All individuals in the County of Funen (approx. 0.5 million inhabitants), Denmark, were followed for occurrences of amputation, diagnosis of diabetes and death during the period 1996–2011, with the aim of describing amputation rates and comparing these between individuals with and without diabetes.
We identified all individuals with a record of amputation from the administrative system at Odense University Hospital. Amputation data were predefined by surgical codes categorising amputation level. Each contact was recorded with one or more diagnoses and procedure codes according to the Nordic Classification of Surgical Procedures  and the ICD-10 system (www.who.int/classifications/icd/en/); “*” was used as a wildcard including all relevant subcodes. Amputations were categorised as follows:
below-ankle amputation (BAAs): ankle, foot and toe, KNHQ0*, KNHQ1* or KNHQ9*
from-ankle-to-knee amputations (BKAs): knee and lower leg, KNGQ0*, KNGQ1* or KNGQ9*
above-knee amputations (AKAs): hip and thigh, KNFQ0*, KNFQ1* or KNFQ9*
We defined amputation status as the highest located amputation performed at any time. Thus, a BAA on the right foot in a person already having had a BKA on the left leg was not considered a BAA event. A person could contribute to all three amputation categories, but only in increasing sequence over time (BAA ➔ BKA ➔ AKA), and only by the first amputation in each category.
We noted a group of very young individuals who were registered as amputees. After a thorough investigation, most of these amputations were shown to have been performed because of congenital malformations (e.g. extra toes), and all individuals below the age of 15 were therefore excluded from the follow-up analyses.
We collected the following variables from the Danish National Diabetes Register (NDR) : ID, sex, date of birth, date of inclusion in the NDR (corresponding to the date of diagnosis of diabetes) and date of death. The NDR comprises persons meeting one of the following criteria, the inclusion date being defined as the date on which the first criterion was met.
A diagnosis of diabetes in the National Patient Register defined as ICD-10 (www.who.int/classifications/icd/en/) DE10–14, DH36.0 or DO24 (excluding DO24.4), and ICD-8 (prior to 1994).
Registration of chiropody in a diabetic patient in the National Health Services Register (NHSR).
Five blood glucose measurements during a 1 year period in the NHSR.
Two blood glucose measurements per year in five consecutive years in the NHSR.
The second purchase of oral glucose-lowering drugs in the prescription register, the Register of Medicines Products Statistics (RMPS), with the exception of women aged 20–39 purchasing metformin only, because metformin is also used as a medication for polycystic ovarian syndrome.
The second purchase of prescribed insulin recorded in the RMPS .
The database is estimated to have a sensitivity of ≥95% and a positive predictive value (PPV) of 80% .
Using Statistics Denmark’s databank, we collected data on population size and number of deaths for the County of Funen for the study period in 1 year age and calendar time classes. Place of residence and date of each change of residence were collected from the Civil Registration System (CRS)  ID in order to account for migration in and out of the county of Funen.
Combination of cohort follow-up and population data
Using this set-up, information from the database includes all individuals residing in the county of Funen who either had a diagnosis of diabetes and/or underwent amputation during the period 1996–2011. We subdivided follow-up among these by diabetes status (no/yes), amputation status (none/BAA/BKA/AKA), sex, age and period in 1 year intervals, giving a total of 614,318 records. Subtracting the follow-up among individuals with diabetes and/or amputation from that of the total population provided the follow-up from individuals without diabetes and amputation, classified by sex, age and period of follow-up. The total follow-up time classified by diabetes status, amputation status, sex, age and calendar time, as well as the number of diagnoses of diabetes, amputations (BAA, BKA, AKA) and deaths, was also provided from these registrations. In this study, we focus on the transitions to the different amputation states separately among individuals with and without diabetes. The states and transitions between all states are shown in Fig. 1.
Only follow-up from age 15 to 99 years was included in the analysis. We computed directly age-standardised amputation rates by type, year and diabetes status, standardised to the population of individuals without diabetes as of 2011. The dataset used in the analysis consisted of time-split records from individuals with diabetes or amputation (612,800 records) combined with records of person-years and number of amputation events among non-diabetic individuals with no prior amputation (2,685 records), classified by sex, age and date of follow-up, and duration of diabetes in 1 year steps. We analysed the incidence of the first amputation at each level as well as the incidence of further amputations, subdivided by level of amputation, as illustrated in Fig. 1.
Amputation rates were analysed with proportional hazards Poisson models describing amputation rates as smooth functions of age, calendar time and duration of diabetes (coded as the midpoint of the intervals) using restricted cubic splines. Models also included effect of diabetes and amputation status as well as interaction between sex and diabetes. Estimates of rates and HRs were reported, with 95% CIs.
All analyses and graphs were made with R version 3.0.2 , using the Epi-package for definition and handling of multistate follow-up [20, 21]. A complete account of all data manipulation, statistical analysis and graphical reporting is published as electronic supplementary material (ESM) and is also available at http://bendixcarstensen.com/SDC/Ampu/FynAmp.pdf
The study is based entirely on register data, and therefore does not require approval from ethics committees according to Danish law. The study was registered with and approved by the Danish Data Protection Agency (Region of Southern Denmark, ID: 2008-58-0035).