Interrupted diabetes eye care in Te Toka Tumai Auckland 2008–2019: analysis of routinely collected health-facility data

Diabetic retinopathy is a leading cause of preventable vision impairment in the working-age population.¹ Optimal management of blood glucose, cholesterol and blood pressure, and avoiding smoking, help to reduce the risk of development or progression of retinopathy.^2,3 If significant disease develops, timely ocular treatment with laser or anti-vascular endothelial growth factor (anti-VEGF) injections is crucial to prevent vision loss.⁴

As symptomatic vision loss due to diabetic retinopathy manifests in the very late stages of the disease, regular retinal screening is necessary to ensure early detection and timely treatment.⁵ The World Health Organization (WHO) recommends that screening should occur every 1–2 years.⁶ Ideally, the screening interval would be personalised and take into account the various risk factors present for progression of diabetic retinopathy for each individual.⁷ However, for pragmatic reasons, the retinal screening interval is often fixed or based on the retinal grade.

In 2022, Aotearoa New Zealand’s (hereafter New Zealand) Ministry of Health – Manatū Hauora estimated that 307,400 people had diabetes (43.1 per 1,000 population).⁸ The retinal screening guidance in New Zealand (Appendix 1) recommends that people with no or minimal retinopathy (grade R0 and R1) attend retinal screening every 2 years (type 1) or 2–3 years (type 2).⁹ If a person develops a threshold severity of retinopathy (R3 and above), they are referred from retinal screening to an ophthalmology clinic for closer monitoring and treatment, and attendance at this appointment should generally occur within 6 months.⁹ Some people whose diabetic maculopathy has stabilised after treatment can be referred back to retinal screening. However, many will need to stay under the care of ophthalmologists, who then decide the follow-up frequency depending on the severity of the disease and the underlying risk factors such as blood glucose level and hypertension.⁹

Each time a patient experiences interrupted care across the diabetes eye care pathway, an opportunity to prevent vision loss might be missed. We recently reported that between 2006 and 2019, only two-thirds of people with diabetes in New Zealand had engaged with diabetes eye care (retinal screening or an eye clinic) at least once every 2 years.¹⁰ This estimate was based on the analysis of administrative data without information on where on the pathway people became disconnected from care.

In this analysis, we address the knowledge gap observed in our previous work by using more detailed facility-based clinical data in Te Toka Tumai Auckland (previously Auckland District Health Board).¹¹ This study aimed to 1) summarise loss to follow-up through the diabetes eye care pathway among people aged ≥15 years in the Auckland Region and 2) identify disparities across population sub-groups by age, sex, ethnicity, area-level deprivation and duration of diabetes.

Method

2.1 Ethical approval and reporting

Ethical approval for this study was obtained from the Health and Disability Ethics Committee (ref: 21/NTB/97), and locality approval was obtained from Te Toka Tumai Auckland (ref: A+9307). We have reported the study in accordance with the RECORD statement (REporting of studies Conducted using Observational Routinely-collected Data, Appendix 2).¹²

2.2 Study design and setting

This is a retrospective cohort study. During the period the data were collected (and before the health sector reforms that commenced in July 2022), Te Toka Tumai Auckland was one of 20 district health boards across the country and was responsible for delivering healthcare services to the approximately half a million people living in and around central Auckland.¹³

2.3 Participants

The study population was identified from the diabetes retinal screening database of Te Toka Tumai Auckland and consisted of people with diabetes residing in Auckland who attended at least one retinal screening appointment at Te Toka Tumai Auckland between 1 January 2008 and 31 December 2019 (inclusive) at which time they were aged 15 years or above. The individuals were followed from the date of the first screening event until death or the study end date (31 December 2019); all events related to retinal screening or diabetes-related ophthalmology visits were analysed.

2.4 Data sources

The cohort was established from the Diabetes Retinal Screening Service dataset of Te Toka Tumai Auckland (held on a bespoke database “HCC” that serves the diabetes service). From this dataset, we received demographic data of individuals (age, sex, ethnicity and area-level deprivation) and clinical information related to each retinal screening event attended, such as the duration and type of diabetes and retinopathy grades.

