Abstract
Purpose: :
Large-scale administrative databases, such as those operated by Medicare and health maintenance organizations, are often used to support clinical research and process improvement in ophthalmology and other fields. Clinical databases, such as the Vermont Oxford Network (VON), contain data from chart review by trained health professionals and have a higher level of quality control. These databases represent secondary sources of original medical data and are therefore subject to error. The purpose of this study was to measure the accuracy of the VON database (http://www.vtoxford.org), a widely used research and quality assurance collaborative network among over 700 neonatal intensive care units (NICU) throughout the world, for characterization of retinopathy of prematurity (ROP) examination findings among hospitalized premature infants at one participating center.
Methods: :
Retrospective chart review was performed to identify all unique infants hospitalized at the Columbia University NICU from 1/1/2002 to 12/31/2005 who received ROP screening exams. Data were collected on ROP parameters in the format used by the VON database: most severe stage in either eye, and whether laser photocoagulation or other surgical treatment was performed in either eye. Chart review findings were compared to patient-matched VON entries.
Results: :
445 patients received ROP screening exams at the study center during this period, of whom 414 met criteria for VON participation based on birth weight and age at hospital admission. Among these 414 study infants, 293 (70.8%) had accurate ROP exam data recorded in the VON database, and 121 (29.2%) had errors in ROP exam data. Of these errors, 79 (19.1%) involved ROP stage, 13 (3.1%) involved ROP treatment, and 35 (8.5%) involved complete absence of ROP exam data.
Conclusions: :
There were substantial inaccuracies regarding ROP exam data from the study center in the VON database. This raises potential concerns about the use of administrative and clinical databases for medical research, quality management, and process review involving ROP and other diseases. Further work is required to determine whether this finding is generalizable across multiple study centers.
Keywords: retinopathy of prematurity • clinical (human) or epidemiologic studies: biostatistics/epidemiology methodology • clinical (human) or epidemiologic studies: prevalence/incidence