A retrospective epidemiological study of Type 1 Diabetes Mellitus in Wales, UK between 2008 and 2018

Abstract

Introduction
Studies of prevalence and the demographic profile of type 1 diabetes are challenging because of the relative rarity of the condition, however, these outcomes can be determined using routine healthcare data repositories. Understanding the epidemiology of type 1 diabetes allows for targeted interventions and care of this life-affecting condition.

Objectives
To describe the prevalence, incidence and demographics of persons with type 1 diabetes diagnosed in Wales, UK, using the Secure Anonymised Information Linkage (SAIL) Databank.

Methods
Data derived from primary and secondary care throughout Wales available in the SAIL Databank were used to identify people with type 1 diabetes to determine the prevalence and incidence of type 1 diabetes over a 10 year period (2008–18) and describe the demographic and clinical characteristics of this population by age, socioeconomic deprivation and settlement type. The seasonal variation in incidence rates was also examined.

Results
The prevalence of type 1 diabetes in 2018 was 0.32% in the whole population, being greater in men compared to women (0.35% vs 0.28% respectively); highest in those aged 15-29 years (0.52%) and living in the most socioeconomically deprived areas (0.38%). The incidence of type 1 diabetes over 10 years was 14.0 cases/100,000 people/year for the whole population of Wales. It was highest in children aged 0-14 years (33.6 cases/100,000 people/year) and areas of high socioeconomic deprivation (16.8 cases/100,000 people/year) and least in those aged 45-60 years (6.5 cases/100,000 people/year) and in areas of low socioeconomic deprivation (11.63 cases/100,000 people/year). A seasonal trend in the diagnoses of type 1 diabetes was observed with higher incidence in winter months.

Conclusion
This nation-wide retrospective epidemiological study using routine data revealed that the incidence of type 1 diabetes in Wales was greatest in those aged 0-14 years with a higher incidence and prevalence in the most deprived areas. These findings illustrate the need for health-related policies targeted at high deprivation areas to include type 1 diabetes in their remit.

Introduction

Type 1 diabetes mellitus results from an immune-mediated destruction of the insulin-producing cells in the pancreas, typically presenting with symptoms related to raised blood glucose concentrations, including weight loss, excessive thirst and urination and hunger with some cases presenting acutely with the life-threatening metabolic disorder ketoacidosis which may lead to coma and death. Type 1 diabetes requires life-long treatment with exogenous insulin therapy accompanied by blood glucose monitoring. The International Diabetes Federation (IDF) reported in 2019 that over 1.1 million children and young people had type 1 diabetes worldwide, with 129,000 new cases diagnosed each year [1]. In the UK during 2010–2011, the direct cost of caring for people with type 1 diabetes to the National Health Service was £1.0 billion, which is expected to rise to £1.8 billion by 2035–2036. In addition, indirect costs were estimated to be 0.9 billion in 2010–2011, rising to £2.4 billion in 2035–2036 [2].

It is essential to have accurate data on the number of people with type 1 diabetes in order to determine both current and future health care resources required to maintain the health of this highly vulnerable population who are at significant risk of short and long-term complications. Although type 1 diabetes may present at any age it is traditionally regarded as a condition occurring predominantly in childhood, which has been the main focus for the majority of studies reporting prevalence and incidence of type 1 diabetes. These studies have demonstrated that in some countries such as the United States the incidence is rising [3], although in others such as Finland [4], Western Australia [5] and Ireland [6] the reported rate of increase has slowed or even stopped. Studies examining the prevalence, incidence and rate of complications of type 1 diabetes must reflect the fact that type 1 diabetes can present at any age [7]. In the UK the Quality and Outcomes Framework (QoF) combined with the National Paediatric Diabetes Audit [8] provided a dataset which allowed for the estimation of diabetes prevalence and incidence in the entire population [9, 10] with the National Diabetes Audit (NDA) providing an estimate of prevalence in adults [11]. Published reports from the NPDA have shown that in the UK, the incidence of type 1 diabetes in children has remained approximately constant in recent years. There are also local audits available, for example the Brecon cohort, a register of people with type 1 diabetes diagnosed under the age of 16 years in Wales [12]. The availability of large-scale research databanks has made it possible to perform epidemiological research without the need for specially gathered registry data. This study builds upon previous work to develop algorithms to identify incident cases of diabetes in the Clinical Practice Research Datalink [13], the UK IMS disease analyser [14], The Health Improvement Network [15], administrative data from Ontario, Canada [16] and Luxembourg [17] and in the Scottish Care Information-Diabetes Collaboration [18]. These methods typically use coded diagnoses and/or medication prescriptions and laboratory tests to identify cases of diabetes, however, due to the anonymised nature of these data sources and the inherent issues related to routinely collected data, robust data cleaning methods are required to ensure the accuracy of the cohort being studied [19].

