Introduction
Life expectancy has rapidly increased among general populations and among the population with type 1 diabetes [
1]. Due to improvements in the prevention and treatment of diabetes-related complications, more individuals with type 1 diabetes are living longer [
2,
3]. However, the incidence of most diabetes-related complications and non-communicable diseases increases with age [
4,
5]. The fact that more individuals with type 1 diabetes are reaching older ages is therefore accompanied by several challenges associated with multimorbidity (the presence of two or more long-standing chronic conditions) [
6]. Two of these challenges are the increasing prevalence of polypharmacy, often defined numerically as being exposed to five or more drugs with potential for drug–drug interactions [
7], and the increasing number of individuals being exposed to drugs that are associated with an increased risk of medication errors and adverse drug reactions [
8].
For general populations, the prevalence and risks of polypharmacy and high-risk prescribing are relatively well explored [
9,
10]. Polypharmacy and the prescription of high-risk drugs were shown to be strongly associated with an increased risk of adverse health outcomes [
11‐
13]. However, identifying the extent to which these risks reflect direct drug effects, drug–drug interactions or the health conditions for which drugs are prescribed is difficult in observational studies [
14]. Despite the complexity of disentangling direct drug effects from confounding by indication, the potential harm from polypharmacy and high-risk prescribing has led to clinical recommendations to minimise the potential risks in Scotland and the UK [
15,
16].
For particular disease groups and subpopulations, including individuals with type 2 diabetes, the prevalence of polypharmacy and its association with adverse health outcomes is increasingly recognised [
17‐
20]. In contrast, very little is known for the population with type 1 diabetes. This is surprising, as the management of type 1 diabetes and its complications often require early pharmacological interventions, implying that individuals are often exposed to complex medication regimes for a long period of time [
21,
22].
Using data for the entire Scottish population with type 1 diabetes, we mapped the number of prescribed drugs over age, by sex and area-based socioeconomic deprivation on 1 January 2017. We then studied the association of each additional drug and the prescription of particular high-risk drugs at baseline with the first hospital admission for falls, diabetic ketoacidosis (DKA) and hypoglycaemia, or death within the subsequent 12 months. All studied outcomes represent important endpoints: hospital admissions for DKA and hypoglycaemia are among the major acute complications of type 1 diabetes; falls and death are among the most reported complications of polypharmacy among general populations. In line with findings for general populations and individuals with type 2 diabetes, we expected the number of prescribed drugs among individuals with type 1 diabetes to increase with age, to be higher among the female sex than the male sex, and to be higher among individuals from more deprived areas. Furthermore, we hypothesised that each additional drug and the prescription of high-risk drug classes would be associated with an increased hazard for hospitalisation for falls, DKA and hypoglycaemia as well as death. These findings will provide important evidence to improve appropriate prescribing among individuals with type 1 diabetes.
Methods
Study population
Using the SCI-Diabetes database, we identified all individuals resident in Scotland, irrespective of their age, who were alive and had a diagnosis of type 1 diabetes at baseline, defined as 1 January 2017 (
N = 28,245). For each individual, we counted the total number of prescribed drugs at baseline. Identical chemical substances, identified using the seventh digit of the Anatomical Therapeutic Chemical Classification System (ATC) of the WHO, were counted as one drug [
24]. Individuals with type 1 diabetes require insulin, which is often prescribed together with drugs to manage hypoglycaemia. Insulins, glycogenolytic hormones and carbohydrates were not considered when counting the number of prescribed drugs. In addition, we did not count devices such as insulin pumps, flash monitors or needles.
We also examined whether individuals were prescribed high-risk drugs. We decided to focus on those second- and third-level ATC classes that were consistently reported to be strongly associated with an increased risk of serious medication errors and adverse drug reactions, leading to hospital admissions, disabilities or death [
8,
25,
26]. While the use of such drugs can of course be clinically appropriate, it is recommended that these drug classes are critically reviewed periodically by the handling practitioner according to the recent Scottish Polypharmacy and Appropriate Prescribing Guidelines [
15]. From a total number of 4747 unique ATC codes, 769 (16.20%) unique ATC codes were captured as high-risk drugs.
