  |
|
|
||||
|
||||||
| Journal of the New Zealand Medical Association,
06-August-2004, Vol 117 No 1199
|
||||||
|
Population need and geographical access to general
practitioners in rural New Zealand
Lars Brabyn, Ross Barnett
Accessible and appropriate health services for people living
in rural areas remains an issue of ongoing concern. The implementation of the
Primary Health Care Strategy, together
with the more recent development of primary health organisations (PHOs), both
build on earlier government policies designed to improve the level and
continuity of service provision in rural
areas.1,2 While such developments provide an
opportunity to improve the effectiveness and sustainability of rural primary
health care services, a number of important questions remain. These include
problems of funding rural care, difficulties in attracting GPs and other health
professionals to isolated rural areas, problems of high doctor turnover and
continuity of care, as well as the influence of geographical and financial
barriers, which may serve to limit the utilisation of needed
services.3,4
New Zealand, with its rugged physical environment and
dispersed rural population, poses particular problems for the location and use
of services. An important requirement, therefore, is to develop quality
information systems that highlight the physical accessibility of primary care,
and the extent to which this access varies for particular rural population
groups.5
Improving the quality of information on accessibility to
rural primary health care is also an important requirement for effective
decision-making by PHOs and District Health Boards (DHBs). The objective of this
paper, therefore, is two-fold:
The Rural Expert Advisory
Group’s report, Implementing the Primary
Health Care Strategy in Rural New Zealand. A Report from the Rural Expert
Advisory Group to the Ministry of
Health, has indicated that high
levels of deprivation are a feature of some rural regions in New Zealand, and
that the extra travel costs that rural people incur make access to primary
health care services particularly difficult for the people of these
communities.6
Therefore, an important task is to identify those areas
where problems of physical accessibility to GPs are compounded by increased
needs for care and conditions of rural disadvantage.
MethodsA geographical information
system (GIS) was used to measure geographical (physical) accessibility to GPs.
Three key methods were compared. First, population/GP ratios were calculated for
each of the 73 Territorial Local Authority areas in New Zealand (using full-time
equivalent GP data for the year 2000 provided by the Ministry of Health and
population data from the 2001 Census).
Second, since population/GP ratios are only a crude
measure of geographical access, two further methods were used: least cost path
analysis (LCPA) and an allocation technique that considers the number of GPs
available and how many people a GP can service. Both methods represent an
improvement on traditional ratio measures of GP access, as they involve more
detailed calculations of travel distances and travel times. In addition, they
are not constrained by area boundaries and aggregation problems of ignoring the
intra-district location of GPs relative to their patients.
LCPA involved calculating a least cost path algorithm
to determine the shortest travel distance and time between each of the 38,336
census meshblocks (origin nodes) in New Zealand, as well as the closest GP
practice (destination nodes). Network analysis capabilities in ARC/INFO were
used to calculate accessibility. The
nodedistance command computes distances
between all possible combinations of origin and destination nodes via the New
Zealand road network.
In this study, nodes closest to the meshblock centroids
were the origin nodes and nodes closest to GP practices (n=1390 practices
representing 3614 GPs) were the destination nodes.
To identify the closest GP for a given Census centroid,
we calculated the minimum distance for each origin to each destination. The
minimum distance to the closest GP for each centroid was the sum of three
calculations; the network distance, plus that from the meshblock centroid and GP
surgery to their closest road nodes, respectively.
The process for calculating the minimum travel time to
the closest GP is similar, except road travel time is used instead of distance.
Estimated road travel times were based on whether the road was inside or outside
an urban area, number of lanes, condition of the road surface (sealed versus
unsealed), and the bendiness (sinuosity). The road travel times, while similar
to those published by the New Zealand Automobile Association, do not take
account of the effects of travel congestion or seasonal differences. Full
details of their derivation and limitations are given in Brabyn and
Gower.7
Since the population characteristics of each Census
meshblock are known, it was also possible to calculate average travel times.
This was accomplished by multiplying the population of each centroid by the
travel time of the centroid (to determine the total time spent travelling if
everyone represented by the centroid visited the closest GP once). To calculate
the total travel time, these values were then aggregated to the level of the TA.