Previous work on incident cases of type 1 diabetes has shown that more people are diagnosed in winter months than summer months [27]. This effect appears to be due to periods of cold weather as it persists in Southern hemisphere countries [28] but its underlying cause is not well understood [29]. We will test our cohort to see if we observe a seasonal variation in incident cases.

This retrospective epidemiological study aimed to identify all persons in Wales, diagnosed in both childhood and adulthood, with type 1 diabetes using anonymised, routinely collected healthcare data held in the Secure Anonymised Information Linkage (SAIL) Databank and to compute estimates of the true prevalence and incidence of type 1 diabetes in this population along with the demographic and clinical characteristics.

Methods

Routine electronic health record data held in the SAIL Databank [20–22] from multiple sources including both primary and secondary care were used for this study. Primary care sources include the Welsh Longitudinal General Practice (WLGP) dataset which covered approximately 80% of the population of Wales over the study period and commenced in 2000, with coverage increasing over time. The data included medications prescribed, laboratory test results and coded diagnoses made by a general practitioner. Inpatient and outpatient hospital records commenced in 1995, covered 100% of Wales and included dates of hospital admissions, diagnoses made and procedures carried out. Demographic and geographical information was drawn from the Welsh Demographic Service (WDS) dataset, which contains administrative data on all persons registered with a primary care practice in Wales. Any event, admission or service received before the index date, which was June 1^st 2018, was included in the study.

People commonly have multiple coded diagnoses of diabetes recorded in the routine data which may or may not specify a particular type of diabetes. Therefore the following criteria were used to identify people with type 1 diabetes: those with a majority of coded diagnoses of type 1 diabetes in both hospitals and general practice, were assigned a diagnosis of type 1 diabetes if insulin was prescribed within 12 months of the earliest recorded date of diagnosis of diabetes, if insulin was prescribed at least 6 months prior to any oral antidiabetic drug (OAD), if a hospital inpatient episode with a diagnosis of diabetic ketoacidosis (DKA) was recorded, or medical devices commonly used in the care of type 1 diabetes (blood glucose monitors, glucose and ketone test strips) were prescribed on at least 5 occasions within 6 months of diagnosis. People who did not have a majority of coded diagnoses of type 1 diabetes were only assigned a diagnosis of type 1 diabetes if insulin was prescribed within 6 months of the earliest recorded date of diagnosis of diabetes and, if concomitant OAD therapy was prescribed, at least six months after insulin initiation. The only permissible OAD therapies were metformin, sulphonylureas, glucagon-like peptide 1 (GLP-1) agonists or sodium-glucose transport protein 2 (SGLT-2) inhibitor agents. People presenting with type 1 diabetes before 2000 may not be identified by this algorithm since data on medication prescriptions were not generally available prior to this date, although coded diagnosis data often is available. For this reason, only new diagnoses of type 1 diabetes from 2008 onwards were used for incidence calculations. However, prevalence calculations involved all people living with type 1 diabetes diagnosed at any time.

The date of diagnosis was either the first recorded diagnosis of type 1 diabetes in any dataset, or the earliest recorded prescription of insulin in the WLGP dataset, whichever was earliest. All persons with a code for type 1 diabetes who also had some other pancreatic condition such as cystic fibrosis or pancreatic cancer prior to type 1 diabetes diagnosis were excluded from the cohort. The complete list of relevant diagnosis codes are included in Supplementary Tables 1 and 2. Deprivation was assigned using the Welsh Index of Multiple Deprivation 2011 (WIMD) score quintile. Each Lower layer Super Output Area (LSOA), small geographic areas where the minimum population is 1000 people and the mean population is 1500 people, is assigned a WIMD score quantifying the deprivation in that area [23]. Settlement type (rural, town and urban areas, based on population density [24]) was also derived from each LSOA.

