How is the user base of general practices associated with Open or Closed Books in Aotearoa New Zealand? An analysis of administrative data

Introduction

New Zealand’s (NZ’s) Primary Health Care Strategy places a strong emphasis on promoting health equity, making it a central goal.¹ The more recent NZ Health Strategy has the aim of ensuring that all are able to access healthcare services when they need it,² implying that access should not depend on people’s health conditions, socioeconomic position, or location.

As a first step, patients can access primary health care at a general practice and, if they choose to enrol, they can access extra services along with obtaining lower out-of-pocket fees. The general practice benefits financially from enrolling patients as they receive government funding based on the number and demographic profile of their enrolees.³

General practices can choose not to enrol any new patients at all, and this is known as ‘Closed Books’. This choice can be seen as a balancing act between financial imperatives and providing an acceptable level of health care.⁴ Closed Books sits alongside other constraints on services, eg wait times for consultations are longer, on average, for all patients at that practice.⁵ In the NZ Health Survey 2023/24, 26% of respondents reported that the wait time to get an appointment to see a GP was a barrier to accessing primary health care.⁶

Geographic disparities, often called the ‘postcode lottery’, are another aspect of healthcare inequity.^7–10 In NZ, this shows up as differences in accessibility to health care due to travel distances to health services and is attenuated by local socioeconomic conditions.

Previous work, based on data from 2022, found that there was no significant association between the proportion of Closed Books practices in a Health District (formerly District Health Board area) and (1) the number of GPs per head of population, (2) the number of consults per enrolees per year, (3) population changes, and (4) the proportion of people enrolled in a district.¹¹ However, there was found to be an association at the general practice level with consultation fees, with general practices with fees just above the middle of the fees distribution being more likely to have Closed Books.¹¹

This paper aims to identify the area where a general practice is located, examine the characteristics of the resident population in that area, ie the user base of the practice, and investigate how these characteristics are associated with general practices’ enrolling status.

Methods

This study looks at information available in areas defined by Statistics New Zealand as statistical area 2 (SA2). These areas were created by Statistics NZ ‘to reflect communities that interact together socially and economically’.¹² In this research, the SA2s for 2023 were used to align with output from the 2023 census.

Data and variables

Closed Books database

The Closed Books database contains the enrolling status of general practices across NZ. A general practice’s enrolling status was retrieved from the Healthpoint website from 5 to 11 September 2024 with the written permission of Healthpoint.¹³ Enrolling status was coded as ‘Yes’ or ‘No’, ie Open or Closed Books respectively, with unknown status coded to ‘Yes’ (n = 3, 0.3%). The types of general practices considered were healthcare facilities defined by Health New Zealand as ‘Enrolling GP Practices’ and were further restricted to have at least one non-virtual GP, were open to the general public, and were in a physical location, eg they did not include aged care practices or telehealth only practices. This database also included the address and/or longitude and latitude of the general practice. A description of how this database was assembled is available in Pledger et al.¹⁴

Geography Boundary Viewer

The Geography Boundary Viewer is a Statistics NZ online Graphical Information System that presents boundaries of different geographic units as well as the Urban Accessibility Index (UAI).¹⁵ The UAI is based on the travel time by road to a major, large, or medium urban area and is coded into eight categories of urbanisation, urban accessibility, or remoteness. An example of the UAI appears in Fig. 1.¹⁵

Fig. 1.

The Urban Accessibility Index for the Greater Auckland, Waikato, and Bay of Plenty regions.¹⁵ This figure is based on Stats NZ’s data, which are licensed by Stats NZ for reuse under the Creative Commons Attribution 4.0 International licence.

NZDep2023 statistics area 2 (SA2) database

The NZDep2023 SA2 database contains NZDep scores and indices based on data from the 2023 NZ census at the SA2 level.¹⁶ NZDep is an area-based measure of socioeconomic deprivation based on responses to selected census questions about households and housing taken from respondents living in an area and assigned to individuals according to their residential addresses. For modelling, NZDep was put into quintiles, with one representing the lowest socioeconomic deprivation areas and five the highest.

