Quantifying unmet secondary healthcare need in New Zealand: a multi-year population study using administrative data

Abstract

Objectives

Examine patterns in declined referrals to secondary healthcare in New Zealand by sociodemographics, region, health specialty and over time, as an important marker of potential unmet healthcare need (UMN) for specialist care. The primary hypothesis was that UMN varies by sociodemographic groups, region and health specialty and has increased over time.

Design

A repeated cross-sectional analysis using administrative data from the National Patient Flow (NPF) Collection (2018–2022).

Setting

Nationwide, encompassing all first specialist assessments (FSA) referrals to public hospital specialists in New Zealand.

Participants

Individuals referred by general practitioners for FSA.

Outcome measure

The primary outcome was FSA referrals being declined at prioritisation.

Results

Among 2 918 557 first referrals for FSA, the observed rate of declined at prioritisation was 13.1%. Among those referred, females had a significantly higher risk of being declined (relative risk (RR), 1.069; 95% CI, 1.062 to 1.075), while those in younger (0–9 years: RR, 0.853; 95% CI, 0.841 to 0.865 and 10–19 years: RR, 0.891; 95% CI, 0.879 to 0.904) and older (80+years: RR, 0.955; 95% CI, 0.944 to 0.967) age groups as well as Māori (RR, 0.817; 95% CI, 0.810 to 0.824) and Pacific peoples (RR, 0.706; 95% CI, 0.695 to 0.716) had a significantly lower risk. There was also significant variation in risk of being declined by region and health specialty. The overall risk of being declined increased by 4.1% annually (RR, 1.041; 95% CI, 1.039 to 1.044). Significant increases in risk of declined over time were also observed across all sociodemographic groups, with higher risks for non-Māori/non-Pacific individuals (RR, 1.045; 95% CI, 1.043 to 1.048) and those in less deprived areas (RR, 1.057; 95% CI, 1.052 to 1.063).

Conclusions

UMN in New Zealand has significantly increased, exacerbating health inequities and straining primary care. Policy interventions are urgently needed to address these disparities, particularly in high-risk specialties and populations. This method of quantifying an important marker of UMN may inform global health equity initiatives.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• Utilised a comprehensive and whole-of-population administrative dataset (National Patient Flow Collection) covering 5 years (2018–2022), allowing robust analysis of unmet secondary healthcare needs at the population level.

• Employed a repeated cross-sectional design, enabling detailed examination of temporal trends and sociodemographic differences in access to specialist care.

• Data quality issues in certain regions and the exclusion of one region due to reporting inconsistencies mean findings should be interpreted with caution and may limit the generalisability of the findings.

• Relied on declined referrals as a proxy for unmet healthcare need, which may not fully capture other barriers to access, such as patients deterred from seeking referrals due to systemic or financial constraints.

Introduction

Unmet need (UMN) refers to the situation where people who need healthcare services do not receive these services or experience care access barriers, meaning their needs are not met.1,3 UMN is a growing challenge in many health systems. UMN can have negative consequences for individuals and society, such as poor health outcomes, high personal spending or downstream health system costs, productivity loss and health inequalities in access, processes and outcomes.⁴ A common example of UMN occurs when a general practitioner (GP) has assessed a patient for a condition that requires specialist medical attention, referring them onto a secondary care professional. The patient may then face barriers to secondary care such as financial barriers or being rejected for treatment as the healthcare system does not have the capacity to provide care. This may be because resources such as funding or workforce capacity are not sufficient to meet demand. In publicly funded health systems, such as the UK NHS, Canada and New Zealand, this is common.5,7 This means that the patient must seek ongoing care with their GP. Measuring UMN is essential for tracking the performance and progress of health systems, which is one of the United Nations Sustainable Development Goals targets.² Of course, UMN also occurs when patients cannot access even basic primary care or other needed services.

There is a well-documented history in the New Zealand public healthcare system of not measuring UMN despite widespread acknowledgement of its existence.⁸ The New Zealand health system is in a UK Commonwealth tradition, with taxpayer funding, free public hospitals and private GPs in receipt of government subsidies. Government-funded services dominate and, at the hospital level, function as a national network across a series of administrative regions. GPs act as gatekeepers, meaning patients must be referred by a GP for any specialist, usually hospital-based, services. Data indicate that around 80% of patient issues are dealt with initially in the GP setting, despite receiving only around 5.5% of government health expenditure.⁹

New Zealand’s GPs are highly organised by global comparison and work through a variety of professionally governed and government supported networks which support multi-disciplinary team work, health promotion and population health planning.¹⁰ Sole or group practice outside of a network is very unusual. Over the years, GPs and specialists have worked on a range of initiatives, with government support, to improve referrals. Specialists use scoring systems to prioritise patients based on need, while GPs and specialists routinely use ‘health pathways’, which are regularly updated, to ensure that referrals follow agreed protocols.¹¹ New Zealand has also placed considerable emphasis on joint ‘whole system’ planning to provide better support for primary care and GP services and release pressure on the hospital system.¹²

Around 80% of New Zealand health funding is public, with the remainder from voluntary private insurance and direct out-of-pocket service payments for GP consultations, pharmacy co-payments and private specialist and other private services. Private hospitals and specialists only provide elective procedures. Over half of public hospital specialists also work in private practice. Public hospital waiting lists for specialist assessments and procedures are ubiquitous, meaning those with means to pay often opt for private services.

