5.4.2
Data sources
5.4.2 Data sources
The different priorities adopted by EU Member States for the collection
of data on chronic diseases hamper the establishment of common tools to monitor
diabetes policy on a large scale, including an accurate measurement of the
prevalence of the disease, risk factors, quality of care, and population
outcomes.
In the year 1989, increasing awareness of the problem led to a meeting
in St. Vincent (Italy) where representatives of Governmental Health Departments
and Patients’ Associations from EU countries joined diabetes experts under the
aegis of WHO regional Offices for Europe and the European Region of the
International Diabetes Federation (IDF), European Region, to agree on a set of
fundamental recommendations (Table 5.4.1). The Declaration acknowledged the
need to optimise health care in all EU countries through the definition of
targets for the reduction of complications and the development of an
information infrastructure allowing constant monitoring of the progress made.
For a number of reasons, among which an objective difficulty in
measuring and exchanging of data on a large scale in a timely manner, the
St.Vincent's objectives are still very relevant 20 years later.
Tracking quality of care is paramount to prevent diabetes complications,
but it is not an easy matter to realise it Europe-wide, for the presence of a
target population that is so large and diverse. Diabetes itself is a
multifactorial disease, with a large number of variables entering the quality
equation that must be taken simultaneously into account. Yet, the organization of diabetes
information would represent an almost ideal model for chronic diseases, with
very little technical obstacle to impede it: a huge pile of evidence is
available, there is agreement on how to contrast it, it is clear what precisely
needs to be measured.
However, collecting standardized and comparable data across countries
remains a difficult job, mainly of collaborative nature, for which the support
of health professionals is crucial, given their role in providing accurate
clinical information.
On a technical level, computerised medical records may be harmonised by
information systems linking individual clinical items to other databases
routinely available. Some of the best diabetes registers in Europe have adopted
innovative strategies to maximise the information available on diabetics,
although their accuracy and reciprocal comparability seems still relatively
limited.
Methodological problems must still be resolved:
·
in most cases
population-based denominators are not known, since disease management programs
and/or diabetes registers only cover those diagnosed with the diseases, and
other sources are not available;
·
diabetes status is
heterogeneously classified: earlier diagnosis due to increasing awareness of
diabetes and to the diffusion of opportunistic screening among high risk individuals
overestimate prevalence; different portion of cases with less severe diabetes
are more likely to be recorded in some regions
·
epidemiological
conclusions can be drawn on the basis of average national indicators (e.g.
blood pressure increase by classes of age), but results would be different by
directly using individual records, or at least sub-national averages (effect of
different sources of variation, or ecological fallacy).
Relative to the last point, the issue of comparability becomes an
obvious limitation in the frequent case where different countries/regions adopt
different sets of indicators, an aspect that still limits the validity of local
outputs at a broader international level.
Table 5.4.1. Recommendations St.Vincent Declaration 1989:
Diabetes mellitus is a major and growing European health problem, a
problem at all ages and in all countries.
It causes prolonged ill health and early death.
It currently (1989) threatens at ten million European citizens.
It is within the power of national governments and health departments
to create conditions in which major reduction in this heavy burden of disease
and death can be achieved.
Countries should be given formal recognition to the diabetes problem
and deploy resources for its solution.
Plans for the prevention, identification and treatment of diabetes and
particularly its complications - blindness, renal failure, gangrene and
amputation aggravated heart disease and stroke, - should be formulated at
local, national and European levels.
Investments now will earn great dividends in reduction of human misery
and in massive savings of human and material resources.
General goals and 5yr targets should be achieved by the Organised
activities of the medical services in active partnership with diabetic
citizens, their families, friends and workmates:
-> Management of their own diabetes and education for it
-> Planning, provision and quality audit of health care
-> National, regional and international organisations for
disseminating
information about health maintenance
-> Promoting and applying research
|
5.4.2.1. National
and regional registries
The construction of population-based diabetes registers represents an
obvious solution to the problem of data collection to fulfil precise
epidemiological requirements: each citizen diagnosed with T1DM/T2DM within a
defined population enters a master index including individual characteristics
at baseline, and at each visit clinical measurements, procedures, and outcomes
are accurately recorded.
To make the profile more accurate, such information can be linked to
administrative data including hospital discharges, pharmaceuticals, etc via a
unique identifier assigned to the patient. This way, denominators are well set
(population from a precise catchment area), and main epidemiological measures
may be correctly estimated.
