Cost-of-illness studies : a review of current methods.

The number of cost-of-illness (COI) studies has expanded considerably over time. One outcome of this growth is that the reported COI estimates are inconsistent across studies, thereby raising concerns over the validity of the estimates and methods. Several factors have been identified in the literature as reasons for the observed variation in COI estimates. To date, the variation in the methods used to calculate costs has not been examined in great detail even though the variations in methods are a major driver of variation in COI estimates. The objective of this review was to document the variation in the methodologies employed in COI studies and to highlight the benefits and limitations of these methods. The review of COI studies was implemented following a four-step procedure: (i) a structured literature search of MEDLINE, JSTOR and EconLit; (ii) a review of abstracts using pre-defined inclusion and exclusion criteria; (iii) a full-text review using pre-defined inclusion and exclusion criteria; and (iv) classification of articles according to the methods used to calculate costs. This review identified four COI estimation methods (Sum_All Medical, Sum_Diagnosis Specific, Matched Control and Regression) that were used in categorising articles. Also, six components of direct medical costs and five components of indirect/non-medical costs were identified and used in categorising articles.365 full-length articles were reflected in the current review following the structured literature search. The top five cost components were emergency room/inpatient hospital costs, outpatient physician costs, drug costs, productivity losses and laboratory costs. The dominant method, Sum_Diagnosis Specific, was a total costing approach that restricted the summation of medical expenditures to those related to a diagnosis of the disease of interest. There was considerable variation in the methods used within disease subcategories. In several disease subcategories (e.g. asthma, dementia, diabetes mellitus), all four estimation methods were represented, and in other cases (e.g. HIV/AIDS, obesity, stroke, urinary incontinence, schizophrenia), three of the four estimation methods were represented. There was also evidence to suggest that the strengths and weaknesses of each method were considered when applying a method to a specific illness. Comparisons and assessments of COI estimates should consider the method used to estimate costs both as an important source of variation in the reported COI estimates and as a marker of the reliability of the COI estimate.