The numbers of people identified with type 1 diabetes over the whole period of data coverage was used to estimate the prevalence and incidence of the condition on the index date. The prevalence was calculated by dividing the number of persons with type 1 diabetes living in Wales (including those diagnosed outside Wales) and registered at a SAIL primary care general practice by the total number of people registered at a SAIL primary care general practice, both on the index date. The incidence was calculated by dividing the number of people newly diagnosed with type 1 diabetes in a calendar year while registered at a SAIL primary care general practice and resident in Wales by the total number of people registered at a SAIL primary care general practice on the 1^st June of each year. Univariate Poisson regression was used to check for differences between population subgroups with the null hypothesis being that there was an equal probability of having type 1 diabetes in each population subgroup. For each model, only the variable under investigation (WIMD quintile, settlement type, month of diagnosis) was included as the independent variable. The dependent variable was the number of incident persons with T1DM or the number of prevalent persons with T1DM. An offset variable was included in each model to account for the differences in population sizes in the different categories. To evaluate the model fit we computed the ratio of the residual deviance to the degrees of freedom, with a value less than or greater than unity indicating under or over dispersion respectively. Lack of under or over dispersion was taken to imply standard error estimates were reasonable.

To illustrate the seasonal variation of newly-diagnosed cases of type 1 diabetes in people under 18 years of age, we took the number of newly-diagnosed cases of type 1 diabetes in each calendar month in the ten years prior to the index date and computed the mean for each month. To account for the difference in the number of days in each month we further divided each mean number by 12nmonth365.25 where n_month is the number of days in the month. Since there were two leap years during the study period, we took the length of February as 28.2 days. Confidence intervals for prevalence and incidence were computed using Jeffrey’s interval [25].

Results

Demographic characteristics of the population with type 1 diabetes living in Wales during the study period are represented in Table 1. There were 7857 people with type 1 diabetes diagnosed prior to the index date that had records in the WLGP data (see Figure 1), giving an overall prevalence of 0.32% (95% CI 0.31, 0.32). More men (n = 4366) than women (n = 3491) had type 1 diabetes, with a prevalence 0.35% (95% CI 0.34, 0.36) and 0.28% (95% CI 0.27, 0.29) respectively. 47.3% of people with type 1 diabetes were diagnosed under age 18, whereas 71.5% of the population with type 1 diabetes were diagnosed under the age of 30 years. 95% of type 1 diabetes diagnoses occurred before age 53 (Table 1).

Table 1: Demographic information.

Statistic	Value
N	7857
Men n (%)	4366 (55.6%)
Population age median (LQ, UQ)	34.5 (23.2, 50.3)
Diabetes duration median (LQ, UQ)	13.5 (6.4, 21.1)
Age at diagnosis median (LQ, UQ)	19.2 (10.6, 32.0)
Percentage diagnosed under 18	47.5%
Percentage diagnosed under 30	71.5%
Percentage diagnosed under 53	95%

Figure 1: Flow diagram detailing cohort construction.

The prevalence of type 1 diabetes was highest in those aged 15–29 years at 0.52% (95% CI 0.50, 0.55). The average incidence in the 10 years prior to the index date was 14.0 cases/100,000 people/year (95% CI 12.5, 15.5), whereas the age group with the highest incidence was those aged 0 to 14 years at 33.6 cases/100,000 people/year (95% CI 28.0, 39.6) (Table 2).

Table 2: The prevalence and incidence of type 1 diabetes by age group up to the age of 60. People with type 1 diabetes aged over 60 were included in the full cohort but not in this table as the numbers were small.

Age range	Prevalence % (95% CI)	Incidence/100,000 people/year (95% CI)
0–14	0.18% (0.17, 0.20)	33.55 (28.02, 39.56)
15–29	0.52% (0.50, 0.55)	26.87 (22.32, 31.83)
30–44	0.45% (0.43, 0.47)	13.15 (10.02, 16.70)
45–60	0.33% (0.31, 0.35)	6.49 (4.39, 8.98)

The prevalence of type 1 diabetes was 31.0% higher in the most socially deprived areas when compared to the least deprived areas. Furthermore, all regions that had greater deprivation than the least deprived areas had a higher prevalence of type 1 diabetes. There was also a difference in incidence rates only when comparing regions in the most deprived and least deprived quintiles (p = 0.040) (Table 3). The prevalence of type 1 diabetes was higher in urban areas compared to rural areas, with a 14.2% difference. There was no difference however in the observed incidence rates across different settlement types (Table 4).

