Unmet need for primary health care and subsequent inpatient hospitalisation in Aotearoa New Zealand. A cohort study

Introduction

Primary health care (PHC), as declared by the World Health Organization (WHO), ‘… is the first level of contact of individuals, the family and community with the national health system bringing health care as close as possible to where people live and work, and constitutes the first element of a continuing health care process.’¹

Thus, PHC has two roles: to provide health care, and to act as guidance to secondary and tertiary level health care and to other health and social care. Through the former role, PHC can improve the health of patients so that hospital-level care is not needed, while the latter role channels patients towards the most suitable care pathways, possibly pre-empting unnecessary utilisation of hospital-level care. Reducing hospital-level care is important, as it holds the potential to lower overall health care expenditure.²

Primary care has been found to improve health as well as health equity.³ However, people who experience barriers to primary care may miss out on these health benefits. In turn, unmet need for primary care may affect how patients access secondary and tertiary level care. Waiting to afford a GP visit may mean it takes people longer to get a referral from their GP to access hospital care, or they may access it more quickly by bypassing primary care and self-referring to the emergency department at a hospital. In the 2022/23 New Zealand Health Survey (NZHS), it was found that 13% of the population aged 15 and over did not have a GP consultation when needed in the previous year because of cost, equating to 541,000 people.⁴

The New Zealand government subsidises the cost of a GP consultation with patients paying the remaining cost. There are various programs that can reduce these costs to patients: the Very-Low-Cost-Access scheme works at the level of the general practice providing lower costs for all enrollees, and the Community Services Card and the High Use Health Card schemes act at the patient level. These programs have the objective of delivering more economical health care to people facing heightened health needs and/or with limited financial means. Additionally, maternity care is free in the public health system.

This paper looks at the characteristics and primary care use of people with unmet need for a GP consultation because of cost, the time it takes them to access inpatient hospitalisations and the characteristics of such hospitalisations.

Methods

The data for this study were accessed via the Datalab at Statistics New Zealand. For a fuller account see the Supplementary material accompanying this paper.

Three datasets were combined via the encrypted National Health Index. These were six NZHSs with data collected between 2013 and 2019 which provided self-reported information about demographics, health care and health status of respondents. The Public Hospital Discharges dataset which records all inpatient discharges from public hospitals. It contains information on the start and end dates of the hospital stay and the diagnoses related to the event, coded in ICD-10-AM.⁵ The Mortality dataset which gives information about deaths.

The NZHS dataset contributed the primary variable of interest. Respondents were asked if during the previous year they had needed to see a GP but did not because of cost. Respondents were put into two groups, called ‘need’ and ‘no-need’, depending on whether they answered ‘yes’ or ‘no’ to this question.

Results

The number of respondents in the six surveys who appear in the Datalab datasets ranged from 10,932 in the 2014/15 survey (out of 13,497 respondents in the total survey or 81%) to 12,579 in the 2018/19 survey (out of 13,572, 93%). Altogether, there were 72,243 respondents in the six surveys. Some respondents were in more than one survey and for them one random observation was chosen to be kept (516, 0.7%, observations deleted). A further 225 people died before follow-up started (0.3% deleted). This left 71,502 respondents for analysis (see Supplementary Fig. S1).

The demographic and health profile of the ‘need’ and ‘no-need’ groups appears in Table 1. We observed notable demographic differences between individuals classified as being in the need group and those in the no-need group. Specifically, a higher proportion of females were found in the need group, constituting 70% of this group, as opposed to 55% in the no-need group. Furthermore, younger adults, (25–34-year-olds and 35–44-year-olds) were most likely to be in the need group while those in the oldest age groups (65–74-year-olds and 75+ year-olds) were most likely to be in the no-need group. Māori make up 30% of the need group compared to 19% of the no-need group; conversely, European New Zealanders/Others make up 57% of the need group and 67% of the no-need group. Those in the need group were more likely to be in the most deprived quintile of NZDep (36%) compared to the no-need group (25%); conversely, 33% of the no-need group were in the top two least deprived quintiles compared to 21% of the need group.

Respondents in the need group were more likely to be in the bottom two categories of self-rated health (26% versus 11%) and less likely to be in the top category (7% vs 14%). They were more likely to have been told by a doctor that they had depression (34% vs 16%), asthma (30% vs 19%) or an anxiety disorder (22% vs 9%).

Respondents were asked if the last visit they had made to an emergency department at a hospital could have been treated at a general practice. For those in the need group, 35% agreed with that statement compared to 24% for those in the no-need group.

Table 2 contains hospitalisation characteristics of the two groups. Compared to the no-need groups, those in need were more likely to be hospitalised (52% vs 44%), had a similar number of hospitalisations (3.4 visits vs 3.1 visits), a shorter length of stay (2.7 days vs 3.3 days) and had a shorter time to hospitalisation (608 days vs 651 days).

Table 3 presents the hazard rates for three proportional hazard survival models for time till hospitalisation. The first model contains the primary variable of interest, need status. It shows that the hazard rate is 28% higher for the need group over the no-need group. The hazard rate for the need group increased from 1.28 in the first model to 1.39 in the second model. Based on a confounder analysis, age group is the variable that accounts for this increase: the need group is younger than the no-need group and this means the unadjusted model does not show the full extent of the consequences of need.

The third model adds in the self-rated health variable, which gives a hazard ratio for the need group of 1.25. The decrease in the hazard rate for the need group from model 2 reflects that the need group is more likely to have people with lower health status.

