Rural–urban variation in the utilisation of publicly funded healthcare services: an age-stratified population-level observational study

In Aotearoa New Zealand, it is possible to monitor geographic variation in secondary health service utilisation due to the presence of a unique identifier for every health service user, well-maintained national administrative health datasets and a single public healthcare system. Contemporary monitoring has included “bench marking” the performance of the country’s 20 district health boards (DHBs).¹ Regional disparities in the range and quality of health services that were identified have been termed a “postcode lottery” and were an important driver of the current health system reforms.² Despite the possibility that even greater variation may exist between urban and rural areas (either within a DHB or at a national level), few rural–urban analyses have been undertaken.

The evidence that does exist is contradictory. The NZ Health Survey 2002/2003, which used self-reported data from approximately 13,000 respondents, failed to demonstrate significant rural–urban differences in hospitalisation rates or access to a medical specialist.³ In contrast, the Rural Health: Challenges of Distance, Opportunities for Innovation report published by the National Health Committee in 2010 used Mānatu Hauora – Ministry of Health administrative datasets and reported age-adjusted utilisation rates that were higher for rural than urban dwellers: outpatient services (11% higher), emergency department (ED) visits (20% higher) and public hospital use (excluding ED) (20% higher).⁴ There is some evidence of lower rural utilisation rates for individual services at a regional level; for example, CT scanning in the Southern Region.⁵

A rural–urban classification designed specifically for use in health research and policy in New Zealand, the Geographic Classification for Health (GCH), was published in August 2022.⁶ The GCH taxonomy comprises two urban categories, major urban centres (U1) and regional cities (U2), and three rural categories (R1, R2 and R3) that denote increasing rurality and remoteness. The GCH has “unmasked” rural–urban differences in health outcomes that were obscured when other rural–urban classifications were used.⁷ Mortality disparities identified by the GCH differ considerably across age bands, with younger rural residents having higher mortality rates than their urban peers, but older rural residents having mortality rates similar to or slightly lower than their urban peers.⁸ Rural Māori have consistently poorer health outcomes than rural non-Māori, frequently exceeding the ethnic inequities observed in the urban context.⁹

In June 2022, New Zealand’s parliament passed into law the Pae Ora (Healthy Futures) Bill.¹⁰ After intense pressure from the rural health sector, the legislation was altered at its final reading to include provision for a Rural Health Strategy, which was subsequently released in July 2023.¹¹ The Strategy is a high-level document that will give rise to specific rural health policy and plans in the coming years. Accurate data on rural–urban variation in health service utilisation is now needed to provide an evidence base for this policy and health service planning.

The objective of this paper is to compare age-standardised and age-stratified utilisation rates across broad categories of publicly funded health services to identify areas of significant rural–urban health service variation that warrant further detailed examination.

Methods

This population-level observational study used deidentified routinely collected data from two New Zealand government agencies: Manatū Hauora – Ministry of Health and Statistics New Zealand (Stats NZ).

Numerators

Extracts of two administrative data collections were obtained from Manatū Hauora – Ministry of Health. This included data for 2015–2019 from the National Minimum Dataset (NMDS) of hospital discharges and the National Non-Admitted Patient Collection (NNPAC) of outpatient and ED attendances. Both datasets included the person’s age at time of event, sex, ethnicity (Māori or non-Māori) and domicile (geographical unit representing the area encompassing their residential address, approximately 2,000 residents in each). Outcome measures derived from the NMDS were all-cause hospitalisations as well as cardiovascular (CVD), cancer, injury, mental and behavioural disorders and ambulatory sensitive hospitalisations (ASH). ASH are defined as hospitalisations of people less than 75 years of age “resulting from diseases sensitive to prophylactic or therapeutic interventions that are deliverable in a primary healthcare setting”.¹² Rural patients are frequently transferred between institutions in order to access appropriate specialist care. This can result in the “overcounting” of rural events. To account for this, contemporaneous admissions for an individual were grouped as part of a single continuous episode of care.¹³ Outcome measures derived from NNPAC were all specialist outpatient and ED attendances, and allied health outpatient events. The Allied Health indicator comprised of all NNPAC events with allied health purchase units, with the exclusion of community radiology. The service descriptions and their frequencies are presented in Appendix Table 1.

Age was categorised as follows: 0–29, 30–44, 45–59, 60–74 or 75+ years. Ethnicity was categorised as Māori or non-Māori. If any of the ethnicities recorded were Māori, the individual was classified was Māori.

