Classification of respiratory functional impairment in chronic obstructive pulmonary disease

Starting from a paper published in 1964 by Wilson et al., we explored the possibility of classifying the clinical and functional deficit of patients with chronic obstructive lung disease into six classes, class 0 representing normality and class 5 greatest severity. Each symptom or sign was classified into six degrees of increasing severity. Next, we looked for a possible dependence of the collegially assigned score on anthropometric, clinical, or instrumental data in each case. More particularly, we tried (1) to identify such combinations of variables as would permit classification of the patient with the smallest possible error, and (2) to determine which of the variables reflected the severity of the case more faithfully. The results emerging from this study suggest the possibility of evaluating and classifying respiratory impairment in three different ways, as follows: (1) On the basis of clinical data only. This method is the easiest to use and affords a fairly good determination coefficient (R2 = 0.812). (2) Using only some combinations of laboratory data (static and dynamic pulmonary volumes, blood gases, etc.), with or without the addition of vital statistics and anthropometric data. These subensembles would allow a posteriori estimates in cases where the subject is no longer available for questioning and examination. In that case the best multiple regression affords a determination coefficient R2 = 0.82. (3) Using all clinical and laboratory data available. In that case, the best multiple regression (R2 = 0.899) for predictive purposes is that which includes the sum of clinical data, the pulmonary volumes before and after pharmacological bronchodilation, and the PaCO2 value. For practical purposes, however, the most convenient function is the one that includes the sum of clinical data plus FEV1 and RV (R2 = 0.863). Even with the best of the three functions proposed in this paper, however, the standard error of estimate entails tolerance limits sometimes amounting to one whole class of severity. Still, the probability of making an error exceeding one class of severity occurs in only 3.7% of the cases, an average which seems quite acceptable from the clinical point of view.