Measuring UMN in New Zealand is important as the less well-off, and Māori and Pacific populations in particular, are most affected.¹³ These groups are more likely to be reliant on the public health system, which is generally less equipped than the private sector to meet healthcare needs. It is widely recognised that New Zealanders with private insurance or the financial means to do so can access services through the private system.¹⁴ Often, this choice is made in anticipation of long wait times for public treatment or the possibility of being declined due to low clinical priority. The decision to seek private care typically occurs at the point of referral and is made in consultation with a primary care provider. This dynamic reflects an unintended consequence of the institutional structure of New Zealand’s health system, rooted in the Social Security Act 1938 and shaped by subsequent policy compromises.¹⁵

Defining and measuring UMN is complex, as it is influenced by a range of factors including the availability, affordability and quality of health services, as well as broader social determinants of health, individual values, expectations, preferences and levels of health literacy. A variety of data sources and methods, such as surveys, clinical guidelines, expert opinions and administrative records can be used to assess UMN. Each approach has its own strengths and limitations and may reflect different dimensions of UMN.⁴

A commonly used approach to measuring UMN involves survey questions that ask individuals whether they were unable to access medical or dental care within a specified timeframe (eg, the past 12 months) due to factors such as cost, travel distance or waiting times. These surveys provide insights into self-reported barriers to healthcare access and help estimate the prevalence of UMN across different population groups. Such questions are included in the New Zealand Health Survey and international surveys such as those conducted by the Commonwealth Fund.¹⁶ However, this method does not capture other important dimensions of UMN, such as how individuals with UMN are managed, issues related to care quality, the burden on services including general practice, lack of awareness or diagnosis or dissatisfaction with care that is received or remains inaccessible.¹⁷

Survey data from 27 OECD countries with comparable measures show that, on average, only 2.6% of people reported experiencing UMN for medical care in 2019. However, rates varied significantly across countries from just 0.2% in Spain to 15.4% in Estonia.¹⁸ UMN was generally more common for dental care than for medical services, reflecting more limited public coverage and higher out-of-pocket costs for dental care in many countries. Rates of UMN were also higher among individuals with lower income, lower educational attainment or poorer health, highlighting persistent socio-economic disparities in access to healthcare.¹⁸

The COVID-19 pandemic significantly worsened the issue of UMN in many countries, as health systems struggled to maintain essential services while responding to the public health crisis.¹⁹ Factors such as fear of infection, lockdowns, financial pressures and reduced service availability all contributed to people delaying or avoiding necessary care. On average, across 23 OECD countries with comparable data, over 20% of individuals reported forgoing a needed medical examination or treatment during the first year of the pandemic.²⁰

In New Zealand, the setting of the present study, the New Zealand Health Survey collects self-reported data and routinely highlights cost as one of the key barriers to accessing GP services. In the 2022/23 survey, 12.9% of the general population reported cost as a barrier to care. This figure was notably higher among disabled adults (21.4%) compared with non-disabled individuals (12.0%), and among women (15.1%) compared with men (10.5%). Māori and Pacific peoples also reported elevated rates of cost-related barriers at 16.9% and 17.6%, respectively. Dental care emerged as a particular area of concern, with 44.0% of the general population, 55.9% of Māori, 58.0% of Pacific peoples and 52.4% of disabled individuals reporting UMN.²¹

Multiple strategies can be employed to address UMN and improve equitable access to high quality healthcare. These require policymakers to adopt a comprehensive, multi-faceted approach that considers both demand- and supply-side factors. Potential strategies include expanding health coverage and reducing financial barriers; improving the availability and distribution of health professionals and facilities; enhancing the quality and safety of services; promoting health literacy and public awareness; streamlining referral and care coordination systems; and strengthening data collection and monitoring systems. Central to these efforts is the need for detailed, reliable information about where access barriers exist, where system bottlenecks occur and where targeted investments will have the greatest impact.¹⁷

There is a need for more in-depth understanding of multiple aspects of UMN. This is outlined in a 2023 paper urging a WHO resolution for member countries to commit to measuring UMN and do so using consistent and comparable approaches.⁴ However, there is a notable lack of published quantitative studies focused on UMN for secondary healthcare. This study contributes to addressing that gap by identifying and applying a practical method for capturing and monitoring an aspect of UMN using national level routinely collected health data. Specifically, it aims to use routine data from the New Zealand public healthcare system to provide a national overview of rates of patients referred by GPs for first specialist assessments (FSA) that were declined, including by sociodemographic sub-group, region, health specialty and over time.

Methods

Study design and setting

The study employed a repeated cross-sectional design, examining all individuals in New Zealand referred for an FSA over the 5 year period from 2018 to 2022 using the National Patient Flow (NPF) Collection.

The National Patient Flow (NPF) Collection

The NPF is a routinely collected nationwide dataset that contains information of key stages in a patient’s journey from primary through secondary and tertiary healthcare.²² It includes patient referrals to specialised services and holds information regarding referral service engagements, detailing a holistic perspective of the patient’s care trajectory. This includes information pertaining to referrals for an FSA from primary care and the outcomes of these referrals. The presenting referral is the first time a patient with a presenting problem is referred for specialist care. The NPF is the only national data source in New Zealand that captures clinician-determined prioritisation outcomes for FSA referrals. As such, it offers a valuable mechanism for routinely monitoring patterns of demand and unmet need within publicly funded specialist services.

Figure 1 illustrates the core business concepts within the NPF and their relationship to key points in the patient referral process. It provides a high-level representation of how referral events are captured. Readers should note that this figure is a simplified subset of the broader business concept model and does not encompass all possible clinical scenarios or referral journeys. For a complete description of data structures and patient flow logic, readers are encouraged to consult the full NPF Phase 3 File Specification.¹⁸

Figure 1. NPF core business concepts. Source: Health NZ.