The average travel time was obtained by dividing this total by the TA
population.
While LCPA approaches represent an improvement on ratio
methods of determining geographic access to GPs, they can be misleading because
not all patients choose to use the closest GP.8
LCPA, therefore, provides estimates of optimum rather than actual travel
distances and times. Furthermore, it ignores the fact that some GPs have
multiple practices, especially in rural areas where these may also be partly
staffed by other health professionals for part of the time. LCPA also neglects
the capacity of a GP practice to service the surrounding population.
Many people may be unable to get an appointment because
the GP is servicing a large population and hence may choose another provider,
especially in more densely populated areas where other alternatives are
available. Despite this caveat, an allocation method was also used to estimate
variations in GP access. This involved allocating potential patients to the
closest GP practice until the practice reaches a specified capacity level. The
model then finds the next closest GP practice. Once a population has been
allocated a GP, the network travel time and distance is calculated.
The capacity used in this study was 1400 patients per
full-time GP—which is the number used by the Ministry of Health for a
full-time work load.9 The output from the
allocation method is similar to LCPA except for the addition of a capacity
constraint. However, the allocation method is also limited because of its
assumption of a uniform capacity constraint, which clearly varies between GPs
especially depending upon their gender and
age.10
Both the LCPA and allocation analyses enabled estimates
of the total population with poor geographic access to GPs to be calculated. For
the purposes of this analysis, a 30-minute threshold was chosen. Thirty minutes
is a long time to travel to a GP and, given the results of US
research,11 most persons would have expressed
dissatisfaction at having to travel for this length of time.
The three methods are first compared, followed by the
analysis of the population composition of rural areas remote from GPs. Four
measures of population need are considered; the 2001 New Zealand Deprivation
Index (NZDep200112), the proportion of the
population of Maori ethnicity, and two indicators of age (percentage less than 5
years old, and the percentage aged 65 years and over). These measures enable the
assessment of population groups that are particularly vulnerable to poor
geographical access.
The method outlined above contains generalisations that
can skew the results but are necessary for practical reasons. First, meshblocks
are represented by one central point, and the location of this point may not
accurately depict the population distribution within the meshblock (which will
be a problem with large rural meshblocks found in the South Island). There is a
new data set being developed in New Zealand (called LandOnLine) that contains
address points, which will map the location of every letterbox in the country.
This data set has been completed for many TAs and can be used to represent the
population distribution within a meshblock.
Preliminary analysis using the geographical mean of the
address points within a meshblock shows that travel times are 2–3% less
than with meshblock centroids. Therefore, the method used for this study
overestimates travel times for rural areas. The second generalisation used by
the method is that it only considers travel speeds during normal flow and does
not consider traffic congestion that is happening in urban centres during rush
hour traffic.
A temporal dimension to accessibility would be a
worthwhile research endeavour if data on travel speeds at different times of the
day were known.
ResultsComparison
of the three techniques as indicators of GP access—Figure 1 shows
the population/GP ratios by TA, while Figures 2 and 3 show the population more
than 30 minutes from a GP using the LCPA and allocation models.
Figure 1: Population (by territorial authority) per
GP
 Figure 2: Population (by territorial authority) more
than 30 minutes from a GP; using LCPA (least cost path algorithm)
 Figure 3: Population (by territorial authority) more
than 30 minutes from a GP; using an allocation model
 There are many other statistics that can be generated from
the LCPA and allocation models (including the average travel time, total travel
time, and travel distance); however, these statistics can be misleading as the
average travel time does not consider the population affected by this time.
Furthermore, a region may have a high average travel time
but only have a low population. For example, Westland District has a high
average travel time (20.8 minutes—based on the allocation model) but only
has a population of 8,091.
It would therefore be inappropriate for the Government to
use only average travel time as a basis for funding. It is possible to present
the total travel time for each district and use this for a comparison. However,
this statistic is high for populated cities because of the large populations
even though the average travel times are less than three minutes. The population
more than 30 minutes from a GP is therefore the preferred statistic to use when
representing geographical access.
