City & Hackney Newham Tower Hamlets

(1)

With support from: Global Health, Policy, and Innovation Unit

City & Hackney

Newham

Tower Hamlets

Diabetes Risk

2011

Clinical Effectiveness Group

(2)

1 | P a g e

Acknowledgements

Project team & authors Project manager Dr. Douglas Noble Data analyst Ms. Rohini Mathur Geospatial analyst Dr. Dianna Smith

Director of Global Health, Policy, and Innovation Unit

Prof. Trisha Greenhalgh

Clinical Effectiveness Group

Dr. John Robson

Contribution to systematic review of diabetes risk scores Co-authors of BMJ paper Tom Dent Catherine Meads Other contributors Helen Elwell Samuel Rigby Sietse Wieringa Kaveh Memarzadeh Nicholas Swetenham Wendy Hu John Furler Other acknowledgements

Julia Hippisley-Cox – QDScore Karen Thomson – NIGB

Muriel McAughtrie – QMUL administration Sean O‟Hanlon and David Stables – EMIS East London GPs and PCTs

Funding

This study was funded by grants from Tower Hamlets, Newham, and City and Hackney primary care trusts, a National Institute of Health Research Senior Investigator award for Professor Greenhalgh, internal funding for staff time from Barts and the London School of Medicine and Dentistry and a Medical Research Council postdoctoral fellowship for Dr Smith.

(3)

2 | P a g e

Executive Summary

This report describes an analysis of the general practice records of a cohort of 519, 288 adults aged 25-79 in Inner North East London (Newham, Tower Hamlets and City and Hackney) to calculate 10-year risk of developing type 2 diabetes. This risk was high across all three boroughs. Overall, 1 in 10 people (51, 061) were at high risk (≥20%) of developing type 2 diabetes within ten years. In Newham 1 in 8 people (22, 513) were at high risk (≥20%), compared to 1 in 11 people (15, 304) in Tower Hamlets and 1 in 13 people (13, 244) in City and Hackney.

10% of males were at high risk (≥20%) compared to 9% of females. As expected,

the risk of developing type 2 diabetes rose with age, but a proportion of young adults were at high risk (≥20%). 2% of 25-39 year olds were at high risk (≥20%) compared to 20% of 40-79 year olds.

South Asian ethnicity increased risk of type 2 diabetes substantially. More than twice as many South Asians (16%) were at high risk (≥20%) compared to the White (7%) population. However, high risk (≥20%) was not confined to South Asians. In

Newham 10% of the White population was at high risk (≥20%) compared to 6% in

Tower Hamlets and 7% in City & Hackney.

Socio-economic deprivation was associated with increased risk and ethnicity increased this association. Those in the lowest band of Townsend Score have the lowest number of people (8%) who are high risk (≥20%) compared to the highest band of Townsend Score (12%). Within the South Asian population at high risk (22, 126) the proportion of people at high risk (≥20%) increased and was higher than the overall population in every Townsend Score band.

Geospatial linkage in Tower Hamlets revealed „hot spots‟ of high risk (≥20%)

corresponding to areas of greater socio-economic deprivation and ethnic diversity. At small area level hypothesized determinants of health such as fast food outlets or lack of green spaces can be displayed to aid local health needs assessment.

Obesity and cardiovascular co-morbidity substantially increased risk of developing type 2 diabetes. For example, 76, 162 people in the cohort had a BMI greater than 30, and 43% of these were high risk (≥20%) for developing type 2 diabetes. 57% of people with cardiovascular disease, 48% with hypertension, 41% with chronic kidney disease, and 18% with gestational diabetes were at high risk (≥20%) for developing type 2 diabetes, compared to 10% of the cohort overall. For QRisk ≥20, 48% (9839) were at high risk (≥20%) for type 2 diabetes.

East London faces vastly increasing rates of diabetes by 2020 with the number of diagnosed cases expected to rise from 40, 000 to 80, 000. Risk of type 2 diabetes is closely associated with development of established disease and could reasonably be expected to rise by the same factor resulting in 1 in 5 adults at high risk of type 2 diabetes by 2020. Urgent public health action needs to be taken not just to improve early diagnosis and management of diabetes but to reduce risk across the whole population. Further local action is required to establish the combination of policy changes, lifestyle interventions and pharmacotherapy that have the potential to reduce risk with a view to decreasing incidence of established disease.

(4)

3 | P a g e

Introduction

Diabetes epidemiology

Global context

Diabetes is becoming increasingly common throughout the world.1 It is one of four

non-communicable diseases – diabetes, cardiovascular disease, chronic respiratory disease and cancers – which account for 60% of global deaths.2 Their cumulative financial burden worldwide in 2008 was estimated to be $2.35 trillion USD and prevalence of disease is projected to increase exponentially.3 Almost 350 million people have diabetes, and the number expected to die from this cause is predicted

to double between 2005 and 2030.4 By 2010, its prevalence in the adult populations

of UK, USA, mainland China and United Arab Emirates had exceeded 7%,5 11%,6 15%7 and 17%8 respectively. Americans born in 2000 or later have a lifetime risk of

more than one in three of developing diabetes.9

United Kingdom

Prevalence of diabetes is estimated to range from 4-7% of the UK population. There are 400 new diagnoses every day and 90% of cases are classified as type two diabetes.10 It has also been estimated that 600, 000 people in the UK may have

borderline undiagnosed diabetes.11 The burden of type 2 diabetes related morbidity

and mortality is considerable for both patients and to local and national health economies. The associated financial costs are significant.12 Expenditure on diabetes in the National Health Service may be as high as 10% of total yearly budget13 and 10-20% of patients in hospital have diabetes (this group also has disproportionately longer in-patient episodes and increased costs).10 As many as 50% of people with type 2 diabetes have complications at diagnosis14, which may have been detectable up to seven years previously and the onset of type 2 diabetes as long as twelve years prior to formal diagnosis.15

Strong social and ethnic gradients in prevalence are well recognized, and among those under 55 years, the poorest 20% of the UK population have twice the prevalence of the richest 20%.16 Type 2 diabetes reduces life expectancy by 10 years with concomitant social inequalities in mortality.17 18 Type 2 diabetes develops a decade earlier and is four times more prevalent in South Asian people, who have

more complications and increased mortality compared with White people.19-21

Diabetes is also more common in Black African/Caribbean populations.

