A cross-national study of hip fracture incidence was carried out in five geographic areas--Beijing, China; Budapest, Hungary; Hong Kong; Porto Alegre, Brazil; and Reykjavik, Iceland--during the years 1990-1992. Cases of hip fracture among women and men of age 20 years and older were identified using hospital discharge data in conjunction with medical records, operating room logs, and radiology logs. Estimated incidence rates varied widely, with Beijing reporting the lowest rates (age-adjusted rate per 100,000 population for men 20 years and older = 45.4; women = 39.6) and Reykjavik the highest rates (man = 141.3; women = 274.1). Rates were higher for women than for men in every area except Beijing. In every area except Budapest, review of the operating room or radiology logs identified additional cases that were not reported in the discharge list, increasing the estimated number of hip fractures by 11% to 62%, depending on the area. Review of medical records identified miscoding of hip fractures (ICD 9820) as 'shaft of femur and other femur fractures' (ICD 9821) in the discharge lists of every area except Budapest, increasing the estimated number of hip fractures by 1% to 30%. The final estimates of hip fracture incidence taking into account all investigated sources of undercount and overcount ranged from 15% lower to 89% higher than an estimate based on the discharge diagnoses alone. Although these results indicate substantial limitations in relying on hospital discharge data alone to estimate hip fracture incidence rates, the extent of errors found in the discharge lists is smaller than the large international variation found here and previously reported in incidence rates. The findings support the conclusion that the differences reported among countries mainly reflect genuine variation in the hip fracture incidence rates.

A quantitative fluid-phase radioassay for autoantibodies reacting with insulin (competitive insulin autoantibody assay, CIAA) was developed. The assay's features include 1) use of a physiologic amount of 125I-labeled insulin, 2) parallel incubations with supraphysiologic cold insulin (competitive), and 3) an incubation time of 7 days and a single-step multiple-wash polyethylene glycol separation. Mean +/- SE CIAA levels in 50 controls were 8 +/- 1.4 nU/ml (range -16-33.3). In 36 cytoplasmic islet cell antibody (ICA)-positive nondiabetic first-degree relatives of type I (insulin-dependent) patients less than 30 yr of age, CIAA levels exceeded the normal range in 20 (55.6%) of 36 (mean 86.8 +/- 17.1 nU/ml). In 26 ICA-positive relatives greater than 30 yr of age, only 5 (19.2%) of 26 exceeded the normal range (mean 26.1 +/- 9.4 nU/ml); P less than .001 compared with younger ICA-positive relatives). Six ICA-negative HLA-identical siblings of type I diabetic patients had normal CIAA levels (mean 3.6 +/- 5.8 nU/ml), and only 2 of 13 ICA-negative identical twins discordant for diabetes (mean 15.4 +/- 6.6 nU/ml) exceeded the normal range. Nine (50%) of 18 ICA-positive schoolchildren exceeded the normal range (mean 105.3 +/- 36.7 nU/ml). Genetically susceptible subjects negative for CIAA (with only 3 exceptions) remained negative for CIAA on multiple determinations (3 conversions observed), and CIAA levels of positive subjects were relatively stable. Linear regression of the first CIAA level versus last (interval between sampling 1 mo to 10 yr) in genetically susceptible individuals showed a highly significant correlation (r = .95, P less than .001).(ABSTRACT TRUNCATED AT 250 WORDS)