Total by topic for individuals (by …, SA2, …) for the 2023 Census database

The 2023 Census database contains confidentialised count data of the usual resident population and medians of variables from the 2023 census based on SA2 areas.¹⁷ The main variables selected were the sociodemographic variables, sex, age, and ethnicity. Personal income was included as a personal indicator of deprivation, as a counterpoint to NZDep, which is based on household measures of deprivation. Of the seven questions asked about health difficulties and health conditions, a summary question, reflecting health status, was selected. Another variable included was whether an individual was born in NZ. While this is associated with ethnicity, it also reflects a geographic context. As levels of urbanisation and accessibility decrease, the likelihood of residents being NZ-born increases. Although some may choose to live in these areas, those with limited social or economic capital may lack the means to relocate to areas with greater accessibility to services and resources, becoming stuck in place.

The variables were modified to reflect that data came from SA2s with different population sizes: (1) sex – the proportion of males, females, and those stating another gender were calculated out of the people stating a sexual identity; (2) Median age was kept as recorded, (3) ethnicity – total count ethnicity for each of Europeans, Māori, Pacific peoples, and Asian were used and transformed to a proportion of those who stated an ethnicity, (4) Born in NZ was the proportion of those born in NZ out of the total classified, (5) Median personal income was kept as recorded, and (6) Heath status was the proportion of people aged 5+ who agreed with the census question ‘Do you have a disability, long-term condition, or mental health condition that limits your ability to carry out everyday activities?’ out of those classified.

Some variables had missing data: 0.9% of responses could not be classified for the variable Born in NZ and 15.1% for Health status. Our current construction of the Health status variable inherently assumes the respondents who were classified would answer similarly to those not classified. An alternative construction would be to take the proportion as the number of people who agreed out of all respondents aged 5+. This latter construction assumes that people not classified would have disagreed with this question. This is a common occurrence when people skip ‘No’ answers when a set of questions aren’t relevant to them.

Results

Each SA2 containing a general practice had its statistics calculated for each of the variables. Each SA2 was then placed in a quintile according to where its statistic fell in the distribution for each variable. Table 1 shows the boundaries for each variable eg for Quintile 1, for the male variable, the proportion of males ranges from 0.406 to 0.476. This means that the 20% of SA2s connected with a general practice with the lowest percentage of males have between 41% and 48% males in them. Similarly, the 20% of SA2s connected with a general practice with the highest percentage of people with a low health status contain between 11% and 28% people with a low health status in them.

Table 2 presents the regression coefficient for the trends across quintiles (multiplied by 100 to be on the percentage scale) according to the absolute value of the strength of the trend.

The variable with the largest trend across quintiles is ‘Born in NZ’. The trend is 6.5, indicating that for a change from one quintile to the next there is an estimated increase of 6.5 percentage points for a general practice having Closed Books. Across the quintile range, that equates to an estimated increase of 26 percentage points. Fig. 2d presents a histogram of Closed Book’s percentages at each quintile. It shows that when the probability of NZ born individuals is low then the probability of having Closed Books is low and that the increase is seen over the first three quintiles.

Fig. 2.

Percentage of Closed Books by quintiles of the (a) proportion of sexes, (b) median age, (c) proportion of total count ethnicities, and (d) proportion of NZ born in the SA2 of the general practice.

The second strongest variable is the Health status variable with a trend value of 4.7. Fig. 3c shows an increasing trend across quintiles with a larger increase seen between the fourth and fifth quintile. The alternative Health status variable had a trend value of 5.2, a reasonably similar value, which suggests that how the missing data were treated did not meaningfully impact the interpretation

Fig. 3.

Percentage of Closed Books by quintiles of (a) Median personal income, (b) the New Zealand Deprivation Index, (c) Health status, and (d) the Urban Accessibility Index in the SA2 of the general practice.

The next strongest variables are the ethnicity variables, with Māori and European having positive trends across quintiles and Asian having a negative trend. This means that as the proportion of people who are Māori and European increases the probability of being a Closed Books practice increases, and the converse for Asian peoples.