Inner North East London

Inner North East London comprises the boroughs of Tower Hamlets (population 241, 747)22, Newham (population 265, 688)23 and City (population 9502) & Hackney (population 229,036).24 These make up three Primary Care Trusts as City & Hackney are grouped together. The estimated population size overall using Greater London Authority estimates from the most recent Joint Strategic Needs Assessments and the Tower Hamlets Population Change and Growth Model is 745, 973. The registered general practice population is significantly higher due to patients remaining on lists after leaving the area and patients who do not live in the boroughs.

(6)

5 | P a g e

In this report the general practice registered population was 881, 896. Over the last six years the registered population has increased by 143,000 (increases of 25% in

Tower Hamlets, 19% in Newham and 13% in Hackney)25. The number of practices

has remained relatively stable at approximately 145.

Quantifying prevalence of type 2 diabetes is problematic due to undiagnosed disease, uncertain population sizes, high population mobility and inaccuracies in disease recording and registration data. Using data from the quality outcomes framework (QoF - a remuneration programme for all GPs based on clinical performance) overall numbers of registered persons with diabetes over 16 in Inner North East London can be estimated to be 45, 688 (Tower Hamlets 13, 770; Newham 18, 467; City & Hackney 13, 451). Number of undiagnosed persons with

diabetes is estimated to be 4840.26 Local data from the Clinical Effectiveness Group

estimate a standardised prevalence of approximately 7% (10% in the South Asian ethnic group). Diabetes registers have been steadily increasing; in 2011 the register size increased by approximately 10%. These changes are more marked in Tower

Hamlets and Newham.25

Importantly Inner North East London is ethnically diverse and of low socioeconomic status compared to England as a whole. Joint Strategic Needs Assessments reveal non-white ethnic groups to make up approximately 50% of the population in Tower Hamlets, 40% in Hackney, 20% in the City and 70% in Newham. Certain ethnic groups suffer low-literacy and obesity. The combination of ethnicity (generational and genetic risk) and cultural/linguistic barriers combined with poverty, increase risk of diabetes significantly. The three areas ranked 3rd (City&Hackney), 4th (Tower Hamlets) and 7th (Newham) for the index of multiple deprivation in 2007 in England.27

Newham

Newham has the highest prevalence of diabetes of all boroughs in London. Increasing rates in children and young people are thought to be linked to the increase in obesity (Newham has the second highest rates of obesity for reception year children in England). Prevalence of diabetes was 1.67 per 1000 population (ONS, Census 2001) in 16-24 year olds in 2008 (one third were type 2 diabetes). The non-white ethnic groups are at increased risk: doctor-diagnosed diabetes was 2.5-5 times greater than the general population for Black Caribbean and South Asian groups. It is common in Newham for diabetes to be thought of as a normal part of life, and it is of concern that this may have affected the population‟s view of the serious health consequences.23

City & Hackney

City & Hackney has the lowest prevalence of diabetes across Inner North East London. The standardised rate of registered persons with diabetes on the register has increased from 4.2-5.4 between 2004-2010, compared to 5.9-7.4 for Newham and 5.9-7.2 for Tower Hamlets.25 It is recognized locally that there is a large undiagnosed population. In the over 40 age group prevalence is 11.1%, and practice level prevalence in this group vary from 4.7% to 20.3%.24

(7)

6 | P a g e

Tower Hamlets

In Tower Hamlets middle age, late diagnosis and co-morbidities feature strongly as characteristics of the existing diabetic population. In 2006 Picker Institute Europe surveyed diabetes patients from ten general practices. Out of 856 selected at random from the practices, 340 returned completed surveys. Eleven percent reported type 1 diabetes, 53% type 2. Findings indicated that for some patients diagnosis was made late – 16% started insulin within three months of diagnosis. The majority (40%) were middle aged (40-59), and 77% were on medications for other conditions.28 High prevalence overall in Tower Hamlets is partly attributed to the increased risk of disease in South Asian communities. Prevalence is steadily increasing and is being seen more often in younger people. Local estimates indicated there were approximately 1800 people with undiagnosed diabetes by the end of 2009-2010. Large increases are expected with an average prevalence of

10.1% expected by 2030.22

Risk of Type 2 Diabetes

Shifting the focus to prevention

Type 2 diabetes makes up the vast majority of diabetes cases. Presently the weight of healthcare resources and public health interventions focus on those who have

undiagnosed disease and those with established diabetes – with efforts mainly

focused on controlling disease, with monitoring of biochemical parameters such as HbA1c. This approach alone is unable to deal with increasing prevalence. This is especially relevant in Inner North East London. Projections indicate that the diabetes register size will expand to approximately 60, 000 by 2015 and 80, 000 by 2020.25 As many of these individuals who will account for this increase in prevalence are currently undiagnosed or at high risk of developing diabetes, now is the time for action to prevent this increase by 2020. The weight of public health intervention needs to urgently shift to the left on the population curve shown overleaf in figure 1. By targeting those at risk of developing type 2 diabetes, disease prevention has the potential to reduce incidence, prevalence, morbidity, mortality, healthcare use and increase quality of life. However, it is also clear that individual approaches to risk factor management are a necessary but insufficient solution. The Foresight Report

concluded “...a bold whole system approach is critical – from production and

promotion of healthy diets to redesigning the built environment to promote walking, together with wider cultural changes to shift societal values around food and activity.” 29

Risk factors for Type 2 Diabetes

Population changes in energy consumption and reduced physical activity are responsible for the current epidemic with susceptibility contributing to individual variation. Type 2 Diabetes results from a complex gene-environment interaction for which a number of risk factors are well-documented. However, the precise interaction of these and other risk factors with one another is a complex process which varies both within and across populations.30-34 A well known and often cited list of risk factors is shown in Table 1.31 35-39

(8)

7 | P a g e

Figure 1: Moving the focus of public health interventions to prevent diabetes Common risk factors for type 2 diabetes mellitus

Age Family history of diabetes

Gender Cardiovascular disease

Ethnicity Smoking

Deprivation Lack of physical activity

Body mass index Diet

Waist circumference Impaired glucose metabolism Antihypertensive medication Diabetes in pregnancy Steroid medication

Table 1: Common risk factors for type 2 diabetes mellitus

Diabetes risk scores

Given the increasing burden of type 2 diabetes to patients and healthcare economies, and the knowledge of amenable risk factors, there has been significant interest in the last 20 years in predicting the risk of developing type 2 diabetes.