Table 3: The prevalence and incidence of type 1 diabetes by Welsh Index of Multiple Deprivation (WIMD) quintile, with p-values from a Poisson regression model. The 1st WIMD quintile represents the most deprived areas, whereas areas in the 5th quintile are the least deprived.

WIMD quintile	Prevalence (%, 95% CI)	p-value	Incidence (/100,000, 95% CI)	p-value
1	0.38 (0.36, 0.39)	<0.001	16.80 (13.40, 20.59)	0.040
2	0.36 (0.34, 0.37)	<0.001	14.85 (11.55, 18.56)	0.196
3	0.34 (0.32, 0.36)	<0.001	14.24 (11.04, 17.84)	0.344
4	0.32 (0.30, 0.34)	<0.001	12.01 (8.91, 15.57)	0.753
5	0.29 (0.27, 0.30)	Reference	11.63 (8.79, 14.86)	Reference

Table 4: The prevalence and incidence of type 1 diabetes by settlement type.

Settlement type	Prevalence (% 95% CI)	p-value	Incidence (/100,000, 95% CI)	p-value
Rural	0.28 (0.26, 0.29)	<0.001	14.45 (12.66, 16.36)	0.403
Town	0.32 (0.30, 0.34)	<0.001	13.43 (9.99, 17.38)	0.589
Urban	0.31 (0.31, 0.32)	Reference	12.27 (8.80, 16.30)	Reference

There was a seasonal trend in the rate of diagnosis of type 1 diabetes in children and young people which was highest during February (p = 0.025) and lowest during the months of July (p = 0.018) and August (p = 0.005) (Figure 2).

Figure 2: The mean number of individuals under 18 years of age with newly-diagnosed (i.e. incident) type 1 diabetes per month in each year of the study period. We have normalised the values by the number of days in the month, so that points are comparable.

Discussion

We found that the prevalence of type 1 diabetes in people of all ages in Wales was 0.32%. The only prior study we are aware of that investigated type 1 diabetes prevalence in people of all ages used QoF data and found the prevalence to be 0.4% in Wales in 2014 [9, 10], which is slightly higher than our finding. This discrepancy could arise from the criteria adopted in this study which required that persons must have had a recorded prescription of insulin within 12 months of the date they were diagnosed.

The SAIL Databank was established in 2007 and contains data going back to 2000 or earlier, with historic data prior to that depending on the data source and quality of electronic data capture. People diagnosed with type 1 diabetes before 2000 will be unlikely to have their early insulin prescriptions recorded, so were not therefore included in our cohort. In 2015 Holman et al. found a prevalence of any diabetes in Wales of 0.2% when restricted to children and young people under the age of 16 years [10].

Importantly, we discovered that the prevalence of type 1 diabetes was 31.0% higher, and the incidence 42% higher in the most deprived areas compared to areas with the least deprivation. In contrast to our findings, a study in Finland found a six fold higher incidence of type 1 diabetes in children under 15 years in a population with a lower level of socioeconomic deprivation [26]. Excessive cleanliness has been hypothesised to explain the greater prevalence of autoimmune conditions such as type 1 diabetes, due to reduced exposure to infectious diseases which would otherwise enhance the immune response. Deprivation defined by the WIMD criteria which were used in this study does include quantification of housing quality, air quality, air emissions and proximity to waste and industrial sites but is not necessarily a good proxy for the cleanliness of the exposed environment [23]. However, comparing different measures of deprivation is problematic, due to the use of different indicators to quantify deprivation. The prevalence of type 1 diabetes was highest in urban areas and it is likely these two observations are related, as settlement type and deprivation quintile are highly correlated. These findings illustrate the need for programmes aimed at areas of highest deprivation to include type 1 diabetes in their remit.

There was a seasonal variation in the diagnosis of type 1 diabetes, with fewer diagnoses in July and August with a peak during February. The size of the seasonal effect observed in this study is in broad agreement with centres of comparable latitude as seen in a multicentre European study [27]. The pattern of increased winter diagnoses persists in both northern and southern hemispheres, but unfortunately there are only a few studies reporting results from the southern hemisphere [28]. Several causes for this seasonal variation have been proposed including seasonal variations in infectious disease, sun or average temperature exposure or patterns of diet and exercise but currently the mechanism is not fully understood [29].