The NPF began its phased data collection in New Zealand in 2014. The present study drew on data from Phase 3 of the NPF, which commenced on 1 July 2017. These data are collected continuously as part of routine health service delivery through standardised electronic health information systems. These systems operate in accordance with the National Patient Flow Phase 3 File Specification v3.6, which defines the required fields, formats and coding standards for consistent nationwide reporting.²²

This study was conducted in accordance with established best-practice guidelines for register-based epidemiology, including those outlined by Sund and Gissler.²³ We worked closely with Health New Zealand (Health NZ, the national health service commissioning agency also known as Te Whatu Ora) to understand the structure, completeness and limitations of the NPF and used this knowledge to inform a study design tailored to the available data. The event-level NPF data used in this research were provided by Health NZ under strict access conditions with formal approval. Data access complied with the requirements of the New Zealand Privacy Act 2020 and the Health Information Privacy Code 2020. Following approval, the data were securely transferred to the University of Otago via Health NZ’s secure File Transfer Protocol (FTP) service on the 17th of January 2024. Access to the event-level dataset was restricted to the lead author and data analyst (Dr Bowden), who was responsible for all data processing and analysis. The handling of the data followed agreed governance protocols, including secure storage and access controls, and was limited to the purposes outlined in the original data access request. Transparent and standardised definitions were applied in line with the NPF Phase 3 File Specification.²²

First referrals

The focus of the present study is first (presenting) referrals for an FSA. This is the first time a patient with a presenting problem is referred for specialist care. FSAs are triaged and a prioritisation outcome is determined by a clinician. This study included all first referrals for an FSA and the associated prioritisation outcome, drawn from the NPF in the 5 year period from 2018 to 2022. The four prioritisation outcomes are: accepted, transferred, not decided and declined. To simplify the analysis and align with the study’s primary focus, rates of declined among referrals for an FSA, we collapsed the original prioritisation outcome variable into a binary indicator (declined vs not declined). This approach also reflected the low frequency of the ‘not decided’ and ‘transferred’ categories (each approximately 1%), which were not analytically informative and would have introduced unnecessary complexity.

Sociodemographic measures

Patient-level sociodemographic information was drawn from the NPF. This included sex (male/female), age in years (categorised as 0–9, 10–19, 20–29, 30–39, 40–49, 50–59, 60–69, 70–79, 80+years) and prioritised ethnicity using Level 1 groups, categorised as Māori, Pacific and non-Māori/non-Pacific (NMNP). Meshblocks associated with domicile information were used to link to area level deprivation measures using the New Zealand Index of Deprivation 2018 (NZDep). NZDep scores were collapsed into quintiles (1 reflecting areas of the lowest levels of deprivation and 5 the highest). The region the individual resided in at the time of the referral was also extracted. Lastly, the health specialty for the FSA was also extracted.

Statistical analysis

Counts of first referrals for FSA were presented by year, sociodemographic sub-groups, region and the twelve highest volume health specialties. Among those referred for an FSA, annual rates of declined were tabulated using prioritisation outcomes overall, by sociodemographic sub-group, region and health specialty.

To analyse the association between being declined at prioritisation and sociodemographic factors and to assess whether declined rates changed over time, a complete case generalised linear regression model with a log link and binomial distribution and robust error variance correction was estimated. The binary dependent variable was whether a referral for FSA was declined at prioritisation. Independent variables in adjusted models included year, sex, age, ethnicity, deprivation and region. To assess whether temporal trends in declined rates varied across population groups, regions and health specialities, we also fitted stratified models which were estimated by sex, age, ethnicity, deprivation, region and health specialty.

Multilevel models were not used in this analysis. While there was potential for clustering of referrals within individuals and within regions, the dataset did not include unique person identifiers, which precluded nesting or clustering by individual. Regarding regional clustering, we opted to include region as fixed-effect covariates in the models which aligned with the study aims and was manageable given the relatively small number of regions in the study.

The objective of regional analysis was to demonstrate variation across regions, as opposed to highlighting specific regions with notable disparities. For this reason, regions were anonymised in both the descriptive analysis and then again in the regression-based analyses. One region was removed due to large variation in referral reporting over time, indicating substantial data quality concerns.

Patient and public involvement

None.

Results

There were 3,034,339 FSA referrals in the raw dataset. Due to data concerns for a particular region, 115 782 referrals were excluded yielding an analytical sample of 2 918 557 first referrals for FSA during the study period. Table 1 shows annual counts of first referrals by year from 2018 to 2022. The data are disaggregated by sociodemographic, region and the 12 highest volume health specialties. Table 1 shows that the annual count of referrals has stayed relatively stable at approximately 560 000 to 6 00 000 per year, with a slight increase in 2020 and 2021. There were more first referrals for females and, generally, referrals increased by age. Referrals by ethnic group were reasonably representative of the NZ population, with a slight over-representation of Māori in the referral data. First referrals also increased with increasing levels of deprivation. As expected, there was substantial variation in the number of first referrals by region which generally reflected their population size. The highest volume health specialties included orthopaedic surgery, general surgery and ENT, which combined for approximately 30% of total referrals.

Table 1. Annual counts of first referrals for first specialist assessments (FSA) by sociodemographic characteristics, region and health specialty, 2018 to 2022.