Unfortunately, only few examples of this kind exist in Europe, showing a
heterogeneous degree of completeness, accuracy, and standardization.
At the National level, it is worth mentioning the cases of Sweden
(Gudbjörnsdottir et al. 2003) and Denmark (Carstersen et al. 2008), while at
the regional level the best referenced example is that of Tayside, Scotland
(Morris et al 1997; Boyle et al 2001), followed by others more recently, among
which Umbria, Italy (Massi Benedetti et al. 2006) has played a role in
proposing its innovative approach as a basis for data exchange across
collaborating registers in Europe (Carinci et al 2006).
In other cases, registers are client-based, i.e. they use only a portion
of the information required, either by collecting data in a disease management
program or one or more specialist clinics (Pruna et al. 2002; Azzopardi et al.
1995; Beck et al. 2001), or by extracting a list of subjects from databases
that do not involve clinical measurements (e.g. hospital discharges),
frequently without the presence of a diabetes diagnosis field validated by an
expert clinician.
Client-based registers may provide valuable information on the average
results obtained on a selected population, but their results are limited for
policy, as we are uncertain on the level of representativeness of denominators,
which in many cases lead to reports that are overoptimistic and do not include
those hard-to-reach, that are more likely to experience bad outcomes.
5.4.2.2. IDF
Diabetes Atlas
The International Diabetes Federation (IDF) has published regularly the
Diabetes Atlas, containing estimates of the prevalence of
diabetes and impaired glucose tolerance. Data of health expenditure are also
provided. Currently,
reported estimates refer to more than 200 Countries for the years 2007 and forecasts are given for 2025.
5.4.2.3. Quality
of care monitoring
The St.Vincent Declaration paved the way for the definition of quality
of care (QOC) information across Europe.
A landmark of QOC monitoring in Europe has been undoubtedly represented
by the “Diabcare” stream of projects, directly favoured by the Declaration and
funded by the EU across the 90s. Diabcare allowed agreeing a basic set of
parameters to be uniformly collected, as well as a common set of targets
(indicators) that can be conveniently used to compare own results against
national and international averages.
The approach of a common benchmarking system for quality improvement is
still valid today and underpins many monitoring systems currently existing in Europe, and beyond. Quality of care can be measured by a range of processes and outcomes
indicators.
The OECD basic set of quality indicators undoubtedly reflect the
original situation of Diabcare, showing that a large number of indicators may
result into an ambitious task if confronted with everyday practice. As a matter
of fact, out of only nine indicators judged to be immediately feasible (Greenfield
et al 2004, see Table 5.4.2), only three were available in the first report
(i.e. lower extremity amputations rates, annual eye examination and poor
control of Glycated HaemoglobinHbA1c, see Armesto et al. 2006) due to problems
in national representativeness of data and other comparability issues.
More recently, an opportunity for fine monitoring of QOC in primary care
has been created by national professional contracts, which have included target
indicators as a basis for budgeting. In the UK, data collection on a range of
indicators (of which several for diabetes) takes place on a regular basis
through the National Diabetes Audit and the Quality and Outcomes Framework.
Table 5.4.2. OECD
indicators
5.4.2.4. Health
Surveys
Health Interview Surveys (HIS) are based on self reported information
provided by the participants, whereas Health Examination Surveys (HES) include
in addition also validated information through physical examination and blood
analysis.
It is well known that the self reporting of the HIS can cause
measurement problems on a range of individual due to recall bias and imprecise
estimates. Differences between different national surveys persist, particularly
for sensitive topics e.g. heavy drinking, smoking etc. The usual recommendation
in this field is that surveys should include national population-based samples
and must be repeated at regular time-intervals.
A HES, provided that the sample is well drawn, delivers the most
reliable information on the disease status. As such, it should be a component
of national/international health monitoring for different parameters. However,
projects to finalise a series of HES at the European level are still in their
initial phase.
Detailed methodologies have been described by the HMP Health surveys in
the EU: HIS and HIS/HES evaluations and models, (01/09/2002).
5.4.2.5. Sentinel
Surveillance Network
Another possibility to obtain information was sought through the
Sentinel Practice Surveillance Network (SPSN). In several EU countries, primary
care based sentinel practice surveillance networks have been established, to
recruit clusters of motivated GP’s to perform surveys on different items. Among
these items, diabetes prevalence and diabetes management have been included
(Fleming, 2004).