The National Health Index (a unique identifier of the individuals) was used to link data from additional sources. Data related to the engagement of individuals with ophthalmology services were obtained from the National Non-Admitted Patient Collection (NNPAC) dataset held by the Ministry of Health – Manatū Hauora (Purchase Unit codes; S40002 = first attendance; S40003 = subsequent attendance).¹⁴

Similarly, the mortality collection dataset¹⁵ was used to obtain the date of death data of all individuals who died during the study period. We sought missing retinopathy grades from the Te Toka Tumai Auckland clinical database, which houses patient records and letters (n=549 individuals had at least one missing retinopathy grade and linkage identified a grade for 502 of these [91.4%]).

2.5 Study variables

Outcome variables

The primary outcome was the interrupted care status at each stage of the diabetes eye care pathway (Figure 1), with each eligible person being categorised as one of:

1. complete loss to follow-up, when an individual never attended the expected next event;
2. delayed care, when an individual attended the expected next event, but not within the expected time frame; or
3. no interrupted care, when an individual attended all expected next events within the expected timeframe.

This status is reported as a proportion of the eligible population at each stage of the diabetes eye care pathway, from retinal screening (stage 1), when referred from screening to ophthalmology (stage 2) and from ophthalmology review (stage 3). Details of the definition (and calculation) of the eligible denominator population, follow-up period and numerator population are provided in Appendix 3.

View Figure 1, Table 1–4.

Explanatory variables

The explanatory variables were: age (categorised as 15–29, 30–39, 40–49, 50–59, 60–69, 70–79, 80+ years); sex (female, male or unspecified); prioritised ethnicity, whereby people are assigned to one ethnic group regardless of how many groups they report belonging to (categorised as Māori, Pacific, Asian, NZ European or Other)¹⁶; area-level deprivation (based on NZDep13¹⁷ mapped from the domicile code on the National Health Index) arranged into quintiles from least (Q1) to most (Q5) deprived; and diabetes type (type 1, type 2 or Other).

2.6 Data processing and analysis

We calculated the expected number of events for each individual (follow-up duration divided by the recommended screening intervals of 24 months for retinal screening, 6 months for referral attendance and 12 months for ophthalmology review) and identified the individuals who experienced 1) complete loss to follow-up, 2) delayed care and 3) no interrupted care at each of the three stages in the pathway (Figure 1). For people with delayed care, the proportion of all expected appointments that were attended was also calculated.

Our use of the 24-month cutoff for retinal screening interval and the 6-month cutoff for referral attendance interval was based on Ministry of Health – Manatū Hauora guidance,⁹ while the 12-month cutoff for ophthalmology review was informed by the modest (generic) follow-up interval practised in Te Toka Tumai Auckland (where co-author DS practiced).

Descriptive statistics were reported for patient characteristics. Counts and percentages were used for categorical data, and numerical data were summarised in terms of median and interquartile range (IQR). The relationship between the outcome variable at each stage of the care pathway and the explanatory variables was analysed for disparities across population sub-groups by age, sex, ethnicity and area-level deprivation quintile using log-binomial regression, controlling for relevant confounders. To explore if there were distinct patterns between people accessing services (fully or partially) and those not accessing services, we ran two models dichotomising the outcome variable as 1) “complete loss to follow-up” (Y=1) versus “no interrupted care or delayed care” (Y=0) and 2) “delayed care” (Y=1) versus “no interrupted care” (Y=0).

We did not include the “diabetes type” variable in the models since 95% of participants had type 2 diabetes, and the models did not converge with this variable. People with missing ethnicity data (n=32) were also excluded from the models. The decision of the model fit was informed by a dispersion statistic value closer to one (not over- or under-dispersed estimates). The direction and strength of association were estimated in terms of relative risks (RR) and their 95% confidence intervals (CIs). All tests were two-sided with a 5% significance level. All analyses were conducted in R 4.5.0 (R Core Team, 2024, R Foundation for Statistical Computing).