The results generated from the LCPA and allocation models
can be represented at a range of scales from individual meshblock units to
national statistics. The average travel time to the GP for the whole of New
Zealand is 4.6 and 5.9 minutes for the LCPA and allocation models respectively.
The population more than 30 minutes from a GP for the whole
of New Zealand is 70,833 and 122,034 for the LCPA and allocation models,
respectively. To compare regional variations in accessibility across New
Zealand, it is necessary to choose a scale where the number of regions is
manageable, and where the regions are not too large so that important variations
within a unit are generalised.
Figures 1, 2, and 3 were compiled according to territorial
authority (TA) regions, which is a good compromise. District Health Board
regions could also be used but these cover large areas and there are significant
variations within them as shown by the territorial authority scale.
A visual comparison of the models (Figures 1–3) shows
that they all provide a different representation of access. Specifically, there
is weak correlation between ‘population per GP’ and the LCPA and
allocation models (0.14 and 0.17, respectively). However, this is to be expected
given that they are completely different methods for measuring accessibility.
Population per GP does not consider the distribution of the
population or the GPs within a particular TA whereas the LCPA and allocation
models are not constrained by area boundaries in the same way that the
population per GP method is. Rather, times from GP locations to the closest
population enumeration point are calculated. The effect of this difference can
be seen in the comparative results for Waikato District. Using the population
per GP method, Waikato District has the highest ratio (2343 people per GP).
However using LCPA and calculating population (more than 30 minutes travel time
from a GP), the district is mid range in its accessibility (33 out of 73).
This disparity is because many of the GPs that service the
Waikato District are located in the city of Hamilton and towns of Cambridge and
Morrinsville, which are all within 8 km of the Waikato District boundary.
Waikato District is predominantly a rural district that is serviced by Hamilton
City, which has its own TA status. This detail is neglected in the ratio
method.
There is a strong correlation between the LCPA and
allocation models (0.88), which is expected since the methods are similar. Where
there is a significant variation between these models, then this indicates that
functional access is a problem.
Selwyn District and Waikato District have substantial
differences between these models (3432 and 3303 people respectively) and there
are 17 TAs with differences between the models of over 1000 people. All these
TAs can be characterised as rural. Virtually all the major urban TAs have no
difference between the LCPA and allocation models. As expected, the population
more than 30 minutes from a GP using the allocation model is either equal or
more than the LCPA model.
The LCPA and allocation models use travel time that only
includes actually time spent travelling by car and not time spent loading the
car and finding a park. The travel time in cities appears very low but the
travel distance is on average less than one kilometre. It needs to be emphasised
that this analysis is intended to produce general statistics rather than
assessment of individual travel times.
LCPA and allocation models produce results that clearly
differentiate between urban and rural districts. Urban territorial authorities
have low travel times, as there tends to be many GP services within a sort
distance. Conversely, high travel times in rural districts describe the
dispersed characteristics of populations and concentrated GP locations in
provincial towns.
Variations between rural districts reflect differences in
population distribution, which in turn is related to land-use, livelihood, and
topography. For example, the Far North District is long and narrow with a large
population living outside of service towns. Conversely, the Waikato District has
a high rural population but also a high number of service towns.
GP access by population
group—While rural regions are more likely to have problems of
access to key services, an important question is the extent to which problems of
GP access vary between TAs depending upon their population characteristics. LCPA
and allocation models combined with the NZ Deprivation Index (a score of 10 is
the most deprived) and census data has been used to produce a range of
statistics for different ethnic, age, and deprivation groups—and this was
completed for each TA and DHB.
This produces large tables that are not possible to present
in a journal publication. Table 1 was generated from the allocation model and
provides a sample of this data. It includes the Far North District and Southland
District, which have the highest population that is more than 30 minutes from a
GP. They are also geographically separated as they are at opposite ends of the
country. For comparison, Table 1 also shows two urban TAs (Waitakere City (which
is part of Auckland), and Christchurch City) and statistics for the whole of NZ.
The average travel time for the different population groups is also
presented.
At a national level, there are variations in geographical
access for the different population groups; however these do not appear to be
significant (this is statistically demonstrated in Table 2, which is discussed
later).