Denominators

Census usually resident population counts for 2013 and 2018, aggregated, simultaneously, by age, sex, ethnicity and rurality, were obtained from Stats NZ. Age was obtained in 15-year bands. Census ethnicity was categorised as “Māori” or “non-Māori” using the same process used for the Manatū Hauora – Ministry of Health data. Annual estimates for 2015–2019 in each of the combinations of these variables (age [5], ethnicity [2] and rurality [5]) were obtained from linear interpolation of the Census 2013 and Census 2018 counts. Total person-years for each of the combinations was obtained from these annual estimates.

Rural–urban status

Rural and urban areas were defined according to the recently published five-level Geographic Classification for Health (GCH).¹⁴ Using the domicile concordance file, the relevant GCH category was assigned to each individual’s domicile code.¹⁵

Statistical analysis

In order to combine the numerator and denominator datasets, the person-level numerator dataset was collapsed, with counts of each outcome produced for each combination of the age (5), ethnicity (2) and rurality (5) categories (50 rows).

Crude rates were calculated separately for the total population, Māori and non-Māori for the age-strata within each of the outcome variables, per 100,000 person-years for the 6 NMDS outcomes and per 1,000 person-years for the 3 NNPAC outcomes. Incidence rate ratios (IRRs) and 95% Confidence intervals (CIs) per age group and outcome were calculated using Poisson regression and represent the ratio of the incidence rate in one of the GCH categories (U2, R1, R2, R3) divided by the incidence rate in U1 (reference category). For each outcome an overall age-standardised rate was calculated; the 2001 Census Māori population was used as the standard population for these directly standardised rates.

Data were prepared using SAS software version 9.4 for Linux.¹⁶ Analysis was undertaken using Stata/SE v17.¹⁷ Figures were produced using R.¹⁸

Results

There was an average of 1,079,000 all-cause hospitalisations per year for the period 2015–2019; 61% of hospitalisations were for residents of U1 (major cities), 20% were for U2 residents and 12%, 5% and 1% were for R1, R2 and R3 residents respectively. Of the 6.3 million specialist appointments per year, 56% were for U1 residents and 23%, 14%, 6% and 1% for U2, R1, R2 and R3 residents respectively. There were on average, 705,000 ED attendances per year, of which 53%, 26%, 13%, 7% and 1% were for U1, U2, R1, R2 and R3 residents respectively. Allied Health events were less likely to be for U1 residents; of the 980,000 per year, 41% were for U1 residents, 36% were for U2 residents and 14%, 8% and 1% were for R1, R2 and R3 residents respectively.

Age-standardised and age-stratified hospitalisation rates for each GCH category are presented separately for Māori (Appendix Table 2), non-Māori (Appendix Table 3) and for the total New Zealand population (Appendix Table 4). Age-standardised hospitalisation incidence rate ratios (IRRs) with U1 as the reference are presented in Figure 1. Results for non-admitted patient events are presented in the same format in Appendix Table 5, Appendix Table 6 and Figure 2.

View Figure 1–2.

For Māori, all-cause hospitalisation rates were highest for those living in U2 areas (regional centres), with the exception of 60+ years, where U1 and U2 rates were the same (Appendix Table 2). Māori all-cause hospitalisation rates for all rural strata were lower than the equivalent age-specific urban strata, the exception being R3 residents aged 75+ years, where all-cause hospitalisations were the same as those in the urban categories. A very similar pattern of lower rural hospitalisation rates was observed for non-Māori, with the exception of the 15–29-year-old age group in the R1 and R2 categories where the rates were higher than U1 but less than U2 (Appendix Table 3).

When the New Zealand population was considered as a whole, a clear gradient of reducing all-cause hospitalisation across the rural categories became apparent. Using U1 as the reference, within each age strata the rate for R2 residents was lower than R1, and the R3 rate lower again (Appendix Table 2). The rate for U2 residents was, however, 5% higher than for U1. Based on these data, if rural residents (R1, R2 and R3) had experienced the same crude rate of all-cause hospitalisation as those living in the cities (U1 and U2), the total number of hospitalisations nationwide would have risen by an average of 5,191 per year (or 0.5%).