There are some limitations to our study. Approximately 80% of people had their GP data available in the SAIL Databank which are not always complete. Also routine databanks only provide access to coded data so free text records are not available for error checking or adjudication and importantly, routine databanks only contain anonymised data preventing follow up to resolve any queries. However, data linkage is a growing field of study in medical research, and new datasets that provide a more detailed picture of people with type 1 diabetes are being added to the SAIL Databank and other repositories on a regular basis. The method used in this study to designate people with type 1 diabetes allowed for a twelve month period from initial diagnosis to receiving a first prescription for insulin in primary care to accommodate for the time between diagnosis (usually in secondary care) and medication prescriptions recorded in primary care. Information on medication prescriptions in secondary care was not available in the SAIL Databank for the purposes of this study. In addition, those people misdiagnosed as having type 2 diabetes and given oral medication prior to the correct diagnosis being established and commencement of insulin therapy will be excluded from the study cohort as not having type 1 diabetes by our chosen algorithm. Also, if the person has type 2 diabetes but is misdiagnosed as type 1 diabetes and the error is not rectified within 12 months, the algorithm would erroneously identify them as having type 1 diabetes. Given that our prevalence findings are broadly in agreement with previous work on the subject it is likely any misclassification error is small. Furthermore, misdiagnosing type 1 diabetes as type 2 diabetes is relatively unlikely, and our lower estimate of prevalence compared with the work of Holman et al. [9, 10] suggests more false negatives than false positives. People relocating into Wales and registering with a GP will be considered a new diagnosis, which would result in a small overestimate in the incidence of type 1 diabetes. Migration within Wales however is correctly accounted for and does not erroneously increase the incidence estimate.

This most recent epidemiological study of people with type 1 diabetes, based on defined diagnostic criteria, in a population of all ages living in Wales has employed the resources of the SAIL Databank, a repository of anonymised routine medical data. This has made it possible to calculate the prevalence and incidence of type 1 diabetes over the stated study period and provide a description of the population being surveyed. This study found that in Wales in 2018 the prevalence of type 1 diabetes was 0.32%, with a higher prevalence in men than women (0.35% vs. 0.28%), highest in people aged 15–34 years at 0.52% and higher in the most deprived areas at 0.38%. The incidence of type 1 diabetes for children and young people was higher in the winter months of January and February, and lowest in the months of July and August. This study provides important additional epidemiological and clinical information about the status of type 1 diabetes in Wales with respect to its prevalence and incidence in relationship to age, gender and socioeconomic status. The findings provide essential evidence to generate future health care policies better able to define and target the needs of this vulnerable group and also encourage the introduction of preventative strategies. This study also forms the basis for future epidemiological studies to monitor the impact of different interventions in clinical care and socioeconomic factors especially the devastating influence of deprivation in Wales. Lessons learnt from conducting this study will result in more comprehensive and improved future epidemiological studies which will be able to provide more accurate estimates of the prevalence and incidence of type 1 diabetes in Wales. Using similar methodology within and between countries/regions will also allow more meaningful comparisons to be made.

Supplementary Files

Supplementary appendices

Acknowledgements

The authors are grateful to John Harvey, Robert French and Rowena Bailey for helpful discussions. AA acknowledges financial support from Health Data Research UK, which is funded by the UK Medical Research Council, Engineering and Physical Sciences Research Council, Economic and Social Research Council, National Institute for Health Research (England), Chief Scientist Office of the Scottish Government Health and Social Care Directorates, Health and Social Care Research and Development Division (Welsh Government), Public Health Agency (Northern Ireland), British Heart Foundation and Wellcome.

This study makes use of anonymised data held in the SAIL Databank, which is part of the national e-health records research infrastructure for Wales. We would like to acknowledge all the data providers who make anonymised data available for research.

Abbreviations

DKA	Diabetic Ketoacidosis
GLP-1	Glucagon-like Peptide 1
IDF	International Diabetes Federation
LSOA	Lower Layer Super Output Area
OAD	Oral Antidiabetic Drug
QoF	Quality and Outcomes Framework
SAIL	Secure Anonymised Information Linkage
SGLT-2	Sodium Glucose Transport Inhibitor
WIMD	Welsh Index of Multiple Deprivation
WLGP	Welsh Longitudinal General Practice