Risk predication is a complex statistical area concerned with estimating the likelihood of any individual (or population) developing an outcome (usually a disease) of interest, based on a series of known risk factors. Epidemiologists and statisticians have been striving to produce weighted models and scores which are perceived as sufficiently simple, plausible, affordable and implementable to be adopted widely in clinical practice.40 41 Identifying ways of assessing risk of type 2 diabetes from existing research populations or routinely collected data have been particularly attractive as it utilises obtainable resources without the need for de novo data collection or invasive procedures, such as blood glucose or HbA1c testing.

(9)

8 | P a g e

Multiple models for assessing risk of developing type 2 diabetes have been developed. Problems with existing models have included sample size, validity, and lack of consideration of the effects of ethnicity and deprivation. The latter weakness is of particular importance given that diabetes is more prevalent in those of South Asian ethnicity and in those of low socioeconomic status.10

Diabetes risk prediction models and scores are calculated from cohorts of non-diabetic individuals who are followed over time until some have developed diabetes, and an ensuing assessment made based on risk factor measurements as to which were significant for those who developed disease (derivation). A good risk score is usually defined as one which accurately estimates individuals‟ risk i.e. predictions based on the score closely match what is observed (calibration); the score distinguishes reliably between high risk people who are likely to go on to develop the condition and low risk people who are less likely to develop the condition (discrimination); and it performs well in new populations (generalisablity). Validating a risk score means testing its calibration and discrimination either internally (by splitting the original sample, developing the score on one part and testing it on another), temporally (re-running the score on the same or a similar sample after a time period), or – preferably – externally (running the score on a new population with similar but not identical characteristics from the one on which it was developed).42 43 Caution should be exercised when extrapolating a risk model or score developed in one population or setting to a different one (e.g. secondary to primary care, adults to children or one ethnic group to another).44

We systematically reviewed 145 risk models and scores for type 2 diabetes

internationally.45 This demonstrated that numerous risk scores now exist based on

readily available data and provide a good but not perfect estimate of the chance that a non-diabetic adult will develop diabetes in the medium-term future. We deliberately did not select a single, preferred diabetes risk score in this review. There is no universal „ideal‟ risk score, since the utility of any score depends not merely on its statistical properties but also on its context of use, which will also determine which types of data are available to be included.46 47 Assessment of context of use enabled us to select one score for use in East London.

QDScore for East London

The QDScore31 is a risk prediction score which gives a ten year estimate of risk of developing type 2 diabetes. The risk factors considered are: age, gender, ethnicity, Townsend score of deprivation (which includes unemployment, car ownership, owner occupation and overcrowding), family history of diabetes, history of cardiovascular disease, smoking status, treated hypertension, current corticosteroid usage and weight and height. It is currently the only risk instrument to incorporate ethnicity and deprivation together and have been validated in a prospective study population (using the QResearch database). The QResearch database is composed of a generalisable sample from England and Wales of 11 million electronic records

(10)

9 | P a g e

The QDScore is suitable for East London based on a number of different factors:

1. Intended users (general practitioners) have access to the EMIS record system in the majority of practices, and the fields required for calculating QDScore have a high level of data completeness.

2. The target population has high ethnic diversity and deprivation which the QDScore accommodates.

3. It is able to be used opportunistically through a web portal, but it is possible that it could be automatically calculated on EMIS allowing automatic calculation for every patient. Resource implications are therefore potentially low.

Methods overview (see appendix for more detail)

It was convenient, based on the distribution of the data, to classify risk as low, medium and high. The table below shows how this was applied to the data.

10 year risk of developing type 2 diabetes (%) Level of risk

0-9.99 Low

10-19.99 Medium

20-100 High

Table 2: Classification of risk

QDScore has been derived and validated on populations aged 25-79 free from diabetes at baseline. A cross-sectional analysis of 519, 288 patient records was undertaken and QDScore calculated for each record in this age range. The table below shows the flow of data through the study.

Tower Hamlets City&Hackney Newham Total Total number of Practices 36 45 64 145

Number of practices with data available 35 40 60 135

Registered Population 268,130 266,577 347,189 881, 896

Aged 25-79 174,596 177,468 216,779 568, 843

Free from Diabetes 163,275 167,685 199,488 530, 448

Data available for analysis 163,088 166,762 189,438 519, 288

Postcode available for analysis (excludes 2 further practices for technical reasons)

157,045 n/a n/a 157, 045

Table 3: Flow of data through study

The QDScore includes some variables which are shown as potential predictive variables in later comparisons such as high BMI and deprivation. Therefore certain tables and results, including high BMI and deprivation, or South Asian ethnicity are associated with greater QDScores. This represents colinearity between the outcome (QDScore) and some predictor variables (e.g. high BMI, high deprivation, South Asian ethnicity) so the results are not unexpected.

(11)

10 | P a g e

Results

Overall risk of type 2 diabetes for Inner North East London

Key finding

1 in 10 people (51, 061) were at high risk of developing type 2 diabetes within ten years.

Other points of interest

The data was heavily right skewed with most people being at low risk of developing type 2 diabetes at ten years (410, 801).

The overall median QDScore was 2.84% with a large difference between those at low risk (1.8%), medium risk (13.8%) and high risk (30.9%).

11.1% of people were at medium risk of developing type 2 diabetes at ten years (57, 426).

Of those at high risk the majority (36, 403) had between 20-40% risk of developing type 2 diabetes at ten years.

10 year risk of developing type 2 diabetes (%)

Number of people in category % of sample in category

0-9.99 (low) 410,801 79.1

10-19.99 (medium) 57,426 11.1

20-100 (high) 51,061 9.8

Total 519,288 100

Table 4: Overall risk of type 2 diabetes

(12)

11 | P a g e

Breakdown by age, sex and ethnicity Key finding

More than twice as many South Asians (16.4%) were at high risk compared to the White (7.5%) ethnic group.

The median age of people at low risk was 34 compared to 49 for those at medium risk and 54 for those at high risk.