 We conducted a sensitivity analysis for loss to follow-up estimates for retinal screening and ophthalmology by adding 2 months to the 24- and 12-month follow-up periods, respectively, to allow for scheduling or rescheduling of appointments around the due date, according to the Ministry of Health – Manatū Hauora guidelines.⁹

Results

We received de-identified retinal screening records for 21,617 people who attended diabetes retinal screening at Te Toka Tumai Auckland during the study period (1 January 2008 to 31 December 2019). We removed records of residents outside of Te Toka Tumai Auckland at the time of their appointment (n=3,240) as well as records provided erroneously (five people aged <15 years at their first appointment and five women with gestational diabetes). Of the remaining 18,367 people, 47 with missing retinopathy grades at their first appointment were excluded, leaving 18,320 people for analysis (study population selection shown in Appendix 4). After removing people ineligible for analysis at each stage of the pathway due to death or an insufficient follow-up period, there were 14,708 people eligible for analysis from retinal screening, 1,909 for their referral appointment and 1,136 for ophthalmology review (Appendix 4). At the retinal screening stage, eligible people had a median (IQR) of 7.4 (4.8–9.9) years of follow-up between their first engagement with screening and the study end date (or their death); people engaged with ophthalmology had a median (IQR) of 6.1 (3.4–8.5) years of follow-up.

Half of the 14,708 individuals eligible for analysis for retinal screening (stage 1) attended retinal screening events at least once every 24 months during the follow-up period and therefore had no interrupted care (n=7,267; 49.4%). A further 30.4% of participants (n=4,471) experienced delayed care, while one in five individuals were completely lost to follow-up (n=2,970; 20.2%) (Table 1). The median proportion of retinal screening events attended by those with delayed care during the study period was 50.0% (IQR 40.0–66.7%). The proportion of people attending all expected retinal screening events reduced as the period of follow-up (and therefore number of expected events) increased, from 73.1% (1,921/2,627) for people with two expected events down to 29.6% (1,011/3,421) for people with six expected events (Appendix 6).

Sensitivity analysis showed that extending the follow-up period by 2 months moved 589 individuals from delayed care to no interrupted care at retinal screening, increasing the proportion of eligible individuals experiencing no interrupted care from 49.4% to 53.4% (4.0 percentage-point increase).

The corresponding increase of people experiencing no interrupted care at referral and ophthalmology review was 3.7 and 7.0 percentage-points respectively (Appendix 5).

The likelihood of complete loss to follow-up from retinal screening was significantly higher among adults aged 30–39 years (RR=1.46; 95% CI 1.13–1.89) and 40–49 years (RR=1.32; 95% CI 1.11–1.58) compared with those aged 50–59 years, and among all ethnicity groups compared with NZ Europeans, being greatest for Māori (RR=1.46; 95% CI 1.26–1.70) and Pacific peoples (RR=1.45; 95% CI 1.28–1.65). There was a gradient of increasing likelihood of complete loss to follow-up with increasing area-level deprivation (Table 2).

The likelihood of delayed care from retinal screening showed a similar pattern across ethnicity groups and area-level deprivation but differed across age groups, with younger age groups being less likely than those aged 50–59 years to have delayed care (Table 2).

When referred to ophthalmology from retinal screening, 43.3% (n=828/1,909) of the eligible individuals attended within 6 months. A further 648 people (34.0%) attended at some time after 6 months, while 433 (22.7%) never reached ophthalmology (Table 1). Of those reaching ophthalmology (n=1,476), the median time between screening and review was 5.1 months (IQR 3.0–12.5 months; range 4 days to 123 months). Almost one-fifth of people reaching ophthalmology (n=268; 18.1%) did so between 6 and 12 months after referral, while one-quarter (n=380; 25.7%) attended after more than 12 months (Appendix 6). There were no significant differences in the likelihood of complete loss to follow-up versus no interrupted care/delayed care during referral across the population groups. However, among the group who experienced delayed care during referral, men were 16% less likely than women to attend their referral within 6 months (RR=0.84; 95% CI 0.75–0.95), and Asian people were 19% less likely than NZ Europeans to attend within that time frame (RR=0.81; 95% CI 0.69–0.95) (Table 3).

