  |
|
|
||||
|
||||||
| Journal of the New Zealand Medical Association,
31-March-2006, Vol 119 No 1231
|
||||||
|
Ethnicity data and primary care in New Zealand: lessons from
the Health Utilisation Research Alliance (HURA) study
Health Utilisation Research Alliance (HURA)
In New Zealand, there is strong evidence of an association
between ethnicity and health exposures, experiences, and outcomes. This is
reflected in differential access to healthcare and significant disparities in
health status.1–6 It is also
consistent with substantial international literature on the relationships
between ethnicity and health.7,8
Comprehensive, high-quality ethnicity data are essential to
the mapping of health trends by ethnicity; the development of effective policies
and strategies; and the measurement, monitoring, and elimination of ethnic
disparities in New Zealand.
Inconsistencies in approaches to the definition of ethnicity
and the collection of ethnicity data have resulted in inaccuracies in New
Zealand ethnicity statistics. In health, for example, this has been reflected in
the under-reporting of both morbidity and mortality for certain ethnic
groups.9
Issues with the consistency and quality of data sets within the health
sector have been noted, including an historical lack of standardisation across
the health sector and missing
data.10
In
addition, gaps in ethnicity data and in the availability of ethnic breakdowns of
health data, have contributed to an incomplete picture of ethnic trends and
disparities.
Information on
health utilisation patterns by ethnicity in New Zealand is not yet
comprehensive. A standardised approach to the collection of ethnicity data for
hospitalisations and self-identification of up to three ethnic groups in line
with the Census definition of ethnicity has been the official policy since 1
July
1996.10
Until
recently, ethnicity data have not been routinely collected in primary care (or
in other areas such as disability and aged care services). It has therefore been
difficult to have an understanding of trends by ethnicity in the primary care
sector, as available data tend to be from surveys of self-report such as the New
Zealand Health Survey, the Commonwealth Fund Survey, and from point-in-time
studies such as
WaiMedCa
and
NatMedCa.11–13
Recent
changes in funding primary healthcare in New Zealand increased incentives for
the collection of ethnicity data in primary care (as ethnicity was included as a
variable in two of the funding formulas for Primary Health Organisations [PHOs]
and as a non-formula criterion in a third) alongside other population-level
variables such as socioeconomic deprivation and age. Accurate and comprehensive
ethnicity data are also a necessary component in meeting high-level strategic
and policy goals of reducing inequalities in health. However, issues remain in
regard to the consistency and quality of ethnicity data within the health
sector.
The
Health Utilisation Research Alliance (HURA) project was a study of general
practice services undertaken to explore the relationship between ethnicity,
socioeconomic deprivation, and utilisation of primary care using routinely
collected data for a 12-month period (1 January to 31 December 2001). A
significant part of the study was working with a sample of general practices to
increase ethnicity data coverage and support standardised data collection.
This
paper outlines the methods used to increase ethnicity data coverage in the
dataset for analysis (i.e. the merging of practice collected and NHI ethnicity),
issues encountered in the study, and the implications of these for New Zealand
health services research.
MethodsGeneral
practices—This study involved data routinely collected by 25 of 37
general practice members of the Wellington Independent Practice Association
Limited (WIPA) using electronic patient management software. In 2001–2002,
when the data collection was undertaken, WIPA general practices were distributed
around Wellington and Porirua Cities and on the Kapiti Coast (up to and
including Waikanae) in the lower North Island of New Zealand and included most
general practices in the locality.
At that time, all WIPA practices (like most New Zealand
general practices) operated under a fee-for-service system with a mixture of
patient payment for consultations and partial government subsidy for low-income
patients. Patients registered with a general practice for their care but there
were no restrictions or funding implications if patients also consulted or
registered elsewhere.
All WIPA practices were invited to take part in the
study (with the exception of one practice not providing primary healthcare to a
general population); 25 practices participated. Over the study period, one
practice split into two and both practices continued to participate. Of the nine
practices declining, there was no systematic difference in practice size and
locality compared to participating practices.
Data
collection—Raw data were extracted from practice computing systems
using custom-written programmes. After extraction, data were transferred to the
WIPA computer system. Data were combined and edited, and small-area deprivation
(NZDep2001) codes were added. Individuals’ National Health Index (NHI)
numbers were encrypted and any unique identifying information removed before
analysis.
Ethnicity
data—The 25 practices participating in the HURA study undertook to
add ethnicity data to their patient registers (five practices were already
collecting patient ethnicity data prior to their involvement in the HURA study).
