Primary data, claims data, and linked data in observational research: the case of COPD in Germany

Respiratory Research

Table 2 Prevalence rates of observed comorbidities in patients in the linked dataset

	Linked dataset (N = 536): percentage of patients with selected comorbidities
	Primary data collected by GPs or pneumologists^a			Primary data collected by GPs			Primary data collected by pneumologists
	N = 536			N = 69			N = 467
Comorbidities	PD	CD	Agreement^a	PD	CD	Agreement^a	PD	CD	Agreement^a
Diabetes (Type 1 or 2)	22.2%	35.3%	87.5% (κ = 0.63)	43.5%	42.0%	98.5% (κ = 0.91)	19.1%	34.3%	84.8% (κ = 0.58)
Depression	6.7%	29.3%	77.7% (κ = 0.15)	15.9%	26.1%	89.8% (κ = 0.59)	5.4%	29.8%	75.6% (κ = 0.09)
Osteoporosis	8.0%	18.5%	90.5% (κ = 0.43)	20.3%	23.2%	97.1% (κ = 0.66)	6.2%	17.8%	88.4% (κ = 0.36)
Hypertension	45.3%	84.0%	63.6% (κ = 0.28)	62.3%	87.0%	75.3% (κ = 0.40)	42.8%	83.5%	59.3% (κ = 0.26)

CD claims dataset, GP general practitioner, PD primary dataset
Significance tests for differences in the percentage of patients diagnosed by GPs and pneumologists in the two datasets: diabetes: p = 0.030; depression: p = 0.007; osteoporosis: p = 0.129; hypertension: p = 0.005
PD values represent the percentage of patients with a particular comorbidity reported in primary data. CD values represent the percentage of patients with a particular comorbidity reported in claims data. Agreement was calculated at a patient level in each of the subgroups (overall linked dataset, primary data collected by GPs, primary data collected by pneumologists) as follows: percent agreement = 100 * ([number of patients in the subgroup with the comorbidity reported in both primary and claims data] + [number of patients in the subgroup with the comorbidity not reported in either primary or claims data]) / (total number of patients in the subgroup)
^aIncludes Cohen’s kappa coefficient for agreement between comorbidities recorded in the primary and claims datasets

ISSN: 1465-993X