Māori CVD hospitalisation rates for rural residents were lower than for U1 residents for 15/18 of the age by GCH combinations (6 age groups x 3 rural categories). At times the difference was large; for example, residents of R3 aged 60–74 years were 27% less likely (20% -33%) to have a CVD hospitalised episode of care than U1 residents of the same age. A similar pattern was observed for non-Māori.

Māori living outside the major cities (U2 and R1–R3) had lower injury-related hospitalisation rates. A slightly different pattern was observed for non-Māori aged 15–44 years, who for those living in U2, R1 and R2 (but not R3) had injury-related rates of hospitalised episodes of care that were similar to or higher than the U1 rates.

For Māori and non-Māori there was no clear pattern of rural–urban variation in cancer hospitalisations, with the possible exception of the lower rural rates in the paediatric (0–14 year) population.

Rates of mental health and behavioural disorder hospitalised episodes of care were lower overall in the rural categories for both Māori and non-Māori. The overall age-adjusted rates for rural Māori, relative to U1 Māori, were estimated to be 0.71 (R1), 0.85 (R2) and 0.77 (R3).

Ambulatory sensitive hospitalisation (ASH) rates for Māori across all rural strata were lower than U1 with IRRs that are consistently less than 0.9. In contrast, ASH rates for Māori living in U2 were at least 6% higher than the rates for Māori U1 residents. Non-Māori exhibit the same pattern with the exception of two strata (U2 10–14 years and R1 15–29 years) where the ASH rate was estimated to be similar to that of U1 residents. For non-Māori, a strong gradient of declining ASH rates across the GCH spectrum from U2 to R3 was evident. For each age strata, the rate for R3 residents was lower than for R2, R2 lower than R1, and R1 lower than U2. For example, in those aged 45–59 years the U2 to R3 IRRs were 1.09 (U2), 0.84 (R1), 0.76 (R2) and 0.43 (R3) respectively.

Residents of U2 communities had the highest utilisation rates for all three categories of non-admitted events, both for Māori and non-Māori. In some circumstances the rate for residents of U2 was triple that for U1 residents; Allied Health service utilisation in the 75+ year-old age group was one example of this. ED utilisation is also much higher for U2 residents, particularly in the 15–29-year-old strata where the IRR for Māori is 1.73 and for non-Māori 2.03.

Residents of R3 communities had the lowest rates of specialist outpatient and ED utilisation, with disparities most apparent in the middle years of life. Examples include the ED IRR for Māori aged between 30 and 59 years of 0.64 and the specialist outpatient IRR for non-Māori aged 45–79 years of 0.59, both compared to the respective rate for U1 residents. Non-Māori in R1 and R2 communities had specialist outpatient utilisation rates that were overall slightly higher than those in U1 but lower than those in U2. On the other hand, Māori aged 30–74 years in these communities had rates that were lower than respective age-strata for U1 Māori. For example, the IRR for 60–74-year-old R1 Māori is 0.90 compared to 60–74-year-old U1 Māori. Residents of R1 and R2 communities had ED utilisation rates that were consistently higher than U1 but lower than U2. The largest differences were seen for 15–29-year-old non-Māori living in R1 and R3 and >75-year-old Māori in R2; IRRs are 1.66, 1.63 and 1.83 respectively.

The utilisation of Allied Health services by residents in R1 and R2 areas were consistently higher than in U1 communities and in some instances approximated the U2 rate. As an example, Māori aged 15–29 years in R2 had a utilisation rate three times higher than U1 (2.99; CI 2.94–3.05); in comparison, the U2:U1 IRR for Māori of the same age group was 2.62 (CI 2.60–2.64).

Discussion

This study identified considerable variation in the rates of publicly funded health service utilisation across the New Zealand rural–urban spectrum. Regional centres (U2) had, overall, the highest hospitalisation rates, and rural areas the lowest. This was most evident for all-cause hospitalisations and ASH where, for non-Māori, clear gradients of declining rates of hospitalised episodes of care with increasing rurality were apparent. ASH rates for some age strata in R3 communities were less than half the comparable U1 rates. Mental health, CVD and injury-related hospitalisation rates are also lower for residents of rural areas compared to U1 areas for the majority of strata, with cancer-related admissions proving an exception. Rates of non-admitted events were consistently higher for residents in U2 areas compared to U1 and all three rural categories. R3 communities had the lowest rates of specialist outpatient and ED utilisation. The pattern was more complex for R1 and R2 communities, where ED attendance was higher than that seen in U1, but for Māori, at least in the middle years of life, specialist outpatient attendance was lower. Allied Health service utilisation in rural areas was higher than in U1 but less than U2. No consistent pattern of variation was evident across the age strata within GCH categories.

