All-cause, premature, and cardiovascular death attributable to socioeconomic and ethnic disparities among New Zealanders with type 1 diabetes 1994–2019: a multi-linked population-based cohort study

Mortality rates among those with type 1 diabetes mellitus (T1DM) diagnosed in childhood and adolescence have been decreasing since the 1970’s, albeit with marked international variation [1]. Despite this trend, T1DM is still associated with an increased risk of death compared with the age- and sex- matched general population [2]. Premature death and deaths with cardiovascular disease (CVD) as the primary cause remain higher among people with T1DM [3‐6]. The importance of minimising hyperglycaemia was confirmed by the Diabetes Control and Complications Trial (DCCT), while the Epidemiology of Diabetes Interventions and Complications (EDIC) follow up study showed a reduction in all-cause mortality and major CVD events [7].

Beyond the DCCT/EDIC (n = 1441) [7], the Allegheny County type 1 diabetes registry (n = 1075) [8] and the Finnish Diabetic Nephropathy Study (n = 5396) [9], recent large T1DM cohort mortality studies are limited. In 2019, the International Diabetes Federation calculated all-cause mortality rates but for many countries, these required substantial extrapolation [10]. In 2014, Morgan et al. compared 23 studies of T1DM mortality across 20 countries [1] revealed that after a follow up of up to 19 years, those diagnosed under 15–19 years had mortality rates of 0–8.12/1000 person years and standardised mortality rates of between 0 and 854 (100 is the background rate). A 1984–1993 New Zealand (NZ) cohort study of those with insulin treated diabetes in Canterbury showed higher all-cause mortality for those with T1DM compared to those people with type 2 diabetes (T2DM) [11]. There are no cohorts of people with T1DM in NZ reporting population-level rates of all-cause, premature and CVD mortality, with a duration of 25 years.

One of the key questions beyond the degree of excess morbidity and mortality compared with the general population [2, 12], is how these compare with those with T2DM. Generally, T1DM has been seen as more “severe” than T2DM because of the dependency on insulin treatment for survival [13]. However, the Diabetes Incidence in Kronoberg Study revealed a standardized mortality ratio of 0.5 in the Swedish T1DM population compared with the T2DM population [14]. Other studies comparing T1DM and T2DM mortality rates have suggested that reduced access to modern care and medications are likely to affect this comparison [13]. Further influences have been identified including low socio-economic status and being from a minority ethnic group, which have been associated with higher mortality [15]. Reduction of these disparities in mortality is a key health policy goal of all governments [16]. However, efforts to lower mortality rates among people with T1DM, especially premature mortality rates and to mitigate these disparities are impeded by a paucity of information on the relative contributions of ethnicity and social circumstances. New Zealand is a developed country with a multi-ethnic population that can provide insights into this issue.

In the current study, using T1DM audit data linked with national databases, we aimed to 1) investigate the all-cause, premature and CVD mortality rates in people with T1DM between 1994–2019 in NZ; 2) estimate the standard mortality ratio with the T2DM population as a reference; 3) quantify socioeconomic and ethnic disparities in mortality rates in NZ.

Consistent with the Diabetes Incidence in Kronoberg Study, the current study found that all-cause, premature and CVD mortality in a T1DM cohort in NZ was 0.39, 0.65, and 0.31 times lower than that in the T2DM population, standardising for age and sex. There was no significant excess all-cause, premature and CVD mortality among non-NZE compared with NZE, while mortality increased significantly with increments in socioeconomic deprivation. The absolute socioeconomic inequality measurement suggested that 8, 5, and 3 of annual all-cause, premature and CVD additional avoidable deaths per 1,000 with T1DM occurred in the most deprived group compared with the least deprived group.

The population attributable fraction of ethnicity for the three mortality outcomes was not significantly different including after adjustment for potential socio-demographic and clinical factors. Conversely, 47% of all-cause mortality, 56% of premature mortality, and 56% of CVD mortality would not have occurred if all people with T1DM had the same risk as those in the least deprived socioeconomic group. These population attributable fractions were lower for all-cause and premature mortality when adjusted for ethnicity, age, sex, diabetes duration, enrolment year, smoking status, BMI, SBP, DBP, HbA1c, TC, triglyceride, LDL-C, HDL-C, eGFR, statin and antihypertensive medicine, which suggests that some of the socioeconomic disparities in all-cause and premature but not CVD mortality can be explained by the combined influences of these clinical characteristics. Adjustment for these characteristics had little effect on the population attributable fraction for socioeconomic disparities in CVD mortality.