In this case participants are motivated and the standards are well
defined. However, representativeness at a national level must be carefully
ascertained, as recruitment on the basis of self-motivation can be very prone
to selection bias. Furthermore, only primary care is included and mainly
diagnoses are monitored. Currently, SPSN is considered an interesting but not
an optimal data source.
5.4.2.6. Hospital
discharge records
Hospital Discharge Records (HDRs) have been used for a long time to
abstract information on procedures and outcomes experienced by patients
identified with a diagnosis of diabetes mellitus.
The establishment of national reimbursement systems in most countries,
mostly based on Diagnosis Related Groups (DRG), has favoured consistent and
uniform coding for all admissions, and the definition of national data formats
that have been fairly constant during the last years.
Normally, an HDR involves the registration of individual
characteristics, plus a variable number of diagnoses and procedures (based upon
the standardized ICD classification) administered during the hospital stay, and
other main aspects related to the admission/discharge (admission from other
hospital, discharge at home, death, etc.). The main hospital discharge
diagnosis usually determines the reimbursement for the hospital, and data are
generally accessible by request for research and analysis through
national/regional health departments.
Several methodological problems exist with the use of HDRs
Coding for primary and secondary diagnosis may be biased by
reimbursement algorithms and highly heterogeneous across hospitals and
territories. An admission for diabetes, unless linked to the occurrence of
complications, is increasingly considered to be inappropriate, particularly as
a first diagnosis, which leads it to be frequently under-reported in all
instances. Furthermore, harmonisation of hospital data is still to be realised
across the EU and it will require considerable work in the future.
Nevertheless, HDRs may considerably increase information content on
diabetes through linkage with other databases: diabetic patients may be
identified by other means, including pharmaceutical prescriptions, ambulatory
and specialist visits, etc. The pattern of hospitalisation may become much more
complete through a unified list, contributing significantly to define estimates
of rates of complications that are frequently recorded in hospitals.
A particular case of outcome is that of mortality: HDRs may provide
information on fatality rates, but they can also hide part of the results as
patients may be discharged when conditions get critical. To obtain unbiased
event rates after hospital interventions is necessary to organize a mortality
register that is lacking in most EU countries at the national level.
5.4.2.7.
Insurance/reimbursement records
Insurance/reimbursement schemes (RS) are increasingly making use of
specific national unique identification numbers, including the use of smart
cards, which involve the use of a specific flag that would identify diabetes as
a specific condition granting access to a range of ad hoc services.
Information on age, sex and linked medication, doctor’s visits and
special diagnostic or therapeutic interventions can be obtained in an anonymous
and very reliable way using this source.
However, there is usually no clinical information being collected,
meaning that most indicators cannot be estimated using this source.
However, RS can provide useful information to complete an existing
register, through a central identification number that can be used for quality
control purposes.
5.4.2.8. National
drug sales
Some studies have utilised national drug sales to estimate prevalence of
different pathologies (e.g. diabetes mellitus). National annual drug sales are
recorded in most countries.
Through this source and the standard classification of ATC and DDDs
(Defined Daily Doses) it is possible to derive an indirect estimate of
prevalence and the pharmaceutical use made by diabetic patients.
A limitation of this method lies in the variations of mean daily doses
adopted by different countries (Papoz 1993), which poses issues of
comparability of the results.
5.4.2.9.
Conclusion
Different data sources provide very different information on diabetes.
Registers may become the most accurate source, but they need to be
complete both in terms of individual measurements and coverage of the target
population. In all cases, they do not provide us with information on
undiagnosed patients and the general population. This must be extracted from
other sources.
Through self-reporting and sentinel networks, all diagnosed patients
will be included in statistical reports, whereas information obtained through
hospitals and health insurances only includes patients receiving a treatment or
having been hospitalised.
Health examination surveys detect both diagnosed and undiagnosed
patients which is impossible health interview surveys. Whether the evaluated
cohort is representative for the population in a given geographical area should
be carefully considered in each of the above.
As a matter of fact, there are only few examples of truly
population-based, harmonised, and complete diabetes sources. The case of Tayside, Scotland provides a good example of diabetes information system through which
clinical practice and statistical information may get the best estimates using
large scale record linkage.
The human as well as economic burden of diabetes necessitates the
development and implementation of a surveillance system on diabetes and its
complications of this nature across Europe. Close monitoring of the impact of
therapeutic approaches and new medications may help to improve the outcome and
reduce the cost of this chronic disease.