2.1% of 25-39 year olds were at high risk compared to 20.1% of 40-79 year olds.

10.3% of males were at high risk compared to 9.2% of females.

The overall median QDScore varied between different ethnic groups: White 1.9%, South Asian 4.4% and Black 4.3%.

Ethnicity Number of people at high risk Number of people in category

Black 10082 82036

White 16069 214542

South Asian 22126 135000

Other 1528 31947

Not recorded 1255 55763

Table 5: Number of people at high risk in each ethnic group overall

(13)

12 | P a g e

Breakdown by level of deprivation and ethnicity Key finding

QDScore and deprivation increase proportionally and this effect is higher in every category for the South Asian population.

15, 262 people are at high risk and in the highest band of Townsend Score. Those in the lowest band of Townsend Score have the lowest number of

people who are high risk (7.7%) compared to the highest band of Townsend Score (12.1%).

Within the South Asian population at high risk (22, 126) proportion of people at high risk increased and was higher than the overall population in every Townsend Score band.

Townsend score Number of people at high risk Number of people in category

-6 to 3 8794 114356

4 ₈₀₃₃ ₈₆₇₄₉

5 ₈₄₉₈ 90988

6 10430 99905

7 to 10 15262 126344

Table 6: Number of people at high risk in each Townsend Score band overall

(14)

13 | P a g e

Breakdown by co-morbidity, BMI and QRisk

Key finding

Cardiovascular co-morbidity, gestational diabetes, increasing BMI and QRisk (only includes those over 30) all substantially increase the chance of being high risk for developing type 2 diabetes.

57.2% with CVD, 48.0% with hypertension, 41.3% with chronic kidney disease, and 18.1% with gestational diabetes were at high risk for developing type 2 diabetes, compared to 9.8% of the population overall.

76, 162 people had a BMI greater than 30, and 42.8% of them were high risk for developing type 2 diabetes.

For QRisk 0-9, 9.7% (15, 516) were at high risk for type 2 diabetes, compared to 31.1% (12, 487) for QRisk 10-19, and 47.7% (9839) for QRisk ≥20.

Comorbidity/BMI/QRisk Number of people at high risk Number of people in category

CVD (IHD/stroke/TIA) 5637 9864 Hypertension 23102 48169 eGFR<60 2905 7026 Gestational diabetes 446 2466 BMI>30 32564 76162 QRisk≥20 9839 20629 Overall 51061 519,288

Table 7: Number of people at high risk in each co-morbidity group overall

(15)

14 | P a g e

Results for Newham

Overall risk of type 2 diabetes

Key finding

10 year risk of developing type 2 diabetes (%)

0-9.99 (low) 142524 75.2

10-19.99 (medium) 24401 12.9

20-100 (high) 22513 11.9

Total 189438 100

Table 8: Overall risk of type 2 diabetes in Newham

(16)

15 | P a g e

Breakdown by age, sex and ethnicity Key finding

A higher proportion of South Asians (15.3%) were at high risk compared to the White (10.5%) and Black (12.2%) ethnic groups.

The overall median QDScore between different ethnic groups was: White 3.2%, South Asian 4.0% and Black 4.3%.

Black 4425 36302

White 5562 52975

South Asian 11712 76701

Other 447 8188

Table 9: Number of people at high risk in each ethnic group in Newham

(17)

16 | P a g e

Breakdown by level of deprivation

Key finding

QDScore and deprivation increase proportionally.

2854 people are at high risk of type 2 diabetes and in the highest band of Townsend Score.

Those in the lowest band of Townsend Score have the lowest number of people who are high risk (10.7%) compared to the highest band of Townsend Score (13.8%).

Median QDScore for those in the lowest band of Townsend Score was 3.4% overall compared to 4.3% in the highest band of Townsend Score.

-6 to 3 ₆₁₀₉ ₅₇₃₀₈

4 ₅₆₆₈ ₄₉₂₀₅

5 ₄₈₄₅ 38607

6 3014 23404

7 to 10 2854 20654

Table 10: Number of people at high risk in each Townsend Score band in Newham

(18)

17 | P a g e

Key finding

31,842 people had a BMI greater than 30, and 44.5% of them were high risk for developing type 2 diabetes.

For QRisk 0-9, 14.0% (6406) were at high risk for type 2 diabetes, compared to 35.8% (4974) for QRisk 10-19, and 51.9% (3970) for QRisk ≥20.

Comorbidity Number of people at high risk Number of people in category

Table 11: Number of people at high risk in each co-morbidity group in Newham

(19)

18 | P a g e

Results for Tower Hamlets

Overall risk of type 2 diabetes

Key finding

10 year risk of developing type 2 diabetes (%)

0-9.99 (low) 131690 80.8

10-19.99 (medium) 16094 9.9

20-100 (high) 15304 9.4

Total 163088 100

Table 12: Overall risk of type 2 diabetes in Tower Hamlets

(20)

19 | P a g e

Breakdown by age, sex and ethnicity

Key finding

Black 1111 11668

White ₄₉₃₉ ₇₇₂₂₆

South Asian ₈₇₃₁ ₄₈₀₇₈

Other 294 10817

Table 13: Number of people at high risk in each ethnic group in Tower Hamlets

(21)

20 | P a g e

Key finding

7405 people are at high risk of diabetes and in the highest band of Townsend Score.

-6 to 3 ₉₇₅ ₂₇₀₂₂

4 ₁₀₀₉ ₁₅₄₉₅

5 ₁₆₉₃ 22885

6 4201 39673

7 to 10 7405 57582

Table 14: Number of people at high risk in each Townsend Score band in Tower Hamlets

(22)

21 | P a g e

Key finding

18, 930 people had a BMI greater than 30, and 44.5% of them were high risk for developing type 2 diabetes.

Table 15: Number of people at high risk in each co-morbidity group in Tower Hamlets

(23)

22 | P a g e

City and Hackney

Overall risk of type 2 diabetes Key finding

10 year risk of developing type 2 diabetes (%) Number of people in category % of sample in category 0-9.99 (low) 136587 81.9 10-19.99 (medium) 16931 10.2 20-100 (high) 13244 7.9 Total 166762 100

Table 16: Overall risk of type 2 diabetes in City & Hackney

(24)

23 | P a g e

Breakdown by age, sex and ethnicity

Key finding

Black 4551 34066

White ₅₅₆₄ ₈₅₃₄₁

South Asian ₁₆₈₉ ₁₀₂₂₁

Other 785 12940

Table 17: Number of people at high risk in each ethnic group in City & Hackney

(25)

24 | P a g e

Key finding

5013 people are at high risk of type 2 diabetes and in the highest band of Townsend Score.