	2018	2019	2020	2021	2022
	N (%)	N (%)	N (%)	N (%)	N (%)
Overall	563 222	564 073	585 683	610 707	594 872
Sex *
Male	257 244 (45.7)	258 502 (45.8)	267 285 (45.6)	277 021 (45.4)	271 769 (45.7)
Female	305 193 (54.2)	304 809 (54.0)	317 414 (54.2)	332 651 (54.5)	321 996 (54.1)
Age (years)
0–9	46 125 (8.2)	45 882 (8.1)	44 449 (7.6)	46 166 (7.6)	46 394 (7.8)
10–19	37 458 (6.7)	38 473 (6.8)	39 458 (6.7)	40 540 (6.6)	40 402 (6.8)
20–29	48 858 (8.7)	49 308 (8.7)	51 799 (8.8)	53 679 (8.8)	50 281 (8.5)
30–39	57 664 (10.2)	58 930 (10.4)	64 976 (11.1)	68 381 (11.2)	65 768 (11.1)
40–49	67 572 (12.0)	66 451 (11.8)	68 252 (11.7)	71 011 (11.6)	67 765 (11.4)
50–59	84 595 (15.0)	83 781 (14.9)	86 059 (14.7)	88 968 (14.6)	85 097 (14.3)
60–69	89 625 (15.9)	90 414 (16)	93 489 (16.0)	97 442 (16.0)	95 607 (16.1)
70–79	80 133 (14.2)	80 184 (14.2)	83 972 (14.3)	87 622 (14.3)	86 872 (14.6)
80+	51 192 (9.1)	50 650 (9.0)	53 229 (9.1)	56 898 (9.3)	56 686 (9.5)
Ethnicity
Māori	106 056 (18.8)	107 907 (19.1)	111 402 (19.0)	117 803 (19.3)	116 852 (19.6)
Pacific	44 836 (8.0)	44 044 (7.8)	45 325 (7.7)	46 479 (7.6)	47 281 (7.9)
NMNP	412 330 (73.2)	412 122 (73.1)	428 956 (73.2)	446 425 (73.1)	430 739 (72.4)
Deprivation†
1 (least deprived)	85 534 (15.2)	86 025 (15.3)	90 216 (15.4)	95 201 (15.6)	92 000 (15.5)
2	97 382 (17.3)	97 233 (17.2)	101 971 (17.4)	107 169 (17.5)	103 513 (17.4)
3	109 143 (19.4)	109 864 (19.5)	114 686 (19.6)	119 329 (19.5)	116 929 (19.7)
4	120 936 (21.5)	121 491 (21.5)	126 132 (21.5)	131 936 (21.6)	128 806 (21.7)
5 (most deprived)	132 678 (23.6)	132 240 (23.4)	135 674 (23.2)	140 446 (23)	138 093 (23.2)
Region‡
1	63 840 (11.3)	63 779 (11.3)	62 451 (10.7)	64 542 (10.6)	65 140 (11)
2	73 997 (13.1)	80 035 (14.2)	81 037 (13.8)	75 425 (12.4)	73 914 (12.4)
3	0 (n/a)	2414 (0.4)	5023 (0.9)	5202 (0.9)	5009 (0.8)
4	69 360 (12.3)	61 788 (11.0)	57 507 (9.8)	60 251 (9.9)	61 817 (10.4)
5	20 373 (3.6)	24 543 (4.4)	27 594 (4.7)	29 728 (4.9)	27 642 (4.6)
6	13 578 (2.4)	18 163 (3.2)	27 269 (4.7)	31 467 (5.2)	31 598 (5.3)
7	28 877 (5.1)	27 300 (4.8)	27 944 (4.8)	29 406 (4.8)	27 947 (4.7)
8	22 925 (4.1)	23 232 (4.1)	21 585 (3.7)	24 408 (4.0)	24 036 (4.0)
9	8816 (1.6)	9477 (1.7)	9509 (1.6)	9084 (1.5)	8525 (1.4)
10	13 810 (2.5)	14 067 (2.5)	11 236 (1.9)	13 338 (2.2)	8755 (1.5)
11	7006 (1.2)	7122 (1.3)	6672 (1.1)	7053 (1.2)	6101 (1.0)
12	25 916 (4.6)	16 904 (3.0)	15 777 (2.7)	15 907 (2.6)	15 457 (2.6)
13	38 891 (6.9)	39 480 (7.0)	41 109 (7.0)	44 039 (7.2)	44 022 (7.4)
14	42 058 (7.5)	41 918 (7.4)	40 218 (6.9)	4,2553 (7.0)	42 021 (7.1)
15	25 566 (4.5)	16 432 (2.9)	33 648 (5.7)	31 112 (5.1)	28 122 (4.7)
16	4054 (0.7)	10 103 (1.8)	9620 (1.6)	1,2071 (2.0)	10 699 (1.8)
17	8071 (1.4)	8654 (1.5)	8751 (1.5)	9829 (1.6)	9737 (1.6)
18	34 435 (6.1)	35 218 (6.2)	34 742 (5.9)	38 701 (6.3)	40 752 (6.9)
19	61 649 (10.9)	63 444 (11.2)	63 991 (10.9)	66 591 (10.9)	63 578 (10.7)
Health specialty
Orthopaedic surgery	57 022 (10.1)	59 331 (10.5)	58 318 (10.0)	60 906 (10.0)	60 030 (10.1)
General surgery	59 345 (10.5)	56 270 (10.0)	56 195 (9.6)	59 562 (9.8)	56 180 (9.4)
ENT	45 490 (8.1)	43 642 (7.7)	41 514 (7.1)	43 374 (7.1)	43 695 (7.3)
Ophthalmology	43 287 (7.7)	41 179 (7.3)	42 284 (7.2)	43 974 (7.2)	43 077 (7.2)
Cardiology	37 360 (6.6)	36 377 (6.4)	36 249 (6.2)	40 060 (6.6)	40 518 (6.8)
Gynaecology	32 476 (5.8)	31 074 (5.5)	32 290 (5.5)	33 941 (5.6)	31 528 (5.3)
Gastroenterology	29 236 (5.2)	30 349 (5.4)	31 136 (5.3)	30 889 (5.1)	30 894 (5.2)
Urology	23 897 (4.2)	22 853 (4.1)	24 294 (4.1)	25 518 (4.2)	24 916 (4.2)
Respiratory	23 789 (4.2)	22 190 (3.9)	22 310 (3.8)	23 882 (3.9)	24 443 (4.1)
General medicine	19 014 (3.4)	20 105 (3.6)	20 619 (3.5)	22 656 (3.7)	21 589 (3.6)
Paediatric medicine	16 791 (3.0)	17 947 (3.2)	17 973 (3.1)	19 540 (3.2)	20 413 (3.4)
Neurology	16 477 (2.9)	17 201 (3.0)	17 336 (3.0)	17 874 (2.9)	16 828 (2.8)