The confounding variables act as would be expected, with those who are older, Māori or Pacific peoples, with greater socio-economic deprivation or with poorer health status having a hospital admission earlier.

On investigating the model it was seen that one particular age group behaved differently in terms of their time to hospitalisation, based on a graph of Kaplan–Meier survival curves (see Supplementary Fig. S3). The 25–34-year age group was more likely to be hospitalised earlier and were less likely to be hospitalised later relative to other age groups. Model 3 was refitted with this age group removed and the hazard rate for the difference between the need and no need groups reduced by 1.8%. This small change gives us confidence that the result is robust to this failure in the proportional hazards assumption.

Supplementary Table S1 groups the ICD codes for primary and secondary diagnoses into their respective chapters by need group for the first hospitalisation during the study period. The chapter showing the biggest difference between groups was for ICD codes involving ‘pregnancy, childbirth and the puerperium’ for the primary diagnosis (17.2% for the need group vs 9.7% for the no-need group). As pregnancy is not an avoidable condition for hospitalisation, we removed people with a pregnancy-related diagnosis and re-ran model 3. The odds ratio of the need group increased slightly by 1.5% indicating that the condition is not causing the difference between the two groups.

The biggest difference for secondary diagnoses was ‘Factors influencing health status and contact with health services’ (59.1% vs 48.4% respectively) where respondents in the need group were more likely to be current tobacco users (‘Z720 Tobacco use, current’, 26% versus 12% respectively) while the no-need group was slightly more likely to have given up tobacco products (‘Z8643 Personal history of tobacco use disorder’, 19% versus 23%).

Discussion

Those who have an unmet need for a GP consultation were more likely to be young, female, Māori or Pacific peoples, more socioeconomically deprived and in poorer health than the no-need group. While previous research has shown that consultation fees were lowest where the population surrounding a general practice was more likely to be composed of Māori, Pacific peoples or the more deprived, it appears that these lower fees are not low enough to eliminate unmet need of a GP consultation due to cost.⁷ It could be that it is not just the individual cost of a consultation that means an individual has unmet need but that an individual may require more consultations and the cumulative cost is an important factor for having unmet need. From the NZHS 2022/23, the average number of GP consultations per year overall was 2.4 visits; it was higher for females (2.8, 95% CI 2.3–2.5), the most deprived quintile (2.6, 95% CI 2.4–2.9) and the disabled (4.9, 95% CI 4.4–5.5).⁸

There may also be multiple instances of unmet need, something not asked about in the NZHS but asked in the Growing Up in NZ study. In that study, caregivers of children of 2 years of age were asked the number of times they had an unmet need for a GP consultation in the prior year. It was found that 44% of the children with unmet need had multiple unmet needs in that time, with 4% reporting more than five instances of unmet need.⁹

The cost of the GP consultation is not the only cost that may impose a financial barrier to care. From the NZHS 2022/23, other financial barriers felt by respondents included the cost of prescription items (4.0% of respondents), as well as transport costs (2.2%), time off work (7.4%) and owing the general practice money (1.4%).¹⁰ It is worth noting that the greatest barrier was due to the wait time for a GP appointment (21.2%).

In general, the need group was more likely to be hospitalised and to have had their first hospitalisation sooner. However, they had shorter stays in hospital on average. Given the diagnoses were generally similar between groups, this could be due to being younger so that their recuperation time is shorter, or it could be that their higher probability of having a pregnancy-related diagnosis produced a shorter stay.

The need group were more likely to have a secondary diagnosis of being current smokers and it could be that smokers in the need group are having to make choices about buying tobacco or seeing their GP. The previous Labour government had put into law a ban on buying cigarettes for those born after 2008 which was to come into force in 2024.¹¹ This would seem an advantageous step since those with unmet need tend to be young and so it would have affected this group first. Subsequently, the National government repealed this legislation.¹²

Studies have looked at which characteristics of PHC providers reduce the need for hospital-level care,¹³ as well as interventions to reduce that need.^14–16 Two studies that put interventions in PHC for people identified at risk of avoidable hospital care found that there was little to no reduction in hospital level care and an increased cost or workload due to identifying people with unmet need.^14,15

We cannot tell from our analysis whether the hospitalisations for the need group could have been avoided if their unmet need for a GP consultation had been met. However, the need group in this analysis had a higher probability (35%) of using the emergency department for a condition that could have treated in PHC compared to the no-need group (24%). This suggests that some in the need group appear to be using emergency departments as an alternative to a GP consultation rather than as an alternative pathway to accessing hospital-level care.

One limitation is that we do not have the interview date of the respondents in the NZHS, which means we had to choose when to start follow-up. In this case, we chose to start follow-up at the end of the survey period for the survey the respondent was in. This meant that some of the respondents could have had their first hospitalisation before follow-up started.

Another potential date for the start of the follow-up period could have been the first day in the survey period that the respondent was in. A sensitivity analysis was done where the start of follow-up was set as the first day of the survey period the respondent was in. For model 3, the hazard rate for the need group decreased by 0.02%. This shows that the hazard rate for the need group is robust to when follow-up started.

The implication from this research, and from previous research on fees,⁸ is that work needs to be done at a policy level so that the out-of-pocket costs to patients are better aligned with their means, especially for people who need on-going care at a general practice. Furthermore, work needs to continue on reducing tobacco use, not only to improve health but with the prospect of reducing unmet need for health care.  

Supplementary material

Supplementary material is available online.