To make targets clearer and more realistic, the EUDIP project, funded by
the European Commission, has delivered a short-list of core and secondary
indicators that is now given considerable attention and are presented in Tables
5.4.3 and 5.4.4.
Table 5.4.3. EUDIP core indicators and their availability in EUCID
over 20 countries
Acronyms:
ICD 9: Diabetes mellitus 250; ICD 10: Diabetes mellitus E10–14; HES: Health
Examination Survey; HIS: Health Interview Survey; UNN: Unique national Number;
RS: Reimbursement Structure; SPSN: Sentinel Practice Surveillance network; HDR:
Hospital Discharge Records.
Definitions
Annual incidence of diabetes in children (0-14 year), with type 1 and 2 not separated, is
defined by EUCID as the number of new cases of diabetes in children in one year
per 100.000 children.
Prevalence of diabetes, defined by EUCID as the ratio of the number of cases
of diabetes present in the population per 1000 individuals in that population
using the definition of diabetes provided by WHO/ADA criteria (1998-9), is
extremely variable.
Prevalence over 25 yrs, an EUCID indicator, is very
similar to the pattern found for all ages. Almost in all countries, prevalence
reaches a peak over 75 years of age, with values all over 100/1000 for EUCID
participating countries. Notably, the difference is really striking with the
IDF Atlas 2006, reporting a range between 40/1000 (UK) and 118/1000 (Germany), and a median of 87.5/1000. These different results, probably deriving from the
lack of uniform surveillance in the EU, emphasize the importance of the problem
as well as the need for an immediate action.
Annual incidence of blindness due to diabetic retinopathy is defined as
the ratio of the number of new cases of blindness in one year due to diabetic
retinopathy/total number of cases of blindness in the same year tracked by
independent blindness national registries, where blindness is defined as legal
blindness (according to the national legislation). EUCID found that the
percentage for this indicator varied from 2% (Turkey) to 15% (Germany) amongst the four countries that could provide data.
Annual incidence of dyalisis and/or transplantation is defined as the number of new cases with
dialysis and/or transplantation (renal replacement therapy) per 100,000
individuals in the diabetic population in one year.
Prevalence (stock) of dialysis/transplantation per 100.000 diabetic population is defined as the number of all cases with
dialysis/transplantation per 100,000 individuals in the diabetic population.
Mortality is the most fundamental outcome indicator of all,
through which countries can evaluate the long term impact of health policies.
This indicator is defined as the annual death rate in patients who have as
primary or any cause of death diabetes mellitus/100,000 general population. Despite
the simplicity to measure such condition, mortality data in diabetes are not very reliable. This
is because, with the exception of diabetes coma, most diabetic patients die
from macrovascular complications, which lead to record diabetes not as a
primary cause, and often not even as a secondary cause.
Table
5.4.4. EUDIP secondary
indicators and their availability in EUCID over 20 countries
Definitions:
In terms of
clinical management in diabetes, parameters
include blood glucose management, blood pressure, blood lipids, kidney
functions and microalbuminuria.
The complete set is difficult to obtain, thus it is important to concentrate on
the most relevant factors.
Impaired
glucose tolerance is
defined as the percentage of general population with impaired fasting glucose,
defined as a fasting plasma glucose equal or above 6.1 mmol/l and below 7.0
mmol/l (fasting plasma glucose ≥6,1 mmol/l and <7,0 mmol/l).
Measurement
of glycated haemoglobin (HbA1c) in the last 12 months is measured as the percentage of the total
diabetic population that had their HbA1c measured in the last 12 months. EUCID
results were obtained with very different sources, and the fact that for many
of them the result was extremely positive (60% countries above 90%) casts some
doubts on the actual reliability of these estimates and are not reported here.
Level of
HbA1c>7.0% is an
indicator of poor management causing intermediate outcomes that can lead to
severe complications. It is defined as the percentage of total diabetic
population with HbA1c tested and a value of HbA1c above 7.0%. This indicator is
also measured by OECD, albeit with a different cut-off value (9.5%) that makes
data less comparable and applicable at EU level.
Measurement
of total cholesterol in the last 12 months is an important parameter for lipids in the diabetic
population. It is measured as the percentage of total diabetic population that
had their total cholesterol tested in the last 12 months.