^*Missing data for 4673 (0.2%) events.

^†Missing data for 83 930 (2.9%) events.

^‡Regional data have been anonymised.

ENT, ear, nose and throat; NMNP, non-Māori/non-pacific.

The data in table 1 should be considered in the context of a NZ population that grew by approximately 7% over the study period. In contrast, the total number of referrals remained relatively stable. However, due to potential limitations in data completeness, we cannot be fully confident in drawing definitive conclusions about referral trends over time.

Prioritisation declined

Table 2 presents annual and total observed rates of prioritisation outcome declined across sociodemographic characteristics, regions and health specialties from 2018 to 2022. Overall, 13.1 percent of first referrals were declined across the 2 918 557 first referrals. There was a gradual increase in declined rates, peaking in 2021 (14.1%) before a slight decline in 2022 (13.9%). Female rates consistently exceeded male rates, with the gap widening slightly over the years. Age-related trends showed higher rates among those aged 30–79 years compared with younger or older groups. Ethnicity patterns revealed consistently lower declined rates for Pacific populations compared with Māori and NMNP. Socioeconomic gradients showed higher rates in less deprived quintiles, with the most deprived group (quintile 5) consistently exhibiting the lowest rates. Regional disparities were stark, both within and across regions. Health specialties displayed variable trends, with gynaecology and paediatric medicine experiencing notable increases in declined rates, while others, like gastroenterology and respiratory medicine, remained relatively stable. Respiratory medicine had the lowest declined rate annually, while the highest rates were observed in ENT.

Table 2. Prioritisation outcomes declined—annual observed rates per 100 referrals of declined by sociodemographic characteristics, region and health specialty, 2018 to 2022.

	2018	2019	2020	2021	2022	Total
Overall	11.6	12.4	13.2	14.1	13.9	13.1
Sex*
Male	11.2	11.9	12.6	13.6	13.2	12.5
Female	11.9	12.9	13.7	14.5	14.6	13.5
Age (years)
0–9	10.6	10.5	10.9	11.1	10.8	10.8
10–19	10.4	11.2	11.4	12.5	12.6	11.6
20–29	11.7	12.3	13.1	14.5	14.2	13.2
30–39	11.8	12.7	13.7	14.7	14.6	13.6
40–49	11.9	12.6	13.4	14.1	14.3	13.3
50–59	11.6	12.7	13.5	14.3	14.2	13.2
60–69	11.6	12.9	13.5	14.4	14.0	13.3
70–79	12.1	13.0	13.9	14.7	14.7	13.7
80+	11.7	12.6	13.7	14.7	14.4	13.5
Ethnicity
Māori	10.7	11.1	11.3	12.3	12.0	11.5
Pacific	6.7	7.0	7.6	8.0	7.9	7.5
NMNP	12.3	13.4	14.3	15.2	15.1	14.1
Deprivation†
1 (least deprived)	11.6	12.7	13.9	14.9	14.9	13.7
2	11.7	12.7	13.7	14.6	14.5	13.5
3	11.8	12.6	13.6	14.5	14.3	13.4
4	11.9	12.8	13.5	14.2	14.2	13.4
5 (most deprived)	11.3	11.8	12.2	13.1	12.7	12.2
Region‡
1	25.3	23.6	18.8	25.4	26.8	24.1
2	19.4	20.3	19.9	20.9	19.8	19.9
3	6.4	7.2	8.1	7.6	7.3	7.3
4	9.9	13.1	3.9	5.0	6.1	7.9
5	n/a	4.8	3.9	3.9	2.8	3.7
6	11.9	3.5	5.1	17.5	12.2	10.2
7	2.6	6.4	19.1	20.4	20.8	16.2
8	14.8	16.1	16.2	14.8	11.0	14.7
9	10.8	13.5	11.6	11.1	10.8	11.5
10	9.6	10.3	12.2	11.5	11.4	11.0
11	7.7	7.0	9.0	8.5	8.0	8.0
12	18.5	20.3	19.6	21.4	21.7	20.3
13	0.8	0.2	0.3	0.2	0.2	0.3
14	11.3	14.0	13.7	12.5	13.4	13.0
15	18.7	19.4	20.1	20.6	20.5	19.9
16	13.2	15.3	15.1	15.5	16.4	15.2
17	11.4	10.8	10.9	11.2	10.3	10.9
18	10.6	10.2	18.9	20.1	20.6	16.9
19	20.6	21.7	18.3	20.9	20.5	20.3
Health specialty
Orthopaedic surgery	15.0	16.3	17.3	17.1	15.5	16.2
General surgery	10.7	11.0	12.0	15.1	14.5	12.7
ENT	18.1	18.8	19.9	20.4	18.6	19.1
Ophthalmology	12.5	12.2	12.9	10.7	11.2	11.9
Cardiology	8.6	9.8	11.8	13.7	14.7	11.8
Gynaecology	11.1	14.9	17.5	15.9	18.1	15.5
Gastroenterology	14.4	11.7	11.5	11.4	11.7	12.1
Urology	11.7	14.1	14.6	15.7	17.6	14.8
Respiratory	7.5	9.3	10.6	9.8	9.3	9.3
General medicine	12.2	12.7	12.8	15.7	16.9	14.2
Paediatric medicine	14.0	13.9	15.4	16.8	17.9	15.7
Neurology	10.9	10.5	11.6	17.4	15.7	13.2

