To determine the accuracy of diagnostic coding of cryptogenic fibrosing alveolitis, the case notes of 166 admissions to four hospitals were reviewed. These consisted of all admissions that had been coded as "idiopathic fibrosing alveolitis" (ICD code 516.3: 97 admissions) or as "postinflammatory pulmonary fibrosis" (ICD code 515.9: 69 admissions). Of 88 available records of admissions coded as idiopathic fibrosing alveolitis, 70 (80%) patients haddefinite cryptogenic fibrosing alveolitis, and six (7%) possible cryptogenic fibrosing alveolitis according to predetermined conventional clinical criteria. Only seven (8%) admissions were clearly coded wrongly. Sixty four records were available for patients coded as having postinflammatory pulmonary fibrosis; 16 (25%) of these patients had definite cryptogenic fibrosing alveolitis, a further 12 (19%) had possible cryptogenic fibrosing alveolitis or fibrosing alveolitis with a connective tissue disorder, and the remainder had a very wide range of diagnoses. In this study the idiopathic fibrosing alveolitis (ICD 516.3) code was relatively reliable, but a substantial proportion of admissions coded under postinflammatory pulmonary fibrosis (ICD 515.9) also had cryptogenic fibrosing alveolitis and code 515.9 was of little diagnostic value. The data are inadequate for case finding, though in respect of cryptogenic fibrosing alveolitis may be adequate for planning purposes. There continues to be a need for more medical input into the process of diagnostic coding.