30