Table 2 Nine year incidence of FEV₁ <80% predicted and of chronic cough or phlegm, medication use because of breathing problems in the past 12 months at the ECRHS II and hospital services utilisation because of breathing problems between the two surveys, according to the presence of airflow obstruction, as defined by the two criteria, among the subjects identified at the ECRHS I and traced at the ECRHS II

		Subjects without AO	Misidentified subjects	Subjects with AO	p Value (misidentified subjects vs those without AO)	p Value (misidentified subjects vs those with AO)
	Fixed cut-off (70%)	AO−	AO−	AO+
	LLN	AO−	AO+	AO+
FEV₁ <80% pred*	No of subjects at risk	5576	295	174	–	–
	Crude incidence rate (1000/y) (95%CI)	1.89 (1.55 to 2.32)	5.66 (3.41 to 9.40)	15.87 (10.72 to 23.49)	–	–
	Incidence rate ratio (95%CI)	0.34 (0.19 to 0.58)	1.00	3.17 (1.66 to 6.05)	<0.001	<0.001
Chronic cough or phlegm†	No of subjects at risk	4865	257	197	–	–
	Crude incidence rate (1000/y) (95%CI)	8.69 (7.85 to 9.61)	14.31 (10.17 to 20.13)	10.69 (6.82 to 16.76)	–	–
	Incidence rate ratio (95%CI)	0.61 (0.42 to 0.87)	1.00	0.78 (0.44 to 1.38)	0.006	0.400
Medication use‡	No of subjects	5634	316	267	–	–
	Crude proportion (%) (95%CI)	13.2 (12.3 to 14.1)	24.1 (19.4 to 29.2)	51.3 (45.1 to 57.4)	–	–
	Risk ratio (95%CI)	0.55 (0.45 to 0.67)	1.00	2.08 (1.71 to 2.54)	<0.001	<0.001
Hospital services utilisation§	No of subjects	5641	317	267	–	–
	Crude rate (1000/y) (95%CI)	5.63 (5.00 to 6.34)	13.43 (9.68 to 18.61)	17.53 (12.81 to 24.00)	–	–
	Rate ratio (95%CI)	0.45 (0.32 to 0.64)	1.00	1.39 (0.88 to 2.19)	<0.001	0.158

*Subjects at risk were those with an FEV₁ ⩾80% predicted at baseline; the incidence rate ratios were obtained by a two level Poisson regression model.
†Subjects at risk were those without chronic cough or phlegm at baseline; 84 subjects at risk with missing information on the outcome were not considered in the analysis; the incidence rate ratios were obtained by a two level Poisson regression model.
‡32 subjects with missing information on the outcome were not considered in the analysis; the risk ratios were obtained by a two level Poisson regression model with a robust error variance and no offset.23
§24 subjects with missing information on the outcome were not considered in the analysis; the crude rates of occurrence of the first emergency department visit/hospital admission between the two surveys were calculated setting the person-years for the subjects who reported at least one hospital contact equal to half the length of the follow-up; the rate ratios were obtained by a two level complementary log–log survival model.24
AO, airflow obstruction; FEV₁, forced expiratory volume in 1 s; LLN, lower limit of normal.