Exercise-induced oxygen desaturation (EID) is associated with increased risk of mortality in chronic obstructive pulmonary disease (COPD). Several screening tests have been proposed to predict EID, including FEV1, DLCO and baseline-SpO2. We aimed to validate a proposed cut-off of baseline-SpO2 </=95% as simple screening procedure to predict EID during six-minute walk test (6MWT). In addition, we studied the prevalence and characteristics of patients exhibited EID to SpO2nadir </=88%. 402 non-hypoxemic COPD patients performed 6MWT. Sensitivity and specificity of baseline SpO2 </=95% as a cut-off to predict EID and determinants of EID were investigated. 158 patients (39%) exhibited EID. The sensitivity of baseline-SpO2 </=95% to predict EID was 81.0%, specificity 49.2%, positive and negative predictive values were 50.8% and 80.0%, respectively. In a multivariate model, DLCO <50%, FEV1 <45%, PaO2 <10kPa, baseline-SpO2 <95%, and female sex were the strongest determinants of EID. Baseline oxygen saturation solely is inaccurate to predict EID. A combination of clinical characteristics (DLCO, FEV1, PaO2, baseline-SpO2, sex) increases the odds for EID in COPD.