^*Missing data for 508 (0.1%) events.

^†Missing data for 8372 (2.2%) events.

^‡Regional data have been re-anonymised.

ENT, ear, nose and throat; NMNP, non-Māori/non-pacific.

Table 3 shows the regression results for pooled first referrals data (n=2,825,865) from 2018 to 2022 of declined at prioritisation on a linear year variable and controls for sex, age, ethnicity, deprivation and region. The regression results show that there was a significant 4.1% increased risk of being declined per year (risk ratio (RR), 1.041; 95% CI, 1.039 to 1.044). It also shows a significantly higher risk of declined for females compared with males. Younger people (0–19 years) and older people (80+years) had a significantly lower risk of being declined compared with those aged 50–59 years. In addition, Māori and Pacific peoples were also at significantly lower risk of being declined compared with NMNP. There was minimal variation in risk of declined by area-level deprivation but significant variation by region.

Table 3. Risk ratios for the association between prioritisation declined and year, sex, age, ethnicity, deprivation and region for pooled first referrals from 2018 to 2022 (n=2,825,865).

	RR	SE	P value	95% CI
Year	1.041	0.001	0.000	1.039 to 1.044
Sex
Male	reference
Female	1.069	0.003	0.000	1.062 to 1.075
Age (years)
0–9	0.853	0.006	0.000	0.841 to 0.865
10–19	0.891	0.006	0.000	0.879 to 0.904
20–29	1.003	0.006	0.658	0.990 to 1.015
30–39	1.020	0.006	0.001	1.008 to 1.031
40–49	1.004	0.006	0.506	0.993 to 1.015
50–59	reference
60–69	0.992	0.005	0.112	0.981 to 1.002
70–79	0.995	0.005	0.387	0.985 to 1.006
80+	0.955	0.006	0.000	0.944 to 0.967
Ethnicity
NMNP	reference
Māori	0.817	0.003	0.000	0.810 to 0.824
Pacific	0.706	0.005	0.000	0.695 to 0.716
Deprivation level
1 (least deprived)	reference
2	1.007	0.005	0.161	0.997 to 1.017
3	1.005	0.005	0.297	0.995 to 1.015
4	1.003	0.005	0.553	0.993 to 1.013
5 (most deprived)	0.998	0.005	0.653	0.988 to 1.008
Region*
1	reference
2	2.021	0.016	0.000	1.989 to 2.053
3	2.202	0.019	0.000	2.166 to 2.239
4	1.477	0.013	0.000	1.453 to 1.502
5	0.047	0.001	0.000	0.044 to 0.050
6	1.085	0.014	0.000	1.058 to 1.113
7	2.679	0.022	0.000	2.637 to 2.721
8	1.361	0.021	0.000	1.322 to 1.402
9	3.162	0.025	0.000	3.114 to 3.211
10	2.071	0.018	0.000	2.036 to 2.107
11	1.485	0.013	0.000	1.460 to 1.510
12	1.023	0.015	0.126	0.994 to 1.054
13	1.780	0.024	0.000	1.733 to 1.828
14	1.513	0.015	0.000	1.485 to 1.542
15	2.660	0.018	0.000	2.625 to 2.696
16	0.477	0.019	0.000	0.441 to 0.515
17	2.699	0.018	0.000	2.664 to 2.736
18	1.888	0.027	0.000	1.835 to 1.942
19	2.702	0.030	0.000	2.644 to 2.761

^*Regional data have been anonymised.

NMNP, non-Māori/non-Pacific.

Figure 2 displays the estimated associations (risk ratios) and corresponding 95% CIs for the linear time trend and declined at prioritisation, both overall and stratified by sex, age, ethnicity and deprivation. In essence, this figure illustrates whether the rate of referrals being declined changed between 2018 and 2022 for the total population and across all sub-populations examined. Findings show that the risk of being declined at prioritisation increased significantly over time for both the full population and for each sub-population. For instance, the overall risk of being declined rose on average by 4.1% per year (RR, 1.041; 95% CI, 1.039 to 1.044). This increase was slightly bigger for males (RR, 1.045; 95% CI, 1.042 to 1.048) than for females (RR, 1.037; 95% CI, 1.034 to 1.041). Age-related trends were generally consistent, though the 0–9 year age group experienced the smallest increase over time (RR, 1.010; 95% CI, 1.002 to 1.018). NMNP experienced the highest rate of increase in risk (RR, 1.045; 95% CI, 1.043 to 1.048), compared with Māori (RR, 1.028; 95% CI, 1.022 to 1.033) and Pacific peoples (RR, 1.026; 95% CI, 1.016 to 1.037). A clear deprivation gradient was also observed: those in the least deprived areas experienced the largest increase in risk over time (RR, 1.057; 95% CI, 1.052 to 1.063) compared with those living in the highest levels of deprivation (RR, 1.026; 95% CI, 1.021 to 1.031).

