Validation of expert opinion in identifying comorbidities associated with atopic dermatitis/eczema

Abstract

Background: The use of expert opinion is widespread in economic studies of healthcare utilisation; however, few studies have attempted to assess the validity of assumptions derived from such sources.

Objective: To examine the use of such expert opinion in determining comorbidities associated with atopic dermatitis/eczema (AD/E), which were assessed as part of a recent third-party payer cost-of-illness study.

Design: To identify the disease-related comorbidities that would represent costs associated with AD/E, physicians on an expert panel were asked individually and then collectively to group all International Classification of Diseases, 9(th) Edition-Clinical Modification (ICD-9-CM) diagnosis codes as 'most likely', 'possibly' or 'definitely not' related to the costs of identifying and treating patients with AD/E. Claims representing $US464 million in payer reimbursements from nearly 125 000 patients with AD/E were identified within two separate claims databases (1997 values). Over 850 ICD-9-CM diagnosis codes were identified in the first-listed position from these claims. For each group of 'most likely', 'possibly' and 'definitely not' related diagnosis codes, prevalence rates were compared within AD/E and non-AD/E populations from the two historical payer claims databases. Adjusted and non-adjusted odds ratios were calculated by comparing prevalence rates between AD/E and non-AD/E patients in the same payer population.

Results: The mean prevalence rate of any diagnosis code in the AD/E population was 0.65 +/- 1.82% (SD) with a mean odds ratio of 1.81 +/- 0.96. Comorbidities considered by the expert panel 'most likely' to be associated with AD/E had higher prevalence rates (3.28 +/- 3.63%) and odds ratios (2.14 +/- 1.14). Comorbidities considered to be 'possibly' related to AD/E had prevalence rates and odds ratios of 3.01 +/- 5.06% and 1.84 +/- 0.82, respectively. Comorbidities considered to be 'definitely not' related to AD/E had the lowest prevalence rates (0.45 +/- 1.09%) and odds ratios (1.80 +/- 0.97).

Conclusions: Comparing the result of consensus panels with actual claims histories validated the use of expert opinion in determining comorbidities associated with AD/E. Expert opinion yielded valid results in terms of identifying comorbidities that manifested frequently and disproportionately in the AD/E population. Limited statistical measurements of comorbidities would have been less specific than expert opinion. Future cost-of-illness studies should consider alternative data sources and methodologies to enhance the validity and importance of expert opinion and to corroborate their findings.