Population with diabetes
The population with diabetes was defined as those having any diabetes-related entries in any of the following registers: the Hospital Discharge Register, the Cause of Death Register, the Finnish Kidney Register, and drug registers of the Social Insurance Institution between 1964 and 2011 and alive in 1 January 1994. For these persons, information regarding the use of hospital services and causes of death was individually linked using the personal identification code unique to each individual. Diabetes type was determined by register entries concerning drug use: persons having entries on continuous use of insulin but no signs of drug use increasing insulin production of the pancreas were defined as having type 1 diabetes. Population at risk was defined as persons with diabetes aged 30 years or older in 1994–2011. We formed annual cohorts of all individuals with diabetes and alive in the beginning of the year and those with incident diabetes during that year. Since we examined complications of diabetes, the diabetes diagnosis needed to precede the complication entry. Persons with gestational diabetes only were excluded from the analyses.
Diabetes-related complications were defined as right for elevated health insurance reimbursement for drug costs due to each of these complications granted by the Social Insurance Institution, or an entry in the Hospital Discharge Register, or Cause of Death Register or the Finnish Kidney Register. We examined the following complications: (1) stroke (ICD-9 and ICD-10 codes: 430*, 431*, 4330A, 4331A, 4339A, 4340A, 4341A, 4349A, I60*, I61*, I63*), (2) AMI (ICD-9 and ICD-10 codes: 410*, I21*–I22*), (3) lower extremity amputation (LEA) (Finnish Hospital League codes: 9571–9575, NOMESCO codes: NFQ10, NFQ20, NGQ10, NGQ20, NHQ10, NHQ20, NHQ30, or NHQ40), and (4) ESRD (entry in the Finnish Kidney Register concerning the onset of renal replacement therapy or cause of death codes 5855, 5856, N185, or N186). We examined the first register entry of each complication only as they are chronic conditions. By only counting the first register entry, we wanted to ensure not counting the same complication several times as each of the complications can result to entries in different registers and several entries in the hospital discharge register.
Statistical analyses
We examined trends in the prevalence of these complications by calculating age-standardized rates per 1000 using the 2011 diabetes population as a standard. Multivariate analyses were based on repeated-measures Poisson regression models controlling for age and diabetes duration. Separate models were calculated for men and women. Trends in the clustering of complications were analyzed by comparing prevalence in 1994–1999 to two later periods (2000–2005 and 2006–2011). We further examined interactions between age and period as well as duration of diabetes and period to study whether the development was similar in different subgroups. In additional analyses, we conducted Cox regression models using counting process data to study the effect of the four complications on survival until death or censoring at the end of year 2011. Models were fitted for people with type 1 and type 2 diabetes separately and adjusted for gender, age, study period, and duration of diabetes. Ethical approval for the study was received from the Research Ethics Committee of the National Institute for Health and Welfare.