Are the Observed Differences Real?

Background and Purpose: We sought to find an explanation for the geographical variation in the incidence of nonfatal stroke detected in the FINMONICA stroke register during the period 1983-1985. Methods: Two separate investigations were made. In the first, a sample of approximately 100 nonfatal events drawn from each of the three monitoring areas (North Karelia, Kuopio, and Turku/Loimaa) participating in FINMONICA was recoded by an independent coder. In the second, 29 nonfatal events from Kuopio and 29 from North Karelia were recoded by the neurologist of the other area. Kappa coefficients (K) were calculated to measure the proportion of agreement beyond chance among the different coders. Results: A good level of agreement with the independent coder was found for each of the three areas; it was best for the cases from Turku/Loimaa (ic=0.896), followed by Kuopio (Kc=0.792) and North Karelia (pc=0.616). In the second part of the investigation, agreement was higher for the stroke cases originating from Kuopio (X=0.861) than for those from North Karelia (pc=0.563). In the latter there was discrepancy in the classification of the cases originally classified as no stroke. Conclusions: The results from the first part of the investigation suggest that the lower incidence of nonfatal stroke in Turku/Loimaa was real, confirming findings in previous studies. Differences in data entry, rather than in the interpretation of signs and symptoms of stroke, were the main cause of disagreement in the second part of the investigation, where the disagreement primarily concerned the cases classified as no stroke in North Karelia. The results also indicate that the different proportions of patients submitted to computerized brain tomography conceivably account for a good part of the difference in the incidence rates of nonfatal stroke between Kuopio and North Karelia. (Stroke