Objective: We assessed the frequency that patients are incorrectly used as the unit of analysis among studies of physicians’ patient care behavior in articles published in high impact journals.
Methods: We surveyed 30 high-impact journals across 6 medical fields for articles susceptible to unit of analysis errors published from 1994 to 2005. Three reviewers independently abstracted articles using previously published criteria to determine the presence of analytic errors.
Results: One hundred fourteen susceptible articles were found published in 15 journals, 4 journals published the majority (71 of 114 or 62.3%) of studies, 40 were intervention studies, and 74 were noninterventional studies. The unit of analysis error was present in 19 (48%) of the intervention studies and 31 (42%) of the noninterventional studies (overall error rate 44%). The frequency of the error decreased between 1994–1999 (N = 38; 65% error) and 2000–2005 (N = 76; 33% error) (P = 0.001).
Conclusions: Although the frequency of the error in published studies is decreasing, further improvement remains desirable.
From the *Department of Pediatrics, Critical Care Division, University of Louisville, Louisville, Kentucky; Departments of †Clinical Epidemiology and Biostatistics and ‡Medicine, McMaster University; Departments of §Pediatrics and ¶Epidemiology and Biostatistics, University of California, San Francisco; ∥Department of Anesthesia, Critical Care Division, Children's Hospital, Boston, Massachusetts; and **Department of Anaesthesia, Harvard Medical School, Boston, Massachusetts.
Reprints: Aaron W. Calhoun, MD, University of Louisville, Pediatric Critical Care, 571 S. Floyd St., Suite 332, Louisville, KY 40202. E-mail: email@example.com.