If people are split into two groups—wealthy (NZ Dep
1-3) and poor (NZ Dep 8-10)—then it can be said that wealthy people
generally have higher travel times to GPs (as many wealthy people purchase
lifestyle blocks on the outskirts of cities, and due to the existence the
wealthy farming communities.
Table 1. Accessibility to GPs by different population
groups
Elderly people generally spend less time travelling to their
GP and this can be explained by the deliberate move many retired people make to
be closer to health services. The under 5 years group appear to be close to the
average. Maori people have higher travel times, on average, which can be
explained by the rural nature of many Maori communities. As for Pacific
Islanders, they have less travel time (on average), and this can be explained by
the large, urban Pacific Island population in Auckland.
When different TAs are studied, there can be significant
variations to these trends. It is clear from Table 1, that there is no
significant problem with geographical access for any of the population groups in
the two urban TAs, both in terms of the absolute number of people and as
percentages. This is the case with all the urban TAs. In the Far North District,
there is a complete reversal of the national trend in terms of wealthy and poor
people, although the absolute population of wealthy people is low (2454).
However, Southland District supports the national trend.
Table 2 illustrates correlations between three measures of
GP access and the socioeconomic characteristics of all TAs, and also provides a
separate analysis of the North and South Islands. Regarding average travel times
to GPs, it is clear that rural areas with lower population densities have poorer
accessibility, as do areas with larger Maori populations, especially in the
North Island.
By contrast, TAs with the highest concentrations of more
affluent groups (deprivation deciles 1-3), in general, had shorter travel
distances to care than was true for more deprived populations (deciles 8-10).
TAs with concentrations of older people (65 years and over) also had smaller
average travel times to GPs, but only in the South Island.
These patterns are also evident with respect to the two
other access measures (the % total population more than 30 minutes from a GP
based on LCPA and allocation methods). However, here the correlations are
magnified between poor access and levels of deprivation and ethnicity,
especially in the North Island. In contrast to the travel time analysis, in both
cases, the relationship between (poor) levels of geographical access and
deprivation is strong (r=-0.57 and -0.55) for the LCPA and allocation measures,
respectively.
These patterns are illustrated in Figures 2 and 3, which
show the absolute population more than 30 minutes from a GP in different TAs and
compare the LCPA and allocation techniques. The pattern is a predictable one.
Many of the traditional Maori heartlands, such as Gisborne or the Far North,
have larger populations with poorer access, but so do many of the more remote
southern TAs such as Southland or Marlborough.
For New Zealand as a whole, the LCPA analysis indicates that
70,833 people (or 1.9%) resided more than 30 minutes from their closest GP. This
figure rises to 128,034 (or 3.4%) when the results of the allocation analysis
are examined. While the latter figure may not seem particularly high (3.4%), the
proportion of the population with poor access rises to 9.9% for all rural TAs
(those outside the main metropolitan areas and regional cities), and exceeds
this margin for over half (24) of the 45 more rural TAs
Discussion
The results presented here suggest that problems of GP
access remain important for many of the more remote rural areas in New Zealand.
Many people in these areas suffer a double burden. Not only do they face long
travel times for obtaining primary care, but also since they are often deprived,
travel difficulties are accentuated.
Other research has indicated that the economic costs of
obtaining care represent a significant deterrent to low income people in New
Zealand.13,14
Malcolm13, for instance, in a survey of eight
health centres providing services to Maori and low-income New Zealanders, found
that rates of GP utilisation were substantially lower (from 37 to 74%) than the
national average of 4.5 visits per capita in 1994/95. But, given that the
centres were set up to improve access to Maori and low-income populations, and
had significantly reduced the financial barriers present in the average general
practice, then cost barriers alone did not appear to be a major factor for the
very low rates of utilisation observed; the effects of geographical and cultural
barriers were just as important.
Although the present study did not examine GP utilisation
rates in areas remote from GPs, Malcolm’s results are consistent with a
large, and longstanding, geographic literature demonstrating links between
geographic barriers and utilisation rates (both for primary and hospital care).
For instance, Haynes and Bentham15 found that
GP consultation rates, outpatient attendance rates, and inpatient admissions in
Norfolk (UK) were all found to decline with decreasing accessibility.