-6 to 3 ₁₆₉₃ ₃₀₀₂₆

4 1356 22049

5 1964 29496

6 3204 36828

7 to 10 5013 48108

Table 18: Number of people at high risk in each Townsend Score band in City & Hackney

(26)

25 | P a g e

Key finding

25390 people had a BMI greater than 30, and 39.2% of them were high risk for developing type 2 diabetes.

CVD (IHD/stroke/TIA) 1570 3037 Hypertension 6924 15360 eGFR<60 61 284 Gestational diabetes 39 266 BMI>30 9958 25390 QRisk≥20 2659 6448 Overall 13244 166762

Table 19: Number of people at high risk in each co-morbidity group in City & Hackney

(27)

26 | P a g e

Mapping sub-section

Percentage at high risk by General Practice

Map 1: Percentage of patients at high risk of type 2 diabetes at general practice level

Map 1 shows the proportion of patients at the level of an individual general practice at high risk of developing type 2 diabetes within ten years. The prevalence of high diabetes risk in this smoothed version of the data varied from 4.1-16.7% across 135/145 practices. A higher risk band stretches from Tower Hamlets in the East, from around the area of the Royal London Hospital, to the North East side of Newham.

(28)

27 | P a g e

Percentage at high risk and Index of Multiple Deprivation

Map 2: Percentage at high risk and Index of Multiple Deprivation

Map 2 illustrates the variation in prevalence of high type 2 diabetes risk across lower super output areas in Tower Hamlets from 0.0-17.3% of the non-diabetic population. General practices and hospitals are also shown. The areas of highest prevalence for type 2 diabetes risk were distributed on either side of the A11 which transects the borough and corresponds with well-known deprived housing estates and high-rise blocks of flats on either side of this road. The map of Index of Multiple Deprivation scores by lower super output area showed a near-identical geographical distribution with high type 2 diabetes risk.

(29)

28 | P a g e

Heat maps of high risk

Map 3: Percentage high risk overall and South Asian at high risk (heat maps)

Map 3 shows similar information as Map 1 but displayed as a globally smoothed surface over the entire geographic area. The prevalence of high type 2 diabetes risk overall in this smoothed version of the data varied from 5.1-13.8%. This way of visualising the data depicts – somewhat more dramatically – a high-risk „hot‟ band running west to east through the deprived housing estates and much lower-risk „cool‟ areas in the more affluent parts of the borough. The second map shows the same technique but only for the South Asian at risk population. The prevalence of high type 2 diabetes risk is higher ranging from 9.5-23.7% and is concentrated more in the West of the borough.

(30)

29 | P a g e

Numbers of people at high and medium risk

Map 4: Percentage at medium and high risk with numbers of people

Map 4 illustrates the variation in prevalence of high type 2 diabetes risk across lower super output areas in Tower Hamlets as in Map 1 (0.0-17.3%) and prevalence of medium type 2 diabetes risk (6.2-34.4%) of the non-diabetic population. Map 3 also displays the number of people at risk per lower super output area.

(31)

30 | P a g e

Increased risk associated with South Asian ethnicity

Map 5: Percentage high risk overall and South Asian at high risk

Map 5 shows the prevalence of high type 2 diabetes risk overall and in the South Asian population across lower super output areas in Tower Hamlets. The fifth quintile differs in the South Asian population as prevalence extends to 30.3% highlighting the higher proportion of South Asians at high risk.

(32)

31 | P a g e

Maps with links to the determinants of health

Map 6: Ring maps highlighting links to selected determinants of health

Map 6 shows two maps of prevalence of high type 2 diabetes risk by middle super output area. In this depiction of the data, prevalence of type 2 diabetes risk ranges from 3.8 to 13.7%. Each middle super output area is shown linked to two separate bands of determinants of health.

Overall, the ring map provides a striking visual display of type 2 diabetes risk and the ring provides a novel way of displaying determinants of health at a small area level.

(33)

32 | P a g e

Discussion

Significance of findings

Risk of adults developing type two diabetes in Inner North East London was universally high: 1 in 10 people (51, 061) were at high risk (≥20%) of developing type 2 diabetes within ten years. In Newham 1 in 8 people (22, 513) were at high risk (≥20%), compared to 1 in 11 people (15, 304) in Tower Hamlets and 1 in 13 people (13, 244) in City and Hackney.

East London already faces vastly increasing rates of diabetes by 2020, with the number of diagnosed cases expected to rise from 40, 000 to 80, 000. Risk of diabetes is closely associated with development of established disease and could reasonably be expected to rise by the same factor resulting in 1 in 5 adults at high risk of type 2 diabetes by 2020.

Increasing age and male sex confer additional risk. The median age of people at low risk (<10%) was 34 compared to 49 for those at medium risk (10-20%) and 54 for those at high risk (≥20%). 2.1% of 25-39 year olds were at high risk (≥20%) compared to 20.1% of 40-79 year olds.