The strengths of this study include the use of a fit-for-purpose geographic classification and the recency of the available data. Primary care utilisation data is a crucial piece of the puzzle, and its absence is a major limitation of this study. The utilisation of all the services considered in this study will be influenced by access to, and the quality of, primary care. Improved primary care data collection should be a priority for the new unified health system. It is the experience of rural health professionals that patients move between urban and rural areas, and between rural areas, in response to age and illness. Since the GCH category used was obtained from patients’ addresses at the time the healthcare event occurred, this may go some way to explaining the variation in healthcare utilisation observed.⁸ This migration and its effect on health data along with primary care utilisation will be the subject of research planned for the near future.²⁴ Possible differences in coding practice between rural and urban hospitals may also impact these rural–urban analyses.

The findings of lower rural hospitalisation rates in this study are consistent with one other New Zealand-based study that noted an association between proximity to care and higher ASH rates for children,¹⁹ but stand in contrast to wider existing New Zealand (that report similar or higher rural rates)^3,4 and international literature (that report higher rural rates). Potentially preventable hospitalisation (PPH) rates, a similar measure to ASH, are between 1.8 and 2.6 times higher in rural and remote Australia than those seen in major cities.^{20, 21} In line with our findings and older New Zealand data, rural resident ED attendance exceeds the urban rates in Australia; something that, along with the higher PPH rates has been attributed to poor access to acute primary care for rural Australians.²² Canadian rural hospitalisation and ED rates mirror those seen in Australia.²³ In addition, rural Canadians have lower rates of specialist outpatient attendance.

Considerable care needs to be taken when interpreting these results from a policy perspective. For example, it should not be assumed that the lower rural ASH rates are indicative of access to quality primary and preventive care or healthier rural communities. These rural communities have New Zealand’s highest amenable mortality rates.⁸ Low ASH rates in this context are more likely to reflect a complex interaction of need, rural models of healthcare delivery and access, and may in part be a consequence of the widespread closure of rural hospital beds that occurred during previous health reforms.²⁵ Equally, the differences in access to Allied Health and specialist outpatient services need further in-depth research in order to understand the causes of the differences, and their implications for policy and service delivery. The high rural:urban mortality rate ratios for the younger age strata⁸ were not matched with higher rates of health service utilisation in this study. This is unexpected and suggests that hospitalisation rates may not be reliable indicators of morbidity in the New Zealand rural context. Other health systems factors that differ between rural and urban areas may be impacting hospitalisation rates. Examples include the structure of the workforce, with a high proportion of locums and international medical graduates in rural areas, and the availability and uptake of private healthcare.^26,27

Until recently, many rural communities shared a DHB with their nearest regional city (U2). The magnitude of the disparities identified in this study are at their greatest when U2 and rural communities are compared. This suggests that greater attention could have been paid to monitoring rural–urban variation within DHBs, rather than focussing on differences between DHBs. Variation in the utilisation of health services between neighbouring rural and urban communities may be larger than the variation between DHBs, and as such a greater example of “postcode lottery”.

The R3 category, which covers 39% of New Zealand’s land area but only 1% of the total population, is home to some of our most vulnerable communities. It has the highest proportion of Māori (33%), and the highest proportion of residents living in the most deprived New Zealand Index of Deprivation (NZDep) quintile (Māori 73%, non-Māori 39%).^28,9 There is evidence that the Māori:non-Māori health outcome “equity gap” is greater in rural areas.⁹ An association between rurality and higher amenable mortality rates (an effect more pronounced in younger age strata, for Māori and for more remote communities) has previously been demonstrated.⁸ This study adds evidence of lower levels of actualised access to secondary care, either as inpatients or specialist outpatient clinics, for the same populations, and in doing so also raises questions about geographic equity in health service design and delivery.

This study has demonstrated large, and previously unrecognised, rural–urban differences in public health service utilisation in New Zealand. These differences are in marked contrast to those seen in comparable countries and warrant further exploration. New Zealand’s new unitary healthcare system and rural health strategy has created an opportunity to address any health disadvantage for rural communities that may be occurring as a result of these differences in health service utilisation.

View Appendix.