Figure 2. Risk ratios for the association between prioritisation declined and year, overall, and stratified by sociodemographic subgroups. NMNP, non-Māori/non-Pacific.

Figure 3 displays a suite of regression results, stratified by health specialty for the association (risk ratio) between prioritisation outcome declined and year. Results show that in 9 out of 12 health specialties, the declined rates increased significantly over time. The risk of declined increased the most for neurology (RR, 1.120; 95% CI, 1.107 to 1.133), gynaecology (RR, 1.118; 95% CI, 1.109 to 1.127) and cardiology (RR, 1.111; 95% CI, 1.101 to 1.120). In contrast, the risk for declined decreased significantly for gastroenterology (RR, 0.962; 95% CI, 0.953 to 0.971) and ophthalmology (RR, 0.964; 95% CI, 0.957 to 0.972).

Figure 3. Risk ratios for the association between prioritisation declined and year, stratified by health specialty. ENT, ear, nose and throat.

Figure 4 displays a suite of regression results, stratified by region for the association (risk ratio) between prioritisation outcome declined and year. Among the 19 regions, the majority experienced significant increases in the risk of declined over time. Among these, there was substantial variation across regions with the highest RR of 1.37 (95% CI, 1.361 to 1.387). Six regions had significant decreases in declined rates over time with the lowest RR of 0.640 (95% CI, 0.606 to 0.676).

Figure 4. Risk ratios for the association between prioritisation declined and year, stratified by region.

Discussion

This study offers new insights into the measurement of UMN for secondary care in New Zealand, with a particular emphasis on patient referrals originating from general practice and primary care settings. As noted, this area remains under-researched, nationally and internationally.^{4 8 17} There is limited understanding of the incidence and patterns of UMN, including patients’ care pathways, experiences and how UMN is managed across different parts of the health system and within the population. As in many other countries, New Zealand has never received explicit government support for measuring UMN in the context of secondary healthcare. This research has therefore provided important new baseline information and assisted with quantifying an important aspect of UMN for secondary healthcare.

The study found that between 2018 and 2022, the overall volume of first referrals from New Zealand GPs to public hospital specialists remained relatively stable. However, the proportion of FSAs declined at prioritisation rose from 11.6% to 13.9%. After adjusting for sociodemographic factors, this corresponded to a significant annual increase of 4.1% in the estimated risk of being declined, indicating a notable reduction in access to publicly funded specialist services. This trend likely reflects growing pressure on the health system, particularly in the context of constrained funding that has not kept pace with population growth and inflationary pressures.²⁴ In addition, ongoing specialist workforce shortages and rising administrative burdens may be reducing clinical capacity to assess and manage referrals.²⁵ These factors together suggest that rising decline rates may be driven less by changes in referral appropriateness and more by systemic constraints on service availability.

While the overall trend shows a significant increase in the risk of being declined at prioritisation over time, the pattern is not strictly linear, with a peak observed in 2021 and a modest decline in 2022. We interpret this as a trend-level rise in unmet need rather than uniform annual growth. The use of a linear term in our regression model offers a parsimonious summary of this broader pattern. Although the annual increases in declined rates may appear small in absolute terms, their population impact is considerable given the scale of national referrals, equating to thousands of additional patients experiencing unmet need each year. The slight drop in 2022 may reflect changes in referral or triaging practices, the evolving impact of COVID-19 disruptions, or variations in workforce and capacity, though further work is needed to disentangle these drivers.

Females were found to have a higher risk of having referrals declined compared with males, while those in young and older age groups, as well as Māori and Pacific peoples, were at significantly lower risk of being declined. The likelihood of being declined at prioritisation also varied significantly by region. Furthermore, all demographic subgroups analysed, including sex, age, ethnicity and area level deprivation, showed significant increases in the risk of being declined over time. Regional differences were especially pronounced, with some areas facing greater challenges than others. There was also considerable variation in the rate of increase in declined referrals over time by health specialty. The findings have important implications for equity and national consistency, as well as for the growing burden that declined referrals place on general practice and the broader primary care sector.

The higher overall risk of being declined for females may reflect broader patterns of gender-based inequities in healthcare access and prioritisation. Similar disparities have been documented internationally, where women are less likely to be referred for specialist care, experience longer wait times and are at greater risk of their symptoms being dismissed or deprioritised in clinical settings.^{26 27} In contrast, the finding of lower risk of being declined for younger and older people, and Māori and Pacific populations, has different possible explanations. One is that New Zealand’s ongoing policy focus on ensuring that certain populations, such as Māori and Pacific and different demographic groups with higher needs, are given a stronger weighting in the prioritisation process has had an impact. Weightings for ethnicity were removed in 2024, so any follow-up study to this one may indicate whether any changes have resulted.²⁸ Another explanation is that the prioritisation process adequately factors in complexity, meaning that the aforementioned groups, which often present with more complex health needs, are prioritised.^{10 29}

The regional variation observed in decline rates may, in part, reflect inherent differences in how health services are organised and delivered across the country, an issue identified in other studies of regional variation in care. These differences may stem from the nuanced and subjective nature of the referral process itself, which involves complex clinical judgments and interactions between patients and referring clinicians.³⁰ International research has shown that such variation in referral and access patterns is common, even in universal health systems, with factors such as local clinical culture, capacity constraints and institutional norms influencing whether patients receive specialist care.³¹

As noted previously, the study and measurement of UMN globally has predominantly been derived from self-reported survey data in the endeavour to understand broader prevalence. Such studies have found considerable variation across and within countries.432,35 The study of UMN has also focused on understanding at the point of clinical consultation through clinically reported data, revealing, for example, socio-economic differences.^{8 36} A systematic review of UMN in Canadian primary care found a range of socio-economic, geographic, demographic, ethnic and educational differences.³⁷ Many studies have focused on particular sectors such as mental health or older people and have continued to debate definitions and categorisation.^{17 38} Through measuring decline rates of clinically-assessed referrals from primary care through to secondary care specialists using national data over a 5 year time period, this study provides new insights into UMN. It also progresses the research field. To our knowledge, this is the first study globally to use this method.