Interventions to reduce risk

We know from cohort studies that early detection of established diabetes improves

outcome, though the evidence base for screening the entire population is weak. 48 49

In those with impaired glucose tolerance or impaired fasting glucose, landmark trials

from China,50 Finland51 and USA52 reduced incident cases of type 2 diabetes by up

to 33%, 50% and 58% respectively via lifestyle changes (increased exercise, weight loss) and/or pharmacotherapy, though changes may be more modest in a non-trial population. Some have argued that because combining known risk factors predicts incident diabetes at least as effectively as impaired glucose metabolism, a diabetes risk score may be a better and more practical means of identifying individuals for preventive interventions than either a glucose tolerance test or a fasting blood

glucose level.53 Others favour targeting assessment of diabetes risk in those whose

glucose metabolism is already impaired on the basis that interventions in these groups are particularly effective.54

Incidence of type 2 diabetes can be reduced by lifestyle and behavioral measures.52

55

Individual lifestyle choices are made in a social and cultural context (e.g. an „obesogenic environment‟) which is (to some extent at least) amenable to locality-based interventions (e.g. in relation to food policy, walkability of public spaces,

community action groups and so on).56 Key to public health prevention is identifying

specific risk groups for targeted interventions as we have done in this study. Prevention may be feasible through national, community and individual measures,

which may reduce development of diabetes by 0.5% - 75%.57-59

Associations with risk

It is well established that social and ethnic diversity of populations heavily influence chronic disease risk e.g. South Asians are four times more likely to develop diabetes than the White population and more likely to die of complications.21 60 In our study ethnicity and risk of diabetes were closely associated. More than twice as many

(34)

33 | P a g e

South Asians (16.4%) were at high risk (≥20%) compared to the White (7.5%)

population. However, high risk (≥20%) was not confined to South Asians. In

Newham 10.5% of the White population was at high risk (≥20%) compared to 6.5%

in Tower Hamlets and 6.5% in City & Hackney.

Socio-economic deprivation was associated with increased risk and ethnicity increased this association. Those in the lowest band of Townsend Score have the

lowest number of people (7.7%) who are high risk (≥20%) compared to the highest

band of Townsend Score (12.1%). Within the South Asian population at high risk (22, 126) proportion of people at high risk (≥20%) increased and was higher than the overall population in every Townsend Score band.

Obesity and cardiovascular co-morbidity substantially increased risk of developing type 2 diabetes. For example, 76, 162 people in the cohort had a BMI greater than 30, and 42.8% of these were high risk (≥20%) for developing type 2 diabetes. 57.2% of people with cardiovascular disease, 48.0% with hypertension, 41.3% with chronic kidney disease, and 18.1% with gestational diabetes were at high risk (≥20%) for developing type 2 diabetes, compared to 9.8% of the cohort overall.

For QRisk ≥20, 47.7% (9839) were at high risk (≥20%) for type 2 diabetes. This overlap signifies the importance of incorporating QDScore within the NHS Health Checks programme, so that identification of risk of type 2 diabetes can move from opportunistic to systematic.

Uses of risk scores

Whilst most authors of papers describing diabetes risk scores have hypothesised (or appear to have assumed) a clinical mechanism of action (i.e. that the score would be used by the individual‟s clinician to target individual assessment and advice), the limited data available on impact studies suggest that a particularly promising area for further research is interventions which prompt self-assessment (i.e. lay people measuring their own diabetes risk).

Not everyone at high risk is interested in coming forward for individual preventive input, nor will they necessarily stay the course of such input. It follows that in areas where aggregated data from electronic patient records is available, the diabetes risk

scores may be used as a population prediction tool – for example to produce

small-area statistics of diabetes risk across a population, thereby allowing targeted design and implementation of community-level public health interventions.61 Small-area mapping of diabetes risk may be a way of operationalising the recently-published NICE guidance on diabetes prevention, which recommends the use of “local and national tools … to identify local communities at high risk of developing diabetes to assess their specific needs”.57

Mapping diabetes risk

In this study geospatial linkage using individual electronic records was only possible in Tower Hamlets. It revealed that high risk of type 2 diabetes was associated geographically with deprivation and ethnic diversity.

At small area level determinants of health can be displayed to aid local needs assessment. Caution needs to apply to the findings in this study which

(35)

34 | P a g e

recategorised risk in terms of components already included within the QDScore. This statistical colinearity potentially skews results making them appear more significant.

Taking a geospatial view of health information such as population at risk of disease complements a traditional epidemiological approach to such data. Epidemiologists use statistical tests, arithmetic adjustments, and critique causality claims and data. By contrast cartographers use geospatial visualisation, utilise classing breaks (e.g. quintiles), and critique symbolisation.62 These different paradigms have an important complementary role. Quantitative analysis identifies statistically significant trends; cartography brings meaning and local relevance. Yet merely converting routine epidemiological data into maps runs the risk of oversimplifying complex data and misunderstands the purpose of geovisualisation, which is to represent data spatially. Grouping and classing data for mapping is an interpretive process, and „points of interest‟ to which the eye is drawn on a map may or may not correspond to statistically significant relationships between variables as determined by traditional epidemiological approaches.

The key aim in health mapping is not to identify statistically significant relationships, but to gain firstly insight, then understanding of the ways in which health status varies over space, and to reveal the potential drivers behind this variation. In our research, by identifying areas of highest prevalence of greater type 2 diabetes risk in relation to small areas, local general practitioners, public health specialists and planners can be aware of increased risk and possible causes in their locality, so as to target individual and population interventions. Such „local‟ information may be unlikely to emerge from statistical analyses alone, especially in social and ethnically diverse inner city areas.

Resources and skills in handling health information in order to commission new interventions and services may be limited, particularly where they relate to dual responsibility of both local authorities and health providers for the health of local populations. Geospatial mapping offers one option to address these deficiencies and present diverse information about health and its wider determinants in an accessible format to support commissioning and planning expertise. It is possible, though somewhat speculative at this stage, that investment in the skill base needed for this approach may prove a sound investment in the longer term.

Conclusion

Urgent public health action needs to be taken not just to improve early diagnosis and management of type 2 diabetes but to reduce risk of type 2 diabetes across the whole population. Further local action is required to establish the combination of policy changes, lifestyle interventions and pharmacotherapy that have the potential to reduce risk of type 2 diabetes with a view to decreasing incidence of established disease.

(36)

35 | P a g e

Appendices

Methods

QDScore variables

QDScore is most accurate when electronic general practice records are complete. However, if variables are missing QDScore approximates a score using assumptions. The table below shows how the score handles missing values.

Missing value QDScore response

Family history Assumes no family history

Townsend score Assumes value is 0 (national average)

BMI Substituted value used based on prediction algorithm using age, sex, ethnicity, smoking status, treated hypertension and cardiovascular disease

If BMI is out of range the processor substitutes a BMI of either 20 or 40 Ethnicity Defaults to White British

Smoking Assumes not a smoker

Table 20: QDScore and missing values

Treated hypertension, cardiovascular disease and current use of corticosteroids only contribute to the score if diagnosed. Townsend score is calculated based on postcode. In Inner North-East London Townsend Score is automatically displayed on electronic records.

