A Validity Review of the National Burn Repository

Abstract

Objective Health-related registries arose due from clinician desire to improve patient quality of care for a specific disorder. As such, disease registries differ from administrative registries in concept, organization, purpose, data recording, and results. Due to their voluntary nature, health-related disease registries are not regularly audited, have a narrow focus, and are designed for clinicians, not administrators. As part of a Department of Defense initiative we conducted an intensive qualitative review of the American Burn Association’s National Burn Repository (NBR). Our objectives are to inform future users of the NBR of issues that could affect statistical analyses and inferences and assist efforts to improve data collection. Methods We obtained a deidentified copy of the 2009 release of the NBR containing 286,293 records. We reviewed this data set for 1) records lacking vital patient information (age, burn size, survival, gender), 2) inconsistencies between data in different fields of the database, and 3) duplicate values. Results Restricting our review to records with an admission year of 2000 or later, vital patient information was missing or invalid for about 60,000 records. Data inconsistencies were found in hospital admission status (initial admission or readmission) for about 12,000 records, survival for about 950 records, and burn injury for about 5,500 records. Depending on the criteria used to identify duplicate records, we found at least 4,000 duplicate records but as many as 14,000 in the database. Finally, significant data quality issues were found for facilities not using the TRACS software. Conclusions All health-related disease registries, unlike administrative databases, are voluntary. Anonymity of data is vital, and data auditing and reporting are challenging. The data contained in the NBR is disease-specific, and, as such, has the potential to provide valuable epidemiologic, treatment, and outcome data as reported by clinicians, not registrars. The NBR provides substantive data on burn injury; however, data review needs to precede data analysis. Revisions to NBR data collection have improved the quality of data submitted, yet data quality issues remain in the current database. Investigators are cautioned to thoroughly assess all fields prior to conducting analyses using the NBR.

Predicting mortality from burns: The need for age-group specific models

“…We also removed 3,218 records identified as probable duplicates, 6,529 records with unreliable information (e.g., total burn surface area greater than 100, records from facilities with questionable ages or mortality rates), 23,084 records associated with readmissions, and 3,690 records of patients with non-burn injuries. [14] This validation left 100,051 records of initial hospital visits (admissions and outpatient visits) with the minimum information for necessary analysis (i.e., patient age, a burn or inhalation injury, and hospital discharge status). We used mixed effect logistic regression models to evaluate the effect of patient age, TBSA burn and the presence of inhalation injury on mortality.…”

Risk Factors for In-Hospital Mortality in Smoke Inhalation-Associated Acute Lung Injury

“…3,7,10 The National Burn Repository of the American Burn Association (ABA) provides comprehensive data on burn victims nationally at burn centers, but is limited by missing data and duplicate records. 7,11 The Nationwide Inpatient Sample, an administrative data source covering one in five US nonfederal hospital discharges lacks details on some key variables. 3 Recent quality improvement-targeted enhancements to some administrative databases, such as encounter-level, datestamped interventions, have facilitated benchmarking, surveillance, and comparative-effectiveness research.…”

Predicting Resource Utilization in Burn Treatment

“…We also removed 6,530 records with unreliable information (e.g., total burn surface area greater than 100, records from facilities with questionable ages or mortality rates), 23,084 records associated with readmissions, 3,690 records of patients with non-burn injuries, and 3,218 records identified as probable duplicates. (2) This screening left 95,579 records of initial hospital visits (admissions and outpatient visits). From this data set, we selected records with valid entries for number of operative procedures (47,792), total operating room visits (48,068), ICU days (77,511), and ventilator days (78,652) in analyzing each of these resource use variables.…”

“…These parameters are 1) the probability of any resource use, 2) the estimated mean level of resource and 3) the estimated mean level of resource use given that the patient has some resource use. The following details how to calculate each of these outcomes.…”