References

1. International Diabetes Federation. 2017.

2. Hex N, Bartlett C, Wright D, Taylor M, Varley D. Estimating the current and future costs of Type 1 and Type 2 diabetes in the UK, including direct health costs and indirect societal and productivity costs. Diabet Med. 2012;29(7):855–862. 10.1111/j.1464-5491.2012.03698.x

   https://doi.org/10.1111/j.1464-5491.2012.03698.x

3. Mayer-Davis EJ, Lawrence JM, Dabelea D, Divers J, Isom S, Dolan L, et al. Incidence trends of Type 1 and Type 2 diabetes among youths, 2002–2012. N Engl J Med. 2017;376(15):1419–1429. 10.1056/nejmoa1610187

   https://doi.org/10.1056/nejmoa1610187

4. Harjutsalo V, Sund R, Knip M, Groop PH. Incidence of Type 1 diabetes in Finland. JAMA. 2013;310(4):427–428. 10.1001/jama.2013.8399

   https://doi.org/10.1001/jama.2013.8399

5. Haynes A, Bulsara MK, Jones TW, Davis EA. Incidence of childhood onset Type 1 diabetes in Western Australia from 1985 to 2016: Evidence for a plateau. Pediatr Diabetes. 2018;19(4):690–692. 10.1111/pedi.12636

   https://doi.org/10.1111/pedi.12636

6. Roche EF, McKenna AM, Ryder KJ, Brennan AA, O’Regan M, Hoey HM. Is the incidence of Type 1 diabetes in children and adolescents stabilising? The first 6 years of a National Register. Eur J of Pediatr. 2016;175(12):1913–1919. 10.1007/s00431-016-2787-6

   https://doi.org/10.1007/s00431-016-2787-6

7. Leslie RD. Predicting adult-onset autoimmune diabetes: clarity from complexity. Diabetes. 2010;59(2):330–331. 10.2337/db09-1620

   https://doi.org/10.2337/db09-1620

8. Royal College of Paediatrics and Child Health, London, UK. National Paediatric Diabetes Audit Report 2017–18. Royal College of Paediatrics and Child Health. 2018.

9. Holman N, Young B, Gadsby R. What is the current prevalence of diagnosed and yet to be diagnosed diabetes in the UK. Diabet Med. 2014;31(5):510–511. 10.1111/dme.12397

   https://doi.org/10.1111/dme.12397

10. Holman N, Young B, Gadsby R. Current prevalence of Type 1 and Type 2 diabetes in adults and children in the UK. Diabet Med. 2015;32(9):1119–1120. 10.1111/dme.12791

    https://doi.org/10.1111/dme.12791

11. NHS Digital. National Diabetes Audit, 2017–18 Report 1: Care Processes and Treatment Targets. NHS Digital. 2018.

12. Sayers A, Thayer D, Harvey JN, Luzio S, Atkinson MD, French R, et al. Evidence for a persistent, major excess in all cause admissions to hospital in children with Type-1 diabetes: results from a large Welsh national matched community cohort study. BMJ Open. 2015;5(4):e005644. 10.1136/bmjopen-2014-005644

    https://doi.org/10.1136/bmjopen-2014-005644

13. Imkampe AK, Gulliford M. Trends in Type 1 diabetes incidence in the UK in 0-to 14-year-olds and in 15-to 34-year-olds, 1991–2008. Diabet Med. 2011;28(7):811–814. 10.1111/j.1464-5491.2011.03288.x

    https://doi.org/10.1111/j.1464-5491.2011.03288.x

14. Hsia Y, Neubert AC, Rani F, Viner RM, Hindmarsh PC, Wong IC. An increase in the prevalence of Type 1 and 2 diabetes in children and adolescents: results from prescription data from a UK general practice database. Br JClin Pharmacol. 2009;67(2):242–249. 10.1111/j.1365-2125.2008.03347.x

    https://doi.org/10.1111/j.1365-2125.2008.03347.x

15. Sharma M, Petersen I, Nazareth I, Coton SJ. An algorithm for identification and classification of individuals with Type 1 and Type 2 diabetes mellitus in a large primary care database. Clin Epidemiol. 2016;8:373. 10.2147/clep.s113415

    https://doi.org/10.2147/clep.s113415

16. Lipscombe LL, Hwee J, Webster L, Shah BR, Booth GL, Tu K. Identifying diabetes cases from administrative data: a population-based validation study. BMC Health Serv Res. 2018;18(1):316. 10.1186/s12913-018-3148-0