Ethnicity grouping

For the purposes of QDScore calculation, ethnicity codes were grouped into 17 categories based on the 2001 census. General Practices in Inner North East London have access to 155 ethnic group codes. These were converted to 17 (using a standard process described with the processor manual) in order to use the QDScore calculator. After combining the calculated QDScore with the clinical variables, ethnic group was collapsed from the 17 categories used in the score calculation to five for ease of analysis. Ethnic categories were collapsed from 17 to 5 as follows: White (British, Irish, other White), Black (White+Black African, White+Black Caribbean, African, Caribbean, other Black), South Asian (Bangladeshi, Pakistani, Indian, other Asian or White+Asian), Other (Chinese, other ethnic groups, other mixed groups), and Not stated or Missing (not recorded). The final category comprised: truly not stated (missing), not disclosed, or was coded at too high a level to be useful (effectively missing). Individuals who reported being of mixed Black or mixed South Asian were grouped with their parent ethnic minority group for reasons of biological plausibility.

Data quality

Completeness of general practice records in our selected cohort aged 25-79 years without diabetes (n = 519, 288) was as follows: age (100%), gender (100%), ethnicity (91.6%), Townsend deprivation score (99.8%), BMI (76.5%), smoking status (96.4%) and family history of diabetes (22.9%).

(37)

36 | P a g e

Distribution of QDScore

The distribution of QDScore was plotted and was determined to be heavily skewed to the right (as shown in the figure below). Hence medians and ranges were deemed to be most suitable for analysis as oppose to means and standard deviations. 0 10 20 30 40 % 0 20 40 60 80 100 QDSCORE value

Distribution of QDSCORE values across east London in adults aged 25-79

Figure 18: Distribution of QDScore

Quintiles and coding

QDscore was grouped into quintiles of type 2 diabetes risk for analysis. In some cases the bottom quintile, relating to a risk of less than 20% was split in half for further exploration. Certain other variables had to be categorised and quintiles of risk were used for much of the analysis, as shown in the table below.

Variable Coding

Quintile of Risk 0-19.99, 20-39.99, 40-59.99, 60-79.99, 80-100 Quintile of Townsend Score -6 to 3, 4, 5, 6, 7-10

BMI 10-19.99, 20-24.99, 25-29.99, 30-34.99, 35-49.99, 50-70 Grouping of QRISK 0-9, 10-19, 20-29, 30-39, 40+

Table 21: Quintiles and coding techniques

Analysis techniques

QDScore results were paired with clinical variables and sub-group analysis performed looking for important trends. This was partly guided by pre-selected analysis, ad-hoc analysis and in response to consultation with local general practitioners and public health specialists.

(38)

37 | P a g e

Mapping

Mapping was not possible in City&Hackney and Newham as postcode was not available. Therefore mapping sub-analysis with individual level data was only performed in Tower Hamlets.

Data sources, extraction and management

We used two complementary data sources: postcode with clinical risk factors for individual residents of Tower Hamlets, drawn from general practice electronic records; and social and environmental determinants of health, drawn from the cross-sectional data set, local authority registers and nationally available data at lower superoutput area level (relating to around 400 households/1000-1500 people) or middle super output area (2000 households/5000-7200 people).

Using the electronic general practice record system, a cohort was identified comprising all non-diabetic individuals aged 25-79 years in Tower Hamlets from 35 out of 36 general practices that used the same computer system. Data download was carried out on secure N3 networks. In order to overcome the information governance hurdle of preventing postcode linking to clinical variables it was necessary to first download clinical variables attached to a pseudonymised identifier (n = 163, 275 – „dataset 1‟). And then, postcode was downloaded separately attached to the same pseudonymised identifier (n = 159,353 – „dataset 2‟). The reduction in numbers was due to two practices that could not share postcode due to technical reasons. We converted postcodes (n = 8911) to lower super output area (n = 130) using an electronic lookup table. Dataset 2 (with lower superoutput area, but without postcode) was linked using the pseudonymised identifier to dataset 1. Thus, each individual record in the final dataset comprised a set of individual-level clinical risk factors plus a lower super output area level indicator of geographical locality which could be related to local and nationally available statistics.

Our local authority dataset, extracted at middle super output area, comprised: [a] fast food outlets per capita (n=371); [b] green spaces per square kilometre, and [c] population density per square kilometre. Fast food outlets were identified using local authority registry data for codes X15 „takeaway‟ and X17 „restaurants‟. All X17 codes were manually reviewed by two researchers and premises unlikely to serve fast food as a major part of their business based on their registration details were removed. This step was necessary because large corporate fast food chains such as McDonalds were registered as „restaurants‟ rather than „takeaways‟. Green spaces are quantified at the lower super output area level using the Generalised Land Use Database (GLUD) from 2005, which provides data on the area (in square kilometres) in each lower super output area dedicated to public green space. This does not include private gardens. Population density was defined as the total population size of the middle super output area divided by the area in square kilometres. This was calculated from the Office for National Statistics (ONS) mid-year population estimates for 2010, the most recent available.

Three different geospatial mapping techniques were employed using ArcGIS version 9.263 and Adobe Illustrator version 10. In the „basic‟ map the high-risk population was displayed by lower super output area as a proportion of the denominator (non-diabetic adults aged 25-79). A second basic map was created of the Index of

(39)

38 | P a g e

Multiple Deprivation score 2010 to allow a visual comparison between high risk of diabetes and a different indicator of deprivation than that used within the QDScore. A list of GP practices and hospitals were located using their postcode. They were located in GIS using the centre of each postcode. This analysis was performed to demonstrate the potential usefulness of informing local practice geographical needs assessment.

The „heat maps‟ assigned the proportion at high risk to the population-weighted centroid for each lower super output area. A kriging procedure (which uses a global semivariogram algorithm) was used to create an interpolated surface of risk. Heat maps thus offered a statistically „smoothed‟ presentation of the data in which the lower super output area blocks were no longer visible.

