89 CASE REPORTS IN EMERGENCY MEDICINE RESEARCH: ARE THE METHODS VALID?

Purpose of StudyAlthough randomized clinical trials and systematic reviews provide the “best evidence” for guiding medical practice, many peer-reviewed medical journals still publish case reports (CRs). Yet there are no accepted methodologic standards for CRs. We sought to determine the proportion of treatment-related CRs that adequately reported their methods, including critical information about the patient, disease, interventions, cointerventions, and measured outcomes.Methods UsedCRs published in four emergency medicine journals from 2003-2005 were identified and categorized according to their purpose (disease description, report of an adverse drug reaction, utility of a diagnostic test, or treatment effect). Treatment-related CRs, including medical and surgical interventions, were reviewed for adherence to 11 methodologic standards. CRs were re-reviewed by a second investigator; interrater agreement was measured with kappa statistics, and disagreements were resolved by consensus.Summary ofResultsOf 646 CRs identified, 37 (5.7%; 95% CI = 0.04-0.08) were treatment CRs. Over half (54%; 95% CI = 0.38-0.69) of the treatment CRs described interventions for a poisoning or overdose. A complete, searchable title, including disease, intervention, and outcome, was present in 23 CRs (62%; 95% CI = 0.46-0.76). A clear diagnostic standard (pathologic, laboratory, or radiologic test or clinical detail) was present in 35 CRs (95%; 95% CI = 0.82-0.98). Most CRs included clear descriptions of disease severity (84%; 95% CI = 0.69-0.92); intervention (medication dose or procedure details) (83%; 95% CI = 0.69-0.92); and all clinically relevant outcomes (75%; 95% CI = 0.59-0.87). Other methods were seldom mentioned: comorbidities (48%; 95% CI = 0.33-0.64); cointerventions (35%; 95% CI = 0.22-0.51); and treatment side effects (35%; 95% CI = 0.22-0.51). Only 22% (95% CI = 0.11-0.37) mentioned key criteria for inferring causality (biologic plausibility, consistency with published literature, and exclusion of “natural history” or other confounding factors). Generalizability of treatment effects to other patients was mentioned in seven CRs (19%; 95% CI = 0.09-0.34). No CR reported its “denominator” (number of patients subjected to the same intervention, whether or not successful).ConclusionsTreatment CRs are published frequently in the emergency medicine literature, but information about the data quality is often lacking. CRs should be held to higher methodologic standards, or