Studies comparing mortality among people with T1DM and T2DM vary between different populations and by ethnicity. In the Finnish cohort including 211 people with T1DM and an 18-year follow up, CVD mortality was similar between people with T1DM and T2DM [26]. In an Australian cohort including 470 people with T1DM aged 15–30 years over 23 years of follow up, 0.5-fold and 0.6-fold less all cause and CVD mortality was found in the T1DM population compared with the T2DM population [27]. In a Hungarian cohort including 11,863 people with T1DM with a mean age of 22 years, a twofold higher risk of all-cause mortality was identified in the T1DM group compared with the T2DM group [28]. In a Korean cohort including 9397 people with T1DM aged 20 years and over, people with T1DM had a 1.9-fold higher risk of all-cause mortality compared to people with T2DM [29]. In the current cohort with 2395 people with T1DM and a median age of 35 years, overall median follow-up of 10 years, overall, 0.39-fold, 0.65-fold, and 0.31-fold decreased risk of all-case, premature, and CVD mortality (respectively), were identified. Similar findings were observed in the non-NZE sub-cohort. In the NZE sub-cohort, similar findings were only observed for all-cause and CVD mortality, with no difference in premature mortality between T1DM and T2DM. The inconsistent findings (higher mortality in T1DM than T2DM) might be due to different health systems, worse diabetes complications than the T2DM population over time ( eg in Asia (25); and/or the younger mean age, for example in the comparison in Europe (24).

Higher raw all-cause mortality and premature mortality rates were observed in Māori and Pacific ethnic groups, while a higher raw CVD mortality rate was observed in the NZE ethnic group. Due to sample size limitations, adjusted mortality rate ratios and adjusted excess mortality were not estimated for specific comparisons between each ethnic group and NZE. Furthermore, comparisons between NZE and combined non-NZE groups did not reveal any significant differences in all-cause, premature, or CVD mortality after adjustment for socioeconomic status, demographic and clinical characteristics. This lack of significance may be attributed to heterogeneity within the non-NZE groups. Future studies with larger sample sizes within each specific ethnic group are needed to enable more precise ethnic-specific comparisons.

Female MRR in certain age groups, particularly within the 25–34 age range, exhibits significant variations compared to male estimations, especially in the Non-NZE population, where no significant estimations were observed. There may be heterogeneity effects within this age group that warrant further investigation through the subdivision into smaller age groups in future studies with a larger sample size. The current study was constrained by limited outcomes within this specific age group of the Non-NZE population. However, significant attributable risks in relation to socioeconomic status were demonstrated for all three mortalities. CVD mortality, in particular, had over half of the excess risk attributable to socioeconomic deprivation, including after adjustment for ethnicity, demographic and clinical factors. Multiple barriers to care and self-care have previously been shown in the 20^th Century among people with diabetes in South Auckland [30] including those relating to socioeconomic deprivation and hence this disparity is not unexpected. More recently, this has been shown across NZ for people with T1DM where it was suggested that the NZ funding model via the Government Drug purchasing agency PHARMAC is unable to allocate equal resource between devices/pharmaceuticals and diabetes support and education [27]. Our study, with the high proportion attributable to deprivation, suggests that greater efforts should be made to reduce these socioeconomic disparities in the three mortalities among people with T1DM. Potentially, 8.04, 4.81, and 2.70 per 1,000 person-years in the average all-cause, premature, and CVD mortality between the least and the most deprived socioeconomic group could be eliminated by applying a ‘place-based’ public health strategy [25], (https://​www.​gov.​uk/​government/​publications/​health-inequalities-place-based-approaches-to-reduce-inequalities/​place-based-approaches-for-reducing-health-inequalities-main-report.) for people with T1DM and socioeconomic disadvantage in NZ. We have not calculated years of life lost, but the cohort had a mean entry age of 34.6 years suggesting that many of those dying would have bene of working age.

The study has a number of strengths including this being the longest (25 years) cohort study of a T1DM population in New Zealand to report all-cause, premature, and CVD mortality. The population basis of the cohort is a further strength with all patients derived from participating general practices. Through data linkage to large, nationally representative databases, we were able to follow up people to ascertain all incident cause-specific death. All health outcomes used in this study were based on the linkage of specific registration datasets, which provide high validity of outcomes. Accuracy of clinical recording and diagnoses in this study have been found to be valid for a range of comorbidities and we also used primary ICD codes for cause-specific outcomes, which have high precision [31]. The study limitations include heterogeneity of the non-NZE group and the degree of representativeness of the population (South and West Auckland vs NZ) and the participating general practices. Misdiagnosis between T1DM and T2DM is a well-known phenomenon in primary care [32], although we restricted the population to T1DM and with insulin treatment. Restricted by sample size, the population-level estimation in the current study could not be broken down to furthermore specific-ethnic groups in the non-NZE population.

In conclusion, our findings suggested that there is a lower risk of all-cause, premature, and CVD mortality for the population with T1DM compared with those with T2DM in this cohort. The lack of significant differences in mortality among ethnic groups with T1DM in NZ may be due to heterogeneity within non-NZE groups necessitating larger samples for precise ethnic-specific comparisons. However, socioeconomic disparities were attributable for a major proportion of the three mortalities. Significant deaths could be reduced if prevention strategies including better access to care, education and continuous glucose monitoring target people with T1DM with socioeconomic disadvantage.