The „ring map‟ is a technique which allows factors of interest (such as putative environmental determinants) to be displayed circumferentially around a map. To produce these, we aggregated data to the level of middle super output area (n = 31) and presented as quintiles of risk. The following data were assembled for each middle super output area: [a] fast food outlets per capita; [b] percentage of non-green space; and [c] population density per square kilometre. Using a validated adjustment procedure, we divided each of these into highest quartile, middle 50% (2nd and 3rd quartiles), and lowest quartile61. The ring map thus gives a less granular picture of the geographical distribution of a variable but allows additional mapping of factors that might influence this variable in each locality.

A second ring map displaying risk factors from the cross-sectional dataset and unemployment score from the index of multiple deprivation was also created using a similar technique. Finally, four choropleth maps were created to compare high risk, medium risk and high risk in South Asians.

We were able to create one map for all three Primary Care Trusts using data at the level of the general practice. We used the EMIS code of each general practice to identify all registered patients aged 25-79 at high risk of diabetes as the numerator, and all patients aged 25-79 without diabetes as the denominator, therefore calculating a proportion at high risk for each general practice. This enabled us to geospatially map high risk of diabetes across a larger area, including Tower Hamlets, Newham and City & Hackney. In total 519, 288 records were used across 135/145 practices. The post code of each general practice was used with the most recent Office for National Statistics (ONS) postcode look-up table (August 2010) which identified an exact location in space for the general practice with a grid reference. One general practice‟s postcode did not have a grid reference as it was a new-build. For this practice we used an adjacent postcode to locate a grid reference. Several practices (n=20) had the same postcode. For these the final digit of the x + y co-ordinate was changed by 1 so that they could be separated in space by approximately 3 metres. Proportions of high risk individuals per practice (n=135) was mapped using a Kriging procedure, which generated a heat map on the basis of a semi-variogram algorithm. Kriging estimates the value of risk between data points where the value of risk is known.

(40)

39 | P a g e

Data quality for mapping

In Tower Hamlets records could not be generated or were removed if [a] The general practice was not able to share the data for technical reasons (n = 3922) or patient permission was withheld (n = 187); [b] the individual record contained no postcode (n = 29) or lower super output area was not calculable from the available postcode (n = 275); [c] the geographic location was outside Tower Hamlets (n = 1813); or [d] there was a mismatch between records in set 1 and set 2 (n=4). This left 157,045 records for analysis (96.2%).

Reducing the list of restaurants to those with a major business purpose of takeaway food resulted in a total sample of 371 outlets, shown in the table below.

Codes X15 +x17 895

Removed no postcode 62

Removed as staff restaurant, kitchen or canteen 142

Removed as usage unclear 74

Removed as Cafe 149

Removed as Bar 8

Removed as Restaurant 88

Removed as closed 1

Final included 371

(41)

40 | P a g e

Figures & Maps

Figure 1: Moving the focus of public health interventions to prevent diabetes ... 7

Figure 2: Overall number of people at high risk ... 10

Figure 3: Percentage of each ethnic group in the high risk category overall ... 11

Figure 4: Percentage of high risk people in each Townsend Score band overall ... 12

Figure 5: Percentage of high risk people in each BMI group overall ... 13

Figure 6: Overall number of people at high risk in Newham ... 14

Figure 7: Percentage of each ethnic group in the high risk category in Newham ... 15

Figure 8: Percentage of high risk people in each Townsend Score band in Newham ... 16

Figure 9: Percentage of high risk people in each BMI group in Newham ... 17

Figure 10: Overall number of people at high risk in Tower Hamlets ... 18

Figure 11: Percentage of each ethnic group in the high risk category in Tower Hamlets ... 19

Figure 12: Percentage of high risk people in each Townsend Score band in Tower Hamlets ... 20

Figure 13: Percentage of high risk people in each BMI group in Tower Hamlets ... 21

Figure 14: Overall number of people at high risk in City & Hackney ... 22

Figure 15: Percentage of each ethnic group in the high risk category in City & Hackney ... 23

Figure 16: Percentage of high risk people in each Townsend Score band in City & Hackney ... 24

Figure 17: Percentage of high risk people in each BMI group in City & Hackney ... 25

Figure 18: Distribution of QDScore ... 36

Map 1: Percentage of patients at high risk of type 2 diabetes at general practice level ... 26

Map 2: Percentage at high risk and Index of Multiple Deprivation ... 27

Map 3: Percentage high risk overall and South Asian at high risk (heat maps)... 28

Map 4: Percentage at medium and high risk with numbers of people ... 29

Map 5: Percentage high risk overall and South Asian at high risk ... 30

(42)

41 | P a g e

Tables

Table 1: Common risk factors for type 2 diabetes mellitus ... 7

Table 2: Classification of risk ... 9

Table 3: Flow of data through study ... 9

Table 4: Overall risk of type 2 diabetes... 10

Table 5: Number of people at high risk in each ethnic group overall ... 11

Table 6: Number of people at high risk in each Townsend Score band overall ... 12

Table 7: Number of people at high risk in each co-morbidity group overall ... 13

Table 8: Overall risk of type 2 diabetes in Newham ... 14

Table 9: Number of people at high risk in each ethnic group in Newham ... 15

Table 10: Number of people at high risk in each Townsend Score band in Newham ... 16

Table 11: Number of people at high risk in each co-morbidity group in Newham ... 17

Table 12: Overall risk of type 2 diabetes in Tower Hamlets ... 18

Table 13: Number of people at high risk in each ethnic group in Tower Hamlets ... 19

Table 14: Number of people at high risk in each Townsend Score band in Tower Hamlets ... 20

Table 15: Number of people at high risk in each co-morbidity group in Tower Hamlets ... 21

Table 16: Overall risk of type 2 diabetes in City & Hackney ... 22

Table 17: Number of people at high risk in each ethnic group in City & Hackney ... 23

Table 18: Number of people at high risk in each Townsend Score band in City & Hackney ... 24

Table 19: Number of people at high risk in each co-morbidity group in City & Hackney... 25

Table 20: QDScore and missing values ... 35

Table 21: Quintiles and coding techniques ... 36

City & Hackney Newham Tower Hamlets

City & Hackney

Newham

Tower Hamlets

Diabetes Risk

2011

Acknowledgements

Executive Summary

Contents

Introduction

Results

Overall risk of type 2 diabetes for Inner North East London

Results for Newham

Results for Tower Hamlets

City and Hackney

Mapping sub-section

Discussion

Appendices

Methods

Figures & Maps

Tables