Among the 1,136 individuals eligible for analysis at ophthalmology review, one in six (n=188; 16.6%) experienced complete loss to follow-up, while approximately one-third (n=366; 32.2%) experienced delayed care and one-half (n=582; 51.2%) attended all expected ophthalmology reviews (at least once every 12 months). Of those with delayed care, the median proportion of the expected annual ophthalmology reviews attended was 56.3% (IQR 40.0–75.0%). The proportion of people attending all expected ophthalmology review events was highest for people with two expected events (70.2%; 85/121) and thereafter varied between 42.5% and 57.1% for up to 11 expected events (Appendix 6).

Similar to the result seen for retinal screening, people aged <40 years and 40–49 years were more likely to experience complete loss to follow-up from ophthalmology review compared with individuals aged 50–59 years. In contrast to retinal screening, Pacific peoples were less likely to experience complete loss to follow-up compared with NZ Europeans; there were no other significant differences across population groups for complete loss to follow-up or delayed care (Table 4).

Discussion

We analysed individual-level clinical data of people with diabetes aged ≥15 years in Auckland, New Zealand, from 2008 to 2019, and found that at each stage of the diabetes eye care pathway (screening; referral to ophthalmology; subsequent ophthalmology care) approximately half of individuals attended appointments within the recommended time interval, one-third experienced delayed care and one in five were completely lost to follow-up. For retinal screening, complete loss to follow-up was higher for Māori and Pacific peoples versus NZ Europeans, younger versus older age groups and people living in more versus less deprived areas. For referral and subsequent ophthalmology clinic care, differences between population groups were less evident.

Our results are broadly consistent with other studies in New Zealand, though there is heterogeneity in how findings are reported.^18–20 This includes our previous study that used national Ministry of Health – Manatū Hauora data and found that between 2006 and 2018, among people with a median of 5.5 years of follow-up, 62% attended retinal screening at least every 2 years.¹⁰ While the analysis reported here has used robust, routinely collected data from Te Toka Tumai Auckland, the lower result is likely due to the inability to monitor the engagement with diabetes eye care of people who have moved outside the region.

This inability to link with data from other regions highlights the need to monitor eye care data at a national rather than regional level. A situational analysis of eye care in 2022 concluded that the health information system needed major strengthening to enable effective monitoring of eye health.²¹ For diabetes eye care, in addition to strengthening data infrastructure and analytical capacity, monitoring efforts will be bolstered when a national diabetes register replaces our virtual register^22,23 to provide a complete list of everyone in need of regular eye health services (either retinal screening or ophthalmology).

A further enhancement of the information system would see key clinical information recorded at each eye care event including risk factors such as most recent HbA_1c,²⁴ retinopathy grade, visual acuity and management. These enhancements would enable assessment of the impact of interrupted care on patient outcomes such as progression of disease and avoidance of vision loss, as well as more readily enable individualised retinal screening intervals to provide more equitable and cost-effective care.⁷ For example, if there was a record of referral from retinal screening to ophthalmology review in the datasets, we would not require the assumptions and data linkage undertaken for this analysis, and progress through the care pathway for people with maculopathy (which has a less distinct referral threshold than retinopathy⁹) in the absence of referrable retinopathy could also be assessed. A further example of the benefits of a robust information system is seen in the Danish context, where 5 years of follow-up data demonstrated each missed appointment by people enrolled in the retinal screening programme roughly doubled the risk of progression to severe disease.²⁵ Ideally such analyses would be possible in New Zealand so that we can evaluate efforts to improve access to care and avoid vision loss from diabetes.