For males, there is a negative trend across quintiles indicating that as the proportion of males increases the probability of a general practice having Closed Books decreases (see Fig. 2a), and the converse is shown for females.

There appears to be no significant trend across quintiles between a general practice having Open or Closed Books and median age quintiles, median personal income quintiles, and NZDep quintiles (see Figs 2b, 3a, b).

Closed Books were more likely in large urban areas (56%, 95% CI 48–63) and medium urban areas (54%, 45–64) compared to other areas (30%, 26–33) (see Fig 3d). The statistics for large and medium urban areas showed little difference between the North and South Islands. However, in the South Island, the proportion of general practices with Closed Books in major urban areas (50%, 41–60) was higher than in the North Island (24%, 19–28), and conversely was higher in remote and very remote areas in the North Island (45%, 31–59) compared to the South Island (10%, 0–21).

For the univariate models listed in Table 2, and where a variable had been found significant, Health District was added as a confounder in the model. After this adjustment, only the variable Born in NZ remained significant.

Discussion

We found several ways in which the location of general practices and the characteristics of their user base were associated with the enrolling status of the practice. General practices were more likely to have Closed Books when the user base had a higher proportion of people born in NZ, had lower health status, were Māori or European, female, or living in large or medium urban areas. It was less likely when the user base had a higher proportion of males or Asian peoples.

Quintiles of Personal income, NZDep, and median age were not associated with Closed Books general practices. This lack of association is surprising for two reasons. Firstly, research in the Waikato district found strong correlations between primary healthcare need and both NZDep and age.¹⁸ This is different to our analysis where the variables NZDep and median age were not significant but the Health status variable, also indicating health need, was highly significant. Secondly, practice fees had previously been associated with Closed Books, whereas here, measures of income and deprivation were not.¹¹

Previous work has shown that Auckland, Counties-Manukau, and Waitemata districts were the least likely to have Closed Books.¹⁴ These areas are highly attractive to international migrants due to the opportunities and services available in NZ’s largest urban area and to internal migrants, especially young people, who move to major urban areas for educational, occupational, and social reasons. The ‘healthy migrant effect’ may mean there is less demand in Auckland for health services, and this is possibly reflected in Asian peoples, males, and young people being less likely to be enrolled in primary health care.¹⁹ In terms of internal migration, the ‘healthy migrant effect’ may leave the provincial cities with higher proportions of people with lower health status, putting strain on general practices in those locations.

The characteristics of the user base that are associated with Open and Closed Books appear to be entwined with location. This is reflected by only one variable, Born in NZ, remaining significant after accounting for Health District. Even so, the strength of this variable has been reduced by over 55%, indicating the importance of the district variable in mediating this effect. This finding seems most likely explained by patterns of settlement.

The fact that we found characteristics of the user base to be associated with the enrolling status of general practices suggests that we are not fulfilling the aims of the NZ Health Strategies. However, the confounding effect of Health Districts suggests that tailoring interventions to geographic contexts could be key to reducing these disparities and achieving a more equitable healthcare system.

A limitation of this analysis is that there are missing values in the Census data that affect our results directly, eg in the Health status variable, and indirectly, as Statistics NZ uses data from elsewhere, where available, to fill in the data for missing respondents.²⁰ Missingness will also affect NZDep, which is based on Census data.

The enrolment data is sourced from the Healthpoint website, and its accuracy depends on general practices updating their profiles promptly.^13,21

We have chosen SA2 as the area that reflects the user base of general practices. However, general practices may have a wider user base than these areas. This may not matter in urban areas where SA2s are likely to be spatially correlated, but may matter in provincial areas where the user base may include urban and rural dwellers whose characteristics are not captured in the SA2 of the general practice. While some patients may go to general practices a distance from their home, this is a choice that is not always possible, as general practices nearing capacity often limit their new enrolees to the local population.⁵

In conclusion, this study highlights that the characteristics of the user base are associated with a general practice’s enrolling status and are mediated by location. The findings contribute to a deeper understanding of the inequalities impacting access to primary healthcare and points to the potential of geographically tailored approaches to minimise disparities and promote healthcare equity.