    https://doi.org/10.1186/s12913-018-3148-0

17. Renard LM, Bocquet V, Vidal-Trecan G, Lair ML, Couffignal S, Blum-Boisgard C. An algorithm to identify patients with treated Type 2 diabetes using medico-administrative data. BMC Med InformDecis Mak. 2011;11(1):23. 10.1186/1472-6947-11-23

    https://doi.org/10.1186/1472-6947-11-23

18. Livingstone SJ, Looker HC, Hothersall EJ, Wild SH, Lindsay RS, Chalmers J, et al. Risk of cardiovascular disease and total mortality in adults with Type 1 diabetes: Scottish registry linkage study. PLoS Med. 2012;9(10):e1001321. 10.1371/journal.pmed.1001321

    https://doi.org/10.1371/journal.pmed.1001321

19. De Lusignan S, Khunti K, Belsey J, Hattersley A, Van Vlymen J, Gallagher H, et al. A method of identifying and correcting miscoding, misclassification and misdiagnosis in diabetes: a pilot and validation study of routinely collected data. Diabet Med. 2010;27(2):203–209. 10.1111/j.1464-5491.2009.02917.x

    https://doi.org/10.1111/j.1464-5491.2009.02917.x

20. Ford DV, Jones KH, Verplancke JP, Lyons RA, John G, Brown G, et al. The SAIL Databank: building a national architecture for e-health research and evaluation. BMC Health Serv Res. 2009;9(1):157. 10.1186/1472-6963-9-157

    https://doi.org/10.1186/1472-6963-9-157

21. Lyons RA, Jones KH, John G, Brooks CJ, Verplancke JP, Ford DV, et al. The SAIL databank: linking multiple health and social care datasets. BMC Med InformDecis Mak. 2009;9(1):3. 10.1186/1472-6947-9-3

    https://doi.org/10.1186/1472-6947-9-3

22. Jones KH, Ford DV, Jones C, Dsilva R, Thompson S, Brooks CJ, et al. A case study of the Secure Anonymous Information Linkage (SAIL) Gateway: a privacy-protecting remote access system for health-related research and evaluation. J Biomed Inform. 2014;50:196–204. 10.1016/j.jbi.2014.01.003

    https://doi.org/10.1016/j.jbi.2014.01.003

23. Welsh Government. Welsh Index of Multiple Deprivation 2011, Technical Report;. Accessed: 2019-06-05. https://gov.wales/sites/default/files/statistics-and-research/2019-04/wimd-2011-technical-report.pdf.

24. Bibby P, Shepard J. Developing a New Classification of Urban and Rural Areas for Policy Purposes – the Methodology;. Accessed: 2019-07-23. https://www.gov.uk/government/statistics/2001-rural-urban-definition-la-classification-and-other-geographies.

25. Brown LD, Cai TT, DasGupta A. Interval estimation for a binomial proportion. Stat Sci. 2001:101–117.

26. Kondrashova A, Reunanen A, Romanov A, Karvonen A, Viskari H, Vesikari T, et al. A sixfold gradient in the incidence of Type 1 diabetes at the eastern border of Finland. Ann Med. 2005;37(1):67–72. 10.1080/07853890410018952

    https://doi.org/10.1080/07853890410018952

27. Moltchanova E, Schreier N, Lammi N, Karvonen M. Seasonal variation of diagnosis of Type 1 diabetes mellitus in children worldwide. Diabet Med. 2009;26(7):673–678. 10.1111/j.1464-5491.2009.02743.x

    https://doi.org/10.1111/j.1464-5491.2009.02743.x

28. Willis JA, Scott RS, Darlow BA, Lewy H, Ashkenazi I, Laron Z. Seasonality of birth and onset of clinical disease in children and adolescents (0-19 years) with Type 1 diabetes mellitus in Canterbury, New Zealand. JPediatr EndocrinolMetab. 2002;15(5):645–648. 10.1515/jpem.2002.15.5.645

    https://doi.org/10.1515/jpem.2002.15.5.645

29. Patterson C, Gyürüs E, Rosenbauer J, Cinek O, Neu A, Schober E, et al. Seasonal variation in month of diagnosis in children with Type 1 diabetes registered in 23 European centers during 1989–2008: little short-term influence of sunshine hours or average temperature. Pediatr Diabetes. 2015;16(8):573–580. 10.1111/pedi.12227

    https://doi.org/10.1111/pedi.12227