Reproducibility of diabetes quality of care indicators as reported by patients and physicians.

INTRODUCTION: Self-report of diabetes care has moderate validity and is prone to under- and over-reporting. We assessed reproducibility of a range of processes and outcomes of diabetes care as reported by patients and physicians. METHODS: In a Swiss community-based survey, patients with diabetes and physicians independently reported past 12 months processes of care (HbA1c, lipids, microalbuminuria, blood pressure, weight, foot and eye examinations) and last measured values of HbA1c, height, weight and blood pressure. For dichotomous variables, we assessed reliability by Cohen's kappa and agreement by uniform kappa. For continuous measures, we used Lin's concordance correlation coefficient and limits of agreement, respectively. RESULTS: Mean age of the 210 patients was 65 years; 40% were women, and 51% had diabetes for >10 years. Agreement was good for recommended processes of care such as blood pressure (uniform kappa = 0.94), HbA1c (0.93), weight (0.88) and lipid (0.78), but lower for microalbuminuria, foot and eye examinations (all <0.50). Cohen's kappa values were all low (<0.25). Comparisons of reported continuous variables showed large limits of agreement for height (±6 cm) and weight (8-10 kg) despite high concordance correlation coefficients (0.93 and 0.97). Concordance correlation coefficients were smaller for HbA1c (0.72) and blood pressure (0.5-0.6), with large limits of agreement (±2% and ±25 mmHg). CONCLUSION: While agreement of routine processes of care was good, agreement was less satisfactory for microalbuminuria, foot and eye examinations. Reports of continuous outcomes yielded good reliability but too wide limits of agreement. Quality of care evaluation relying on self-report only should be made cautiously.