The research contributes to related work on challenges the New Zealand health system faces with service access and care barriers, workforce capacity and funding.³⁹ There are implications which should be addressed with some urgency, particularly for patient suffering, patient management and the workload of those caring for patients rejected by specialist services. Such implications are noted in other studies globally.^{4 35 37} The increasing rates of declined referrals not only shift the burden of care back onto primary care providers but may also exacerbate existing health inequities, particularly for populations already disadvantaged by socioeconomic, geographic or systemic factors. Addressing these challenges will require a multifaceted approach, including investment in workforce capacity and enhancements in regional health system planning.

It is essential to deepen our understanding of the impact on patients who are declined access to specialist services within publicly funded systems like New Zealand’s. These patients are often left with no alternative but to return to the primary care sector, as the specialist system is unable to meet their needs. This challenge is common across health systems constrained by limited funding and capacity. In the New Zealand context, it places a significant burden on GPs, who are required to care for patients with a legitimate and clinically assessed need for specialist input. In some instances, GPs may choose not to refer patients because they anticipate the referral will be declined; in others, they are left managing complex needs that extend beyond their usual scope of practice. There is a wealth of research and practical experience available that could help guide policy and service responses to this issue.40,42

The research also highlights a growing challenge within the New Zealand healthcare system that arguably is occurring in other healthcare systems where there are insufficient resources to accommodate all referred patients. Therefore, the method employed for quantifying UMN reported in this article could be replicated elsewhere. Of course, national datasets would need to support this.

Strengths and limitations

This study is one of the first to use a comprehensive national administrative dataset to quantify unmet secondary healthcare needs in New Zealand, offering robust and novel insights into healthcare access disparities. By stratifying data by sociodemographic factors, regions and health specialties, the research highlights critical inequities and trends over time, providing actionable evidence for targeted policy interventions. The method employed is replicable and scalable, making it a valuable model for other countries aiming to measure UMN in a systematic and data-driven way.

However, this study must also be viewed in the context of several limitations. The analysis in this article relies on relatively new administrative data which may be subject to quality issues such as incomplete or inconsistent reporting. Notably, quality issues and data coverage appear to vary considerably by region. Therefore, regional results should be interpreted with caution. While our intended population includes all individuals eligible for public healthcare in New Zealand, systematic undercounting may occur, particularly among people living in high-deprivation areas who are less likely to be enrolled in a PHO and therefore under-represented in the available data.⁴³ The generalisability of these results may be limited by potential variations in referral processes and access across different healthcare systems, which warrant further exploration. Our use of generalised linear regression models reflects the event-level structure of the NPF. Because the data we were provided did not contain unique person-level identifiers across time, multilevel models with events nested within individuals were not feasible. While robust standard errors were applied to account for clustering, it is possible that individuals with multiple referrals are overrepresented. We also note that potential sources of residual confounding, including clinical severity, comorbidities and unobserved practice variation, were not able to be included in the analysis. These limitations are consistent with those observed in related register-based studies, including in the New Zealand multimorbidity literature.⁴⁴ Although missing data were minimal for key variables, we employed a complete case analysis approach, which may introduce bias if data are not missing completely at random. While declined referrals were used as a proxy for UMN, this measure does not capture other dimensions of UMN, such as patients who are deterred from seeking referrals due to systemic barriers (eg, cost to see a GP, distance, or waiting times). Furthermore, the analysis does not quantify or account for underlying differences in population health need, access to primary or secondary care or the clinical appropriateness of referrals. Without this context, it is not possible to determine whether variations in declined referrals reflect differential levels of need, differential access to care or differences in referral and triage practices across the health system. The study did not account for potential changes in triage criteria or referral practices over time, which could influence observed trends. Finally, while sociodemographic disparities were examined, the analysis does not explore intersectional impacts (eg, ethnicity and deprivation combined), which may provide further insights into inequities in access to care.

Future research should address these limitations to build a more comprehensive understanding of UMN in New Zealand. There is a need to better understand the drivers of the increasing rate of referrals being declined across New Zealand, why there is regional variation and why some groups are disproportionately affected. Future studies should aim to quantify the underlying levels of population health need, assess barriers to accessing both primary and specialist care and identify the specific clinical, systemic or administrative reasons why referrals are being declined. Understanding these factors is critical to ensuring more equitable access to specialist care and addressing unmet health need.

Conclusion

There is a history of UMN in New Zealand and globally that has not been adequately documented or evaluated. This study aimed to quantify an important marker of UMN for secondary healthcare. Deploying a quantitative method, it provides a new understanding of the level and growth over time of UMN. The study revealed a range of challenges that warrant a response. It will be important for policy makers to consider a series of activities in order to relieve the pressures being borne by New Zealand patients referred for specialist care and not receiving this. It would be useful, also, for researchers and policy makers in other jurisdictions to develop data and analytical methods for assessing UMN as this study has.

Data availability statement

Data are available upon reasonable request.