA-VTE) in the United States have important limitations. Key challenges include distinguishing probable or confirmed from possible cases of VTE; distinguishing new from recurrent VTE; and identifying events diagnosed after hospital discharge. Two types of administrative healthcare data are commonly used for estimates of HA-VTE in inpatients: hospital discharge databases and health insurance claims databases. Analyses of both types of data cannot confirm VTE diagnoses from medical records or reliably assess the timing of onset to distinguish postoperative or HA-VTE events. In addition, hospital discharge databases are limited to diagnoses occurring prior to discharge (before or during hospitalization). Although health insurance databases include outpatient records, the reliability of outpatient records is unclear. Because of the higher rate of HA-VTE among surgical patients, efforts to prevent and monitor HA-VTE often focus on postoperative VTE, which is the approach taken here.
Methods
This study used electronic health record (EHR) data from the Veterans Health Administration (VHA) to quantify the frequency of postoperative VTE within 30 and 90 days post-surgery among inpatient admissions of surgical patients at VHA hospitals during 2005-2010. Records were restricted to VHA surgical admissions of patients who had no record of a VTE event within 365 days preceding a surgery and were alive at either 30 or 90 days post-surgery without a repeat surgery. Inpatient VTE events were identified using diagnosis codes while outpatient VTE events were identified using a combination of diagnosis codes, procedure codes, pharmacy records, and the narrative text of EHR clinical notes. A natural language processing (NLP) system was developed to automatically find evidence of acute VTE events based on mentions in clinical notes. To confirm an outpatient event, we required within 14 days after the VTE diagnosis either a prescription for an anticoagulant or a CPT code for thrombectomy, embolectomy, vena cava filter placement, or thrombolysis and a positive finding of the event in the patient’s narrative clinic notes identified using an NLP tool. For each VTE, we distinguished whether it was diagnosed (1) post-surgery but pre-discharge, (2) post-discharge in a VHA outpatient setting, or (3) post-discharge in a VHA inpatient setting (readmission). Admissions were classified into 1 of 3 mutually exclusive types of surgery (1) major orthopedic (total knee or hip replacement or hip fracture surgery), (2) abdominal-pelvic, and (3) other.
Results
A total of 648,851 inpatient admissions occurred nationwide during 2005-2010 at one of 114 VHA facilities, of which 442,410 (from 363,545 unique patients) and 420,858 (from 347,794 unique patients) met the inclusion criteria for 30 and 90 days post-surgery, respectively. VTEs were documented in 3,845 (0.87%) and 5,383 (1.3%) surgical admissions where the patient was alive at 30 or 90 days, respectively. Postoperative VTEs occurred before discharge following 2,140 (0.48%) surgeries tracked through 30 days. Roughly one-half of postoperative VTEs were diagnosed after discharge, 44.3% of 30-day postoperative VTE events and 61.7% of 90-day postoperative VTE events. A VTE was diagnosed within 90 days of surgery in 1.9% of orthopedic surgery admissions, followed in frequency by abdominal-pelvic (1.4%), and other surgeries (1.2%). Among 3,483 VTE diagnoses identified post-discharge, 2,676 (76.4%) resulted in a VHA hospital readmission within 90 days of surgery, accounting for 3.4% of 78,473 90-day readmissions.
Conclusion
This study makes three key methodological contributions for identifying HA-VTE. First, detailed VHA inpatient data allowed us to isolate VTE events that occurred after the patients’ surgery and to exclude VTEs present prior to surgery. Second, these data allowed us to track HA-VTE events occurring up to 90 days following surgery. Third, we harnessed the information in unstructured narrative text using an NLP tool to verify an outpatient VTE diagnosis. To monitor HA-VTE, it is essential to track patients after discharge to identify potential VTE events diagnosed in outpatient settings; this study confirms previous findings that 40-50% of postoperative or HA-VTE events are diagnosed after hospital discharge.
Disclosures:
No relevant conflicts of interest to declare.
ICD-10 Coding of Musculoskeletal Conditions in the Veterans Health Administration