As ethnicity was not routinely collected by WIPA
general practices at the time of the study, researchers met with practices once
they had agreed to participate in the HURA study to establish their baseline
ethnicity coverage, discuss ethnicity data collection, and support the
development of appropriate methods for individual practices.
Ethnicity data collection was standardised as much as
possible by supporting practice personnel to collect self-reported data using
the ethnicity question in the 2001 New Zealand Census of Population and
Dwellings. The research team also supported practices by developing resources
(including a patient pamphlet and a staff card), assisting in setting up alerts
on practice software, and providing individualised feedback on progress. This
support was ongoing throughout the study.
Ethnicity data was extracted alongside other
demographic variables in 2002, for the 2001 calendar year. Prior to analysis,
missing ethnicity data were supplemented with data from the National Minimum
Data Set linked by NHI number. Ethnic groups were combined into broad aggregate
categories for analysis. People who identified with more than one ethnic group
were assigned a single ethnicity using a prioritisation process used by
Statistics New Zealand at the time of the study.
In prioritised grouping, any
patients identifying
as Māori were defined as Māori, whether or not they also identified
with another ethnic group. Priority was given secondly to Pacific, then to
Asian, other ethnic groups, and finally to European.
As NHI data included a grouping defined as
‘Other’,
people defined in this category were divided into those where a specific
ethnicity was identified other than Māori, Pacific, Asian, or European
(specified other) and those where there was no further definition (unspecified
other).
Analysis—Databases
for managing the data were constructed and edited in Microsoft Access software
and data transferred to SAS software for analysis. Comparison of ethnicity data
collected by general practices in the study with that held on the National
Minimum Data Set (NMDS) were based on 216,132 patient records from 25 practices
for whom a date of birth and NHI number were available. As it was possible for
patients to be recorded at more than one practice, this number does not reflect
individuals.
Ethics
approval—The Wellington Ethics Committee approved the study.
ResultsThe coverage of ethnicity data achieved varied by practice,
ranging from below 10% to over 90% of all patients (Table 1). Data collection
processes were developed by practices to suit their individual requirements, as
practice management software and registration procedures differed. All practices
increased their coverage during the study period. Practices that were already
collecting ethnicity data prior to joining the study were more likely to achieve
higher coverage by the end of the study period, as were smaller practices.
In general, ethnicity data coverage (as for coverage of
other demographic data) was higher at practices for patients registered with the
practice than for non-registered patients. Ethnicity was also more likely to be
recorded for registered patients who had consulted during the study period, with
ethnicity data missing for 67.2% of registered patients who did not consult in
2001 (Table 2). (This is consistent with the ethnicity data collection method of
self-identification.)
To conduct analyses on service utilisation (published
elsewhere),14 data from practice records
and NZHIS (giving priority to practice collected ethnicity) were combined to
increase coverage to the extent that 23 of the 25 practices in the study (92%)
had ethnicity data for at least 70% of registered patients. The proportions of
different ethnic groups in the data collected by the practices and from NZHIS
are compared in Table 2.
The most noteworthy differences in comparing practice
collected ethnicity with NZHIS data were differences in the proportion of people
recorded as ‘Other’, with a lower proportion of patients in the
unspecified ‘Other’ category in general practices. While only 1% of
registered patients and 0.3% of non-registered patients were coded as
‘Other’ by practices, 21% of registered and 7.7% of non-registered
patients were coded as ‘Other’ on NZHIS data.
To
understand the implications of combining data from two different sources,
ethnicity data collected by general practices were compared with corresponding
ethnicity data from the NHI at an aggregate ethnic group level (e.g. Māori,
Pacific, Asian, Other, European). The comparison showed that there was a level
of disagreement between ethnicity data collected at the practice and that from
the NZHIS (Table 3).
Of
those records where there was no ethnicity data recorded at the practice,
approximately 50% had an ethnicity coded on the corresponding NHI record. Of
those records where ethnicity data was available from the practice (85,391),
there was 47% agreement overall with the ethnicity coded on the corresponding
NHI (47.8% for Māori, 54.4% for Pacific, 39.3% for Asian, 44.9% for Other,
and 46.6% for European). NHI ethnicity data was missing for a further 30% of
records overall where ethnicity data was available from the practice. The
remaining 23% of records were coded as a different ethnic group on the
corresponding NHI record.
DiscussionAt the end of the study period there was considerable
variation in the coverage achieved at different practices. Although based on the
standardised process recommended by the HURA study, data collection processes
varied between practices as a result of differences in practice management
software and registration procedures.
