Generalizability in epidemiology: an investigation within the context of heart failure studies

Purpose Generalizability is characterized by the relevance of a studyʹs results when applied to a larger population. The Dictionary of Epidemiology (J.M. Last et al., 4th ed., 2001) defines generalizability as “unbiased inferences regarding a target population (beyond the subjects in the study).” A goal of public health research is to apply study findings to an external population, and the ability to effectively evaluate generalizability is crucial to the real-world application of results. Current descriptions of generalizability fail to adequately assist in this evaluation; therefore, an alternative conceptual framework is proposed. Generalizability can be assessed using seven key determinants that provide a more systematic evaluation of this concept than can be accomplished by using widely known definitions. Results Generalizability as a construct in textbooks is discussed, as is the formation of our determinants based on these constructs. The seven key determinants are: (i) population definition, (ii) definition of outcome, (iii) recruitment of subjects, (iv) inclusion and exclusion criteria, (v) data collection, (vi) subject retention, and (vii) length of follow-up. The main study designs in epidemiology (ecologic, case–control, cohort, and randomized clinical trial) are used to demonstrate the application of the determinants. In addition, heart failure studies using each type of epidemiologic study design are reviewed for application of the determinants. Concerns regarding the generalizability of the results from all studies are discussed in depth, using the seven key determinants. Strengths and limitations of the generalizability of the findings from these studies are compared in summary tables. Conclusion Generalizability typically pertains only to the external validity of a study, but the key determinants include elements of internal validity. This additional information lends strength to the ability to evaluate a studyʹs results. Short-term and long-term solutions to obtaining and presenting generalizable information on heart failure patients are presented. These solutions can be used to provide clinicians and epidemiologists with more representative information on heart failure populations. This review can serve as a model for the discussion of the determinants of generalizability in other chronic disease studies.