We used the Scottish Index of Multiple Deprivation (SIMD) 2016 as an area-based measure of socioeconomic deprivation. The SIMD is an area-level index that captures social deprivation across multiple aspects of life, including unemployment, income, education and crime rates at an individual’s place of residence [
27].
Measures of baseline characteristics for the study population were identified within a 2 year window prior to baseline. If multiple measurements were available, the measurement closest to baseline was used. These measures were mainly recorded in primary care and included diabetes duration, HbA1c, systolic BP, diastolic BP, HDL-cholesterol, LDL-cholesterol, total cholesterol, BMI, diabetic foot risk score, retinopathy grading, eGFR, smoking status, and whether individuals used continuous subcutaneous insulin infusion (CSII). The diabetic foot risk score reflects the maximum score of either the left or the right foot. The retinopathy grading is based on a score combining the maximum grading of each eye. Measures of eGFR were adjusted for individuals receiving renal replacement therapy and categorised as <30 (ml min−1 [1.73 m]−2).
In addition, we obtained information on whether individuals were previously admitted to hospital for CVD, hypoglycaemia and DKA using all available information on hospital admissions. We obtained information on the number of hospital admissions within the 2 year period prior to baseline, not considering admissions for DKA and hypoglycaemia.
The daily dose of insulin at baseline was conceptualised as the mean daily dose of insulin per day within a 360 day window ranging from 180 days before baseline to 180 days after baseline. For each individual, we combined information on quantity, pack size and strength of all insulin prescriptions to estimate the total amount of issued insulin within this 360 day window. We then assumed that 20% of all prescribed insulin is not taken due to damage, loss, a passed expiry date or deviance from the established treatment regimen. While no study has quantified this ‘waste factor’ explicitly for the population with type 1 diabetes in Scotland, we followed results discussed in the literature [
28]. We then divided the corrected, total amount of issued insulin by the number of days individuals were observable within this 360 day window. A small number of unrealistically high and low daily doses of insulin were identified using the 0.5% (left) and 1% (right) tails of a fitted log-normal distribution, the shape of which described the original data best.
The insulin dose for all identified outliers and all missing sociodemographic and health information at baseline were imputed using multiple imputation methods, based on all covariates presented in this study. An overview on the fraction of all imputed missing data at baseline is provided in ESM Table
1. Imputations were carried out using the R-package Amelia (R version 3.6.0, Amelia version 1.7.6; downloaded via
https://cran.r-project.org/).
Acknowledgements
We thank the Scottish Diabetes Research Network (SDRN) Epidemiology Group: J. Chalmers (Diabetes Centre, Victoria Hospital, UK), C. Fischbacher (Information Services Division, NHS National Services Scotland, Edinburgh, UK), B. Kennon (Queen Elizabeth University Hospital, Glasgow, UK), G. Leese (Ninewells, Hospital, Dundee, UK) R. Lindsay (British Heart Foundation Glasgow Cardiovascular Research Centre, University of Glasgow, Glasgow, UK), J. McKnight (Western General Hospital, NHS, Edinburgh, UK), J. Petrie (Institute of Cardiovascular & Medical Sciences, University of Glasgow, Glasgow, UK) R. McCrimmon (Division of Molecular and Clinical Medicine, University of Dundee, Dundee, UK), S. Philip (Grampian Diabetes Research Unit, Diabetes Centre, Aberdeen Royal Infirmary, Aberdeen, UK), D. McAllister (Institute of Health & Wellbeing, University of Glasgow, Glasgow, UK), E. Pearson (Population Health and Genomics, School of Medicine, University of Dundee, Dundee, UK) and S. Wild (Usher Institute, University of Edinburgh, Edinburgh, UK).
Some of the data were presented as an abstract at the 56th EASD Annual Meeting in 2020.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.