The groups most affected in rural areas were those with the
highest relative need for healthcare. Other research has similarly found that
distance barriers disproportionately affect poorer
patients.16 For higher-status patients,
distance barriers will have less effect on utilisation not only because of
greater levels of affluence and car ownership, but also because of a preference
to take advantage of non-local providers of both primary and hospital
care.17
However, as Girt18 and
others19,20 have found, distance may have both
a positive and negative effect on consulting behaviour. Individuals are likely
to become more sensitive to the development of disease the further they live
from a doctor, but (at the same time) distance negatively affects their
propensity to consult. The distance at which this effect changes seems to depend
upon the extent of the self-perceived illness or
condition.20
The effect of ‘distance to GP’ on ‘rates
of use’ also has implications for the use of hospital services. In rural
New South Wales, Walmsley21 found that the
chances of admission diminished the further a patient lived from hospital.
Haynes et al22 similarly found that (after
controlling for needs and provision) distance to hospital produced a
17–37% reduction for different types of admissions.
Of particular importance was their finding that distance to
GP surgeries had the effect of reducing hospital inpatient episodes—an
effect which was greatest for elective and psychiatric admissions.
These findings suggest that ‘distance’ and
‘travel time’ are important considerations, especially for rural
dwellers (who frequently express the greatest dissatisfaction with problems of
access to care).11 An accumulated body of
research thus suggests that policymakers should give greater weight to such
parameters when evaluating the availability and quality of primary care in rural
areas.
Traditionally, analyses of future directions in primary
healthcare have either neglected the importance of spatial analysis
approaches23or, where analyses have taken
place, they have been on the basis of crude GP population
ratios.24. However, such an approach, on its
own, is an insufficient basis for assessing the effects of poor access and
planning future needs.
Population-based ‘travel time’ and
‘distances to health services’, as well as an ‘analysis of the
characteristics of the population most affected by geographic barriers’
are more useful measures of GP accessibility—and we suggest that future
primary healthcare policy should pay more attention to such factors. Such
considerations will become more important as PHOs take on the task of
identifying and addressing those groups in their populations that have poor
health or are missing out on services.
The application of GIS approaches, such as those discussed
in this paper and which are beginning to be widely used in health
research,25,26 therefore provide a valuable
tool for assisting such organisations in improving access to services and the
health of the most disadvantaged. The use of GIS tools, however, requires access
to quality data. One of the most time consuming challenges of this research was
obtaining a geographically referenced database of GP practices.
Currently the GP register maintained by the NZ Medical
Council does not contain reliable information on the geographical location of GP
practices. An address of each practice is kept, but this could be the GP's
residential address. The addresses of GP practices were obtained from a
commercial data supplier, whose usual clients would likely be pharmaceutical
companies. The conversion of addresses to a GIS layer is labour intensive and
should only be done once.
Lastly, it is imperative that the Ministry of Health or the
New Zealand Medical Council maintains a geographically referenced data set that
contains New Zealand grid reference coordinates of GP practices, along with the
number of GPs working in the practice and the hours they work.
Author information:
Lars Brabyn, Lecturer, Department of Geography, University of Waikato, Hamilton;
Ross Barnett, Associate Professor,
Department of Geography, University of Canterbury, Christchurch
Acknowledgements:
This research has been supported by Public Health Intelligence, which is part of
the Ministry of Health, New Zealand. In particular, Dr Chris Skelly has been a
key person in establishing the project and suggesting possible goals. Ron King,
also from the Ministry of Health, supplied the GP locations and assisted in
reviewing the analysis results. Paul Gower, a graduate student from the
University of Waikato, provided analysis assistance. We would also like to
acknowledge and thank these people for their involvement in this
research.
Correspondence: Lars
Brabyn, Department of Geography, University of Waikato, Hamilton. Fax: (07) 838
4633; email: larsb@waikato.ac.nz
References
|
|
||||
| | Current
issue | Search journal |
Archived issues | Classifieds
| Hotline (free ads) Subscribe | Contribute | Advertise | Contact Us | Copyright | Other Journals |
|