Figure 1 shows a time series analysis of mean respiratory frequency in one patient who required hospitalization because of ECOPD during follow-up wherein, compared with baseline (clinical stability), mean respiratory frequency increased progressively during the 5 days that preceded hospitalization. A similar upward trend was observed in 21 of the 30 patients (70%) (C statistic, P < .05). Analysis of variance confirmed that the mean respiratory frequency was significantly higher during the 5 days that preceded hospitalization than at baseline (19.1 ± 5.9/min vs 15.2 ± 4.3/min, respectively). This was not the case in patients who did not require hospitalization because of ECOPD (16.1 ± 4.8/min vs 15.9 ± 4.9/min, baseline vs 5 random days, respectively) (Figs 2A, 2B, respectively). ROC analysis comparing the mean daily respiratory frequency of the 30 patients who required hospitalization because of ECOPD with that of those 59 who did not showed that the area under the curve was 0.79 (P < .05) and 0.76 (P < .05) for the absolute increase of mean breathing rate between baseline and that of 24 h or 48 h before hospitalization, respectively (Fig 3). Numerically, an increase of 4.4/min provided the best combination of sensitivity and specificity (66%; 95% CI, 48%-80%; and 93%; 95% CI, 84%-97%, respectively) with a positive predictive value of 88% and a negative predictive value of 84% (95% CI, 69-96 and 73-91, respectively) for the former (24 h before hospitalization). Whereas an increase of 2.3/min was the best combination (72%; 95% CI, 55%-84%; and 77%; 95% CI, 65%-86%, respectively) for the latter (48 h before hospitalization), with a positive predictive value of 64% and a negative predictive value of 84% (95% CI, 48-76 and 72-91, respectively). Given that the mean baseline breathing rate determined in these patients was about 15/min, these values (4.4/min and 2.3/min) represent an increase of about 30% and 15%, respectively, from baseline, which is in keeping with our a priori working hypothesis. Finally, we observed that there was a significant relationship between the increase in mean breathing rate before hospitalization and length of hospital stay (χ2, P = .02).