Total
cholesterol level>5 mmol/l is an important indicator of lipidic control that is measured as the
percentage of diabetic population with their total cholesterol tested in last
12 months showing a total cholesterol above 5 mmol/l.
Measurement
of LDL cholesterol in the last 12 months is another important aspect of lipid management, as
measured by the percentage of the diabetic population that had their
LDL-cholesterol tested in the last 12 months.
LDL
cholesterol level >2.6 mmol/l is an important indicator of lipidic control that is
measured as the percentage of the diabetic population with LDL-cholesterol
tested in the last 12 months and presenting a value above 2.6 mmol/l.
Measurement
of HDL cholesterol in the last 12 months is also taken as an indicator for lipids, measured by
the percentage of the diabetic population that had their LDL-cholesterol tested
in the last 12 months.
HDL
cholesterol level <1.0
mmol/l for men and <1.25 mmol/l for women is measured as the percentage of the diabetic
population with their HDL-cholesterol tested in the last 12 months and HDL
cholesterol below 1.0 for men and 1.25 mmol/l for women.
Measurement
of triglycerides in the last 12 months is the last important parameter for lipids in the
diabetic population. It is measured as the percentage of total diabetic
population that had their total triglycerides tested in the last 12 months.
Triglycerides
level >2.3 mmol/l is an important indicator of lipidic control
that is measured as the percentage of diabetic population with their total
cholesterol tested in last 12 months showing a total cholesterol above 2.3
mmol/l.
Microalbuminuria
should be also tested
every year. The clinical management indicator is defined as the percentage of
total diabetic population that had microalbuminuria tested (or proteinuria
tested if persistent proteinuria exists) in the last 12 months.
An abnormal
level of albuminuria and proteinuria is an important element of clinical management. It is
measured as the percentage of diabetic population with
microalbuminuria/proteinuria tested in last 12 months with a microalbuminuria
above the threshold of normal (locally defined) or a proteinuria.
Blood
pressure control is measured through the percentage of diabetic subjects that had
their blood pressure measured in the last 12 months. The median is 95% for Germany. This clinical parameter seems to have the highest priority in Europe with a total
of 8 countries achieving a percentage of at least 80%. Interestingly, there
seems to be no clear association between the measurement of blood pressure and
age.
Level of
blood pressure is
measured as the percentage of diabetic subjects that had their blood pressure
measured in the last 12 months and reported a diastolic blood pressure above 90
and/or a systolic blood pressure above 140 mm Hg.
Smoking is defined as any type of smoking and
relates to the percentage of smokers in the diabetic population.
Levels of
BMI among diabetics are
measured as the percentage of
diabetic population that had their BMI measured or had both weight and height
available and had a value above or equal to 25/30 kg/m2.
Age at
diagnosis by age bands
is an extremely difficult indicator to collect, unless there are sophisticated
registers in place. Indeed, this fact was not reported by EUCID, due to the
small number of countries submitting data (N=2).
Fundus
inspection is a
fundamental examination to check for the existence of ophthalmologic
complications. The process indicator is measured by the percentage of diabetic
population that had their eye fundus inspected in last 12 months. This
indicator is the only one reported by the OECD (Armesto et al 2006) as fit for
international comparisons in diabetes.
Proliferative
retinopathy is defined
as the percentage of diabetic population that had their eye fundus inspected in
the last 12 months and were diagnosed with a proliferative retinopathy.
Indicator on
timely laser treatment of retinopathy is defined as the
percentage of diabetic population that had their eye fundus inspected in the
last 12 months and were diagnosed with a proliferative retinopathy and had
laser treatment within 3 months.
Control
of serum creatinine is
defined as the percentage of diabetic population that had their serum
creatinine tested during the last 12 months.
Renal failure is measured as the percentage of
diabetic population with creatinine tested in the last 12 months and End Stage
Renal Failure (ESRF), which in turn is defined as serum creatinine above or
equal to 400 umol/l (WHO).
The annual incidence of major amputations is defined as
the number of new cases of major amputations (above ankle) per 100,000
individuals in the diabetic population in one year. Amputations are also
recorded by OECD, albeit not included in the list of indicators fit for
international comparisons.
The annual incidence of stroke is defined as the number
of new cases with stroke (both ischemic and bleeding) per 100,000 individuals
in the diabetic population in one year.
The annual incidence of any myocardial infarction is defined as the
number of new cases with any myocardial infarction per 100,000 individuals in
the diabetic population in one year.