Limitations of administrative databases in spine research: a study in obesity

Abstract

Background contextThe use of national inpatient databases for spine surgery research has been increasing. Unfortunately, without firsthand knowledge of each specific database, it can be difficult to judge the validity of such studies. Large databases that rely on administrative data, such as International Classification of Diseases, Ninth Revision (ICD-9) codes, may misrepresent patient information and could thus affect the results of studies that use these data. PurposeThe present study uses obesity, an easily quantified and objective variable, as an example comorbidity to assess the accuracy of ICD-9 codes in the setting of their continued use in spine database studies. Study design/settingA cross-sectional study at a large academic medical center. Patient sampleAll patients spending at least one night in the hospital as an inpatient between April 1, 2013 and April 16, 2013. Obstetrics and gynecology, psychiatry, and pediatric patients were excluded. Outcome measuresProportion of patients for whom ICD-9 obesity diagnosis codes assigned at hospital discharge match chart-documented body mass index (BMI). MethodsThe medical record was reviewed for each patient, and obesity ICD-9 codes were directly compared with documented BMI. ResultsThe study included 2,075 patients. Of 573 “obese” patients (calculated BMI 30–39.9), only 109 received the correct code (278.00), giving this ICD-9 code a sensitivity of 0.19. Of 174 “morbidly obese” patients (calculated BMI >40), only 84 received the correct code (278.01), giving this ICD-9 code a sensitivity of 0.48. ConclusionsUsing obesity as an example, this study highlights the potential errors inherent to using ICD-9–coded databases for spine surgery research. Should a study based on such data use “obesity” as a variable in any analyses, the reader should interpret these results with caution. We further suggest that obesity is likely not the only comorbidity to which these results apply. As database research continues to represent an increasing proportion of publications in the field of spine surgery, it is important to realize that study outcomes can be skewed by data accuracy, and, thus, should not be blindly accepted simply by virtue of large sample sizes.