Reasons for the interrupted care we observed are likely to be multifactorial, with a combination of service- and patient-related factors involved. Service-related factors, such as delays in appointment scheduling, inflexible appointment times, poor communication with patients, geographic distance from eye services and transport links, reduce a person’s ability to access timely services in many settings including New Zealand.^26–28 During the study period, diabetes eye services in Te Toka Tumai Auckland was predominantly run from a single hospital setting with a relatively expensive car park that is frequently full. This centralised service and car park were key barriers expressed by people disconnected from care at Auckland diabetes eye services that was not shared by people enrolled at Counties Manukau, which has a more decentralised service.²⁸ The mobile screening service in Te Tai Tokerau provides another example of a service delivery strategy that can assist with improving access to eye care services for people with diabetes.^10,29

Patient-related barriers include discordance between presence of diabetic retinopathy and patient awareness of having diabetic retinopathy,^28,30 self-perceived need for eye services, and time constraints (e.g., work and family commitments).^28,31 A key finding of our analysis is the higher complete loss to follow-up from retinal screening among younger adults (30–39 years and 40–49 years) compared with people aged 50–59 years. This may reflect higher mobility among this age group compared to older adults, and they may be accessing services in other regions.³² However, given this younger group have decades of care ahead of them to maximise their health, they are an important group to focus on when planning how to improve access to diabetes eye care.³³ Our finding that there were fewer differences between population groups in loss to follow-up at the referral and ophthalmology review stages may be explained by an increased self-perception of need for services if they are experiencing vision loss or other impacts on their day-to-day functioning, whereas many people at retinal screening are likely asymptomatic.³⁴

Among underserved groups, barriers in accessing eye care are even greater, and consequently it is unsurprising that their loss to follow-up rates are higher.³⁵ The Pae Ora (Healthy Futures) Act, launched in 2022, and the New Zealand Health Strategy emphasise the role of the health system in improving equity and access through people-centred approaches.³⁶ To achieve equity in diabetes eye services, these approaches need to address systemic barriers that disproportionately affect underserved groups. Indeed, we have extensive evidence on barriers experienced by underserved groups with diabetes when accessing eye care^28,35,38 or other health services.³⁹ Future efforts must move towards developing and evaluating solutions that improve access to eye care for people with diabetes.⁴⁰ Fortunately, there is evidence on successful strategies used elsewhere on which we can draw, including moving screening to primary care or community settings via teleretinal screening^41,42 or mobile services,⁴³ and AI-assisted screening.⁴⁴ The impact of any new interventions on equity must be monitored to avoid unintended widening of health disparities.³⁷

Our results should be interpreted in the context of several limitations. First, the available data did not include the follow-up period decided at each individual consultation, so we used generic follow-up intervals for each stage of the pathway. The recommended follow-up period may have been shorter for some people who were assessed to be at higher risk of developing diabetic retinopathy,²⁴ so we may have under-estimated delayed care at all stages of the care pathway. Second, due to how ophthalmology services are coded, there was no ophthalmology consultation code specific to diabetic retinopathy. However, as this cohort was created from people who were referred from retinal screening, any ophthalmology consultations provided to people in the cohort for reasons unrelated to diabetic retinopathy will likely be small enough to not inflate the loss to follow-up rates at the referral and ophthalmology review stages reported here. Third, because we used retinopathy grading as the criteria to assign need for referral, we have not assessed time to ophthalmology appointment for people referred for other reasons, such as severe maculopathy in the absence of referrable retinopathy, poor image quality or cataract. If some of these people were experiencing reduced vision, they may be prioritised in the referral process, or their care-seeking may be more proactive compared to the generally asymptomatic people included in our analysis, meaning we may have over-estimated interrupted care by omitting these groups from our analysis.

In conclusion, our study found high levels of loss to follow-up and delayed care at every stage of the diabetes eye care pathway in Auckland, with pronounced differential access observed between population groups for retinal screening. Interrupted care is likely to contribute to vision loss, particularly as diabetic retinopathy is asymptomatic until it reaches an advanced stage.

View Appendices.