Predicting walking-induced oxygen desaturations in COPD patients: A statistical model

BACKGROUND: Oxygen desaturation during walking can have important consequence on prognosis of COPD patients. However, a standard 6-min walk test (6MWT), useful in detecting desaturation in COPD patients, can be difficult to execute in some settings of COPD management, as in the community healthcare service. We evaluated a new scoring system for the risk of oxygen desaturation during walking in COPD patients: the walking desaturation score. METHODS: We collected data from symptomatic COPD in-patients admitted for rehabilitation (derivation cohort) and out-patients referred to the local community health service (validation cohort). S_pO2 was monitored during 6MWT, and the subjects were classified as walking desaturators or non-desaturators. By a regression analysis model we assigned a weighted score proportional to the measured percentage of explained variance for each variable. Risk estimates were computed as odds ratios. A receiver operating characteristic curve analysis and a Hosmer-Lemeshow goodness-of-fit test were then performed to measure discrimination and calibration of walking desaturation score. RESULTS: Baseline characteristics in the derivation cohort (n= 435, 74% of whom were walking desaturators) and the validation cohort (n =238, 37% of whom were walking desaturators) were different. Resting arterial oxygen saturation measured from an arterial blood sample, P_aO2, and percent-of-predicted FEV₁ were the variables that predicted walking desaturation. The proportion of walking desaturators (and odds ratio estimate) gradually increased according to walking desaturation score (range 0-6) and associated categories of desaturation risk (total walking desaturation score: low 0 or 1, high 2-3, very high 4-6) (chi-square P