For many participating practices there was limited baseline
ethnicity data already available on practice registers. Most practices therefore
had to collect this data for the majority of their patients, and this provided
challenges for some practices (including software issues, time costs, and
competing priorities). Ethnicity data were more likely to be collected for
patients who consulted, as the recommended method of ethnicity data collection
is self-identification. It thus requires some form of contact between patient
and practice.
The study period coincided with significant changes to the
way in which primary healthcare was funded and organised, including the
introduction of funding formulas for which ethnicity was a variable. While this
provided a new incentive for the collection of ethnicity data in primary care,
these changes were also associated with other impacts on practice workloads.
The study also found a divergence between ethnicity recorded
on the NHI (often collected in hospitals) and that collected in general
practice. When compared to NHI ethnicity data, practices were more likely to
have ethnicity data missing in general. However, where data were present, it was
more likely to be more specific than that on the NHI. Practices did not tend to
have a large ‘Other’ (not elsewhere specified) group, but were more
likely than NHI to have recorded a specific ethnicity within the
‘Other’ category. The size of the ‘Other’ category on
the NHI data for this study (21%) was substantially higher than that of the
total population as reflected by census data.
The proportion of people in the ‘Other’ category
in the last census was 0.69%.15 When
restricted to records with ethnicity data available from both sources, we found
that over 75% of those coded as ‘Other’ on NHI were coded as
European by the practice. A 1999 study of ethnicity data in Wellington Hospital
also found a higher than would be expected proportion of admissions coded as
‘Other’ (no further
defined).16 The likely over-representation
of the ‘Other’ category will need to be taken into account when
interpreting NHI ethnicity data.
It
is important to note that NHI data is likely to be historical and could have
been collected continuously over several years prior to corresponding
practice-collected data. Although the collection of self-identified ethnicity
data on the NHI has been official policy since 1996, there have been problems
with the consistency and quality of the data. For example, the New Zealand
Census Mortality Study (NZCMS) found a 30–40% undercount of Māori,
Pacific, and Asian ethnicities on the NHI when compared with census ethnicity
data.17
When
restricted to patients with ethnicity data available from both sources
(practice-collected and NZHIS), we found similar patterns although not as
extreme. After the merging of NHI ethnicity data with practice data (where
ethnicity was missing in the practice data), the residual misclassification
within NHI data will remain, specifically a likely undercount of Māori,
Pacific, and Asian ethnic groups, and overcount of the ‘Other’
ethnic groups.
It
is difficult to establish the extent to which the disagreement between
practice-collected and NZHIS ethnicity data is due to miscoding, issues with
data collection, or a true difference in response to the question. Indeed, there
will always be some level of disagreement between ethnicity data collected at
different times and within different contexts, even where a standardised process
is in place. This is because it is possible for ethnicity to change over time
and/or for people to respond to the ethnicity question differently within
different contexts. However, a standardised approach to data collection will
reduce the variation that is due to inconsistent data collection approaches.
Since
the study was undertaken, the Ministry of Health has launched Ethnicity Data
Protocols for the standardised collection of ethnicity data throughout the
health sector. If these protocols are implemented consistently and
comprehensively, they may address some of the issues in divergences between
practice collected and NZHIS ethnicity data. In addition, there will need to be
ongoing monitoring and auditing of the accuracy of ethnicity data within the
health sector.
High
quality ethnicity data is a valuable public health resource in terms of
monitoring health status and health indicators, and is essential for the
measurement of ethnic disparities in health. General practices need to be
supported in their collection of consistent, appropriate, quality ethnicity data
for their practice populations.
Author
information:
The
Health
Utilisation Research Alliance (HURA) includes:
Acknowledgements:
HURA is a collaborative research partnership between the Wellington School of
Medicine and Health Sciences and the Wellington Independent Practice Association
Limited (WIPA).
The research team is grateful to the general practices who
took part in the study and shared the data they collected, and to Stella Ramage
for assistance with database development. We also thank the Health Research
Council of New Zealand for funding the research and the Ministry of Health for
providing funding for the development of pamphlets about ethnicity data
collection.
Correspondence:
Donna Cormack and Bridget Robson,
Te Rōpū
Rangahau Hauora a Eru Pōmare, Wellington School of Medicine and Health
Sciences, PO Box 7343, Wellington South. Fax: (04) 385 5924; email: bridget.robson@otago.ac.nz
or donna.cormack@otago.ac.nz
References:
|
|
||||
| | Current
issue | Search journal |
Archived issues | Classifieds
| Hotline (free ads) Subscribe | Contribute | Advertise | Contact Us | Copyright | Other Journals |
|