 
|
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Journal of the New Zealand Medical Association,
05-March-2010, Vol 123 No 1310
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
A population-based approach
to the estimation of diabetes prevalence and health resource utilisation
James Smith, Gary Jackson, Brandon
Orr-Walker, Rod Jackson, Siniva Sinclair, Simon Thornley, Tania Riddell, Wing
Cheuk Chan
Counties Manukau District Health Board (CMDHB) is
responsible for the funding and provision of health services for the people of
Counties Manukau, serving around 480,000 people in 2009. It has the
fastest-growing population of any district health board (DHB), the highest
number of young people (aged ≤14 years), the highest numbers of Māori
and Pacific peoples and the highest number of people living in the most deprived
two deciles of the New Zealand deprivation
index.1
Diabetes is a leading cause of morbidity and mortality in
New Zealand2,3 and is particularly common in
Counties Manukau. Māori and Pacific peoples, who together make up around
40% of the Counties Manukau population, have a disproportionate burden of
diabetes compared with other groups.4–6
The 2006/07 New Zealand Health Survey (NZHS) estimated that Counties Manukau had
the highest prevalence of self-reported diabetes in those aged ≥15 years
of any DHB area.7 Furthermore, in Counties
Manukau around one-third of adult respondents to the NZHS were found to be
obese, with body mass indices ≥30
kg/m2,7
Initiatives to reduce the growing burden of diabetes in
Counties Manukau are underway. Let’s Beat Diabetes (LBD) is a
long-term, intersectoral strategy which draws on wide-ranging activities such as
community-based programmes, social marketing, support for primary care and
collaboration with the food industry, to prevent or delay the onset of Type 2
diabetes, limit disease progression and improve quality of life for those with
diabetes in CMDHB.8 The Chronic Care
Management (CCM) programme, a community- and primary care-focused
programme, has a diabetes module aimed at structured management of individuals
with complicated diabetes in the community.
In order to plan effectively for future community needs
related to diabetes and understand the impact of initiatives such as LBD and
CCM, timely and readily updateable estimates of the prevalence of diabetes in
Counties Manukau are necessary. To take a ‘whole of community’
approach to planning, it is also important to gain insight into community care
of diabetes, by examining factors such as community pharmaceutical prescribing
patterns and laboratory monitoring.
This study estimated diabetes prevalence and utilisation of
healthcare services in Counties Manukau using routinely collected administrative
data and compared estimates with findings for three other DHBs in close
geographic proximity.
MethodStudy population—We linked three
routinely collected administrative health databases in four DHBs: Counties
Manukau, Northland, Waitemata and Auckland. Data from the National Minimum Data
Set (NMDS—a national database containing details of all public and private
hospital discharges from 1990 onwards) were linked with 30 months of community
laboratory and pharmacy subsidy claims data for the period July 2005 to December
2007. Laboratory and pharmacy claims data containing details of reimbursement
claims for subsidised community laboratory tests and pharmaceuticals dispensed
on the New Zealand Pharmaceutical Schedule9
were obtained from the Ministry of Health.
The three databases were linked using unique National
Health Index (NHI) numbers for each case record and all NHI numbers were
encrypted by the Ministry of Health prior to analysis, to avoid identification
of individuals. Individuals were included in the aggregated data set if they had
any health service recorded in one or more of the three data sets in the 2 years
between January 2006 and December 2007 (inclusive). All deceased individuals
were removed from the reconstructed group using encrypted data from the Ministry
of Health Mortality Collection.
Data were not available for those who did not have
hospital events recorded in NMDS, or did not have claims made for subsidised
pharmaceuticals or laboratory tests (with NHI numbers documented) during the
study period. These aggregated data created a ‘reconstructed’
population of around 1.4 million people.
Identification of diabetes
cases—In this study, ‘diabetes’ refers to all forms
of diabetes mellitus. An individual from the study population defined above was
identified as a ‘case’ if he or she met any of the following
criteria: a hospital event with a principal or secondary International
Classification of Diseases, 10th edition,
Australian Modification (ICD-10-AM) diagnosis code E10-E14 ‘Impaired
glucose regulation and diabetes mellitus’, or the codes O24.0 to O24.3
which cover pre-existing diabetes in pregnancy since 1990, three or more HbA1c
test claims within 2 years (2006/07), or two or more community pharmaceutical
dispensing claims for New Zealand Pharmaceutical Schedule therapeutic group (TG)
level 2 categories ‘Diabetes’ and ‘Diabetes management’
in 2006/07.
Ethnicity—Prioritised ethnicity
was used to classify ethnicity, so that individuals were categorised into only
one ethnic group, according to a prioritised schedule, if more than one
ethnicity was recorded in different databases. Consistent with the standard
prioritisation protocol recommended by the Ministry of Health, ethnicity was
prioritised in the following order: Māori, Pacific, Asian, NZ
European/Other.10
Deprivation—Average NZDep2006
scores for the census area unit (CAU) in which a case lived were applied to each
individual as a measure of socioeconomic status. NZDep2006 is a
multi-dimensional index of deprivation used for small
areas.11 It includes nine dimensions covering
income, home ownership, household occupancy, education, employment and access to
transport and to a telephone.
Statistical analyses—Statistical
analyses used Microsoft Excel™, SPSS®
(version 13.0) and SAS® software
applications. Prevalence estimates have been calculated using both the
reconstructed population and 2006 national census estimates as denominators.
Point estimates are reported for descriptive statistics and 95% confidence
intervals are included where appropriate (95% CI). For direct standardisation,
prevalence proportions were divided into the following age groups: 0-4 years,
10-year age intervals until age 84 years, 85+ years. Standardisation for age and
sex used Statistics New Zealand national population estimates for 2006/07.
Further detail on statistical methods is available in a background
report.12
Ethical considerations—Ethical
review was not required, as all unit record data were de-identified by the New
Zealand Ministry of Health and no contact was made with participants. Only
aggregated results have been reported.
ResultsStudy population—The combined data
sets identified about 1.4 million people in all four DHBs, compared with 1.5
million in the March 2006 census estimate. The reconstructed population of CMDHB
in 2006/07 contained 427,350 people, 6% fewer than the 454,790 identified in
CMDHB in the March 2006 national census estimate. Social and demographic
characteristics of the CMDHB reconstructed population are compared with 2006
national census estimates for CMDHB in Table 1.
Diabetes prevalence—Almost 27,000
diabetes cases were identified as resident in Counties Manukau in 2006/07, while
51,000 diabetes cases were identified in the remaining three northern region
DHBs. This translated to a crude prevalence of 6.3% (95%CI 6.2%-6.4%) for
Counties Manukau and 5.3% (95%CI 5.2%-5.3%) for the remaining three DHBs, using
the reconstructed denominator. Counties Manukau had the highest age- and
sex-standardised prevalence of diabetes of any of the four DHBs.
Using the reconstructed denominator, the age- and
sex-standardised prevalence of diabetes in Counties Manukau was 7.1% (95%CI
7.0%-7.2%), compared with 5.2% (95%CI 5.1%-5.2%) for the remaining three DHBs
combined. While a difference of 0.4% was found between the two crude Counties
Manukau prevalence estimates using reconstructed and census denominators, this
difference narrowed considerably with standardisation (Table 2).
Age-standardised prevalence of diabetes by ethnicity is
presented in Table 3. Pacific women had the highest prevalence of any group,
with an age-standardised prevalence (using reconstructed denominator) of 15.0%.
Women of NZ European/Other ethnicity had the lowest diabetes prevalence of any
group, at 4.0%.
Table 1. Social and
demographic characteristics of reconstructed CMDHB population and estimated
population in 2006 national census
 Table 2. Crude and age-
and sex-standardised diabetes prevalence estimates in CMDHB, 2006/07, using
reconstructed and 2006 census denominators
  * Using the ‘reconstructed’ population as
the denominator
Community care of diabetes—In
Counties Manukau, 83% of diabetes cases had at least two glycated haemoglobin
(HbA1c) community test claims in 2006/07, compared with 82% in the other DHBs.
Just over half (52%) of Counties Manukau diabetes cases had four or more HbA1c
test claims in 2006/07, versus 44% of cases in the other DHBs. Almost 40% of
Counties Manukau diabetes cases had five or more HbA1c tests in 2006/07.
Frequency of HbA1c test monitoring appeared consistent both by ethnicity and by
socioeconomic status. Figure 1 shows the similarity between distributions of
HbA1c test frequency by ethnicity in Counties Manukau, while Figure 2 compares
HbA1c test frequency for those living in the most deprived areas with the
remainder of diabetes cases in Counties Manukau.
 Ninety-two percent of diabetes cases in Counties Manukau had
fasting lipid studies performed at least once in 2006/07 (90% in other three
DHBs) and 81% of Counties Manukau cases had two or more fasting lipid tests in
this two-year period (75% in other DHBs). Around one third of Counties Manukau
cases had five or more claims for lipid tests in 2006/07. As with HbA1c, no
important differences in serum lipid test frequency were noted by ethnicity or
by socioeconomic status.
 Use of a range of different medications by diabetes cases in
Counties Manukau was examined. Of particular value in our understanding of
quality of care was use of medications to treat secondary complications of
diabetes, such as lipid lowering agents and agents affecting the
renin-angiotensin system (like ACE inhibitors).
It was not possible from the data available to identify
which diabetes cases ‘should’ have had subsidy claims for lipid
lowering agents, although current guidelines suggest a high proportion of
diabetes cases may benefit from them. In CMDHB, 64% of diabetes cases had at
least two pharmaceutical claims for lipid lowering agents, the majority of which
were HMG CoA reductase inhibitors (statins). Minor differences were identified
in the proportion of diabetes cases accessing these medications by ethnicity and
by deprivation (Figures 3 and 4).
Seventy percent of diabetes cases had at least two
pharmaceutical claims for blood pressure-lowering medications. As expected from
guidelines, claims for agents affecting the renin-angiotensin system were
particularly common, with 61% of diabetes cases recording regular claims for
these medications. The distribution of subsidy claims for agents affecting the
renin-angiotensin system showed a relatively even spread across all ten
deprivation deciles (Figures 3 and 4). Asian diabetes cases were found to have
slightly lower utilisation of these medications than the other three groups.
Figure 3. Proportion of diabetes cases with regular
claims for agents affecting the renin-angiotensin system and statins, by
ethnicity in Counties Manukau, 2006/07
 Figure 4. Proportion of diabetes cases with
regular claims for agents affecting the renin-angiotensin system and statins, by
socio-economic status in Counties Manukau, 2006/07
 Hospital service utilisation—In 2007,
11,800 CMDHB medical and surgical hospital discharges were recorded in NMDS
among diabetes cases, 17% of all hospital discharges in that year. The average
length of stay for diabetes cases was 3.6 days, compared with 2.4 days for those
without diabetes.
Twenty-three percent of diabetes cases had one or more
medical or surgical hospital admissions in 2007, compared with only 11% of those
without diabetes in the reconstructed population. When these figures were age-
and sex-standardised, 29% of diabetes cases were found to have had one or more
admissions in 2007, compared with 11% of those without diabetes. Considerably
more Māori diabetes cases had one or more hospital admissions in 2007 than
diabetes cases of other ethnicities (Table 4).
Table 4. Proportion
of diabetes cases in CMDHB with medical/surgical hospital admissions in 2007, by
ethnicity
 DiscussionThis study used routinely collected administrative data from
community laboratory and pharmaceutical subsidy claims, together with data on
hospital discharges recorded in NMDS, to create a reconstructed population for
four northern region DHBs. It took a ‘whole of community’ approach
to explore both diabetes prevalence and utilisation of monitoring tests,
medications and hospital services. Record linkage of administrative databases
has previously been used to examine the epidemiology of diabetes in Denmark,
Scotland, Canada and Sweden.13-17
Administrative data has also been used to estimate prevalence and to review
quality of care of conditions like cardiovascular disease and
asthma.18, 19 To our knowledge, this is the
first study to estimate the prevalence of diabetes and resource utilisation in a
DHB setting in New Zealand, using record linkage of administrative data
sources.
We found diabetes to be common in Counties Manukau, with
almost 27,000 people identified as diabetes cases. The age- and sex-standardised
prevalence of diabetes was seven percent. This was the highest prevalence of any
of the four DHBs studied and almost two percent higher than the average
prevalence in the other three DHBs. Standardised prevalence estimates were
similar using reconstructed and census denominators. Māori and Pacific
people, in particular Māori men and Pacific women, had the highest
prevalence of diabetes.
Our prevalence estimates are consistent with other estimates
of diabetes in Counties Manukau. The 2006/07 New Zealand Health Survey estimated
the total number of adults aged 15+ years with self-reported diabetes to be
around 26,400, only around 600 less than our
estimate.7 The crude adult prevalence of
diabetes in this survey was 8.2%, similar to findings in our study. The LBD
2006/07 benchmark survey of 2,520 people in Counties Manukau found an
age-standardised, self-reported prevalence of 7.0% in those aged 16+ years, a
finding consistent with our estimates.5 Our
findings were also comparable to diabetes prevalence estimates by ethnicity for
adults aged 35-74 years in the Diabetes Heart and Health Survey (DHAH)
undertaken in Auckland in 2002/03, where 9.3% of respondents were found to have
Type 2 diabetes (previously- or
newly-diagnosed).4 While there were important
differences between the study populations in DHAH and our study, analysis of the
same age group for the three Auckland DHBs using the reconstructed population
gave an estimated diabetes prevalence of 9.5%.
Key strengths in this study were currency of the data, low
additional cost to the DHB of data collection, extensive detail in the three
contributing databases and the ability to link numerators with denominators from
the same dataset. Laboratory and pharmaceutical data came from the HealthPAC
General Transaction Processing System for reimbursement of community
laboratories and retail pharmacies, meaning that data became available only a
few months after claims for late 2007 were processed. All data had already been
collected for administrative purposes and there was no requirement for the DHB
to undertake any further data collection.
Substantial data were available for analysis in the final
data set. The pharmaceutical data alone contained 42 separate variables. All
numerators and denominators in the analyses came from the same reconstructed
population and social and demographic variables in the data analyses were
directly linked, thereby avoiding numerator-denominator bias in the analysis of
ethnicity20 (although ethnicity data in
hospital records can still differ substantially from self-identified
ethnicity21).
The validity of the decision rules has not yet been formally
established. Only subsidy claims for HbA1c were recorded in the data and not
laboratory values. Also, only claims for community laboratory and pharmaceutical
tests were available. Such data were unavailable for hospital admissions. Three
approaches were taken to assess the suitability of the rules. A literature
review substantiated the scope of HbA1c use and use of diabetes medications for
conditions other than diabetes. Both HbA1c and diabetes medications were
predominantly used in diabetes, although medications like metformin are
sometimes used for other purposes.22, 23
Individuals with local expertise in diabetes care, primary care, epidemiology
and clinical coding were consulted about suitability of the rules and
sensitivity analyses were undertaken using a range of test frequency and
pharmaceutical dispensing claim thresholds. Further work is underway, validating
the rules using capture-recapture methods. In the interim, we believe our
findings are a reasonable reflection of diabetes in Counties Manukau. Not only
are our estimates consistent with those of other studies, but work validating
similar rules has demonstrated high levels of sensitivity and positive
predictive value.24 Some individuals at high
risk of developing diabetes may have been misclassified as having diabetes. Such
misclassification is acceptable for planning purposes, as these people share
many of the risks experienced by those with
diabetes.25-29
Only individuals with records of hospital events or subsidy
claims during 2006/07 were included in the reconstructed population, meaning
exclusion of the remainder from the reconstructed group. ‘Residential
churn’ over the period of data collection also meant obsolescence of some
addresses, despite most recent health encounter being used to allocate address.
However, our estimated population for CMDHB was only six percent less than that
estimated by the 2006 national census and only small differences were seen
between the socio-demographic profiles of the two groups. Use of the smaller
reconstructed population as denominator resulted in over-estimation of crude
diabetes prevalence, although the effect of over-estimation was lost when
prevalence was age- and sex-standardised, as the bulk of diabetes cases occurred
in those aged 35+ years and this group had the best coverage in the
reconstructed population compared with census data.
Analysis of ethnicity using high-level categories such as
‘Pacific’ and ‘Asian’ is not without limitation and
assumes ethnicities aggregated within these groups have similar health
characteristics. This is not necessarily the case, as was highlighted in a
report on Asian health needs which found considerable heterogeneity in health
indicators between different Asian groups (such as Indian and
Chinese).30
No data were available on appropriateness of prescribing and
laboratory monitoring for individual clinical cases. For example, similarities
in community care by ethnicity did not account for differences in diabetes
severity. Results of the descriptive analysis were checked using logistic
regression and findings for resource utilisation were consistent with those
described. However, pharmaceutical and laboratory utilisation for Māori and
Pacific was similar to other groups, despite substantial differences in hospital
admission frequency. Further work is needed to understand whether resource use
and quality of care are appropriate to disease severity and are equitable across
these groups.
This study used a set of timely, comprehensive and
inexpensive data to examine diabetes prevalence and resource utilisation in
Counties Manukau. The analysis will be repeated regularly to inform planning for
the diabetes epidemic. The methods may also be used to examine other long-term
conditions and other aspects of diabetes care, such as the additional health
care cost to the DHB of diabetes.6 Disparities
in the prevalence of diabetes by ethnicity were demonstrated, highlighting again
the need for community-based primary prevention programmes such as LBD, which
aim to address this inequity.
Competing interests: None known.
Author information: James Smith, Public
Health Medicine Registrar, Planning and Funding, Counties Manukau DHB, Manukau;
Gary Jackson, Manager Public Health, Planning and Funding, Counties Manukau DHB,
Manukau; Brandon Orr Walker, Clinical Leader, Let’s Beat Diabetes,
Counties Manukau DHB, Manukau and Clinical Head Endocrinology and Diabetes,
Middlemore Hospital; Rod Jackson, Professor of Epidemiology, Section of
Epidemiology and Biostatistics, School of Population Health, University of
Auckland, Auckland; Siniva Sinclair, Public Health Physician, Planning and
Funding, Counties Manukau DHB, Manukau; Simon Thornley, Assistant Research
Fellow, Planning and Funding, Counties Manukau DHB, Manukau; Tania Riddell,
Senior Lecturer, Section of Epidemiology and Biostatistics, School of Population
Health, University of Auckland, Auckland; Wing Cheuk Chan, Honorary Research
Fellow, Section of Epidemiology and Biostatistics, School of Population Health,
University of Auckland, Auckland
Acknowledgements: This research was funded
within operational budgets of Counties Manukau District Health Board to inform
ongoing service planning within the DHB. We are very grateful to Craig Wright
(Senior Advisor Statistics and Epidemiology, Health and Disability Intelligence,
Ministry of Health) for his assistance in development of the rules for
identification of individuals with diabetes.
Correspondence: Dr James Smith, Counties
Manukau District Health Board, 19 Lambie Drive, Manukau City,
Private Bag 94052, South Auckland Mail Centre, Manukau
2240, New Zealand. Email: jameskgsmith@yahoo.co.nz
References:
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| | Current
issue | Search journal |
Archived issues | Classifieds
| Hotline (free ads) Subscribe | Contribute | Advertise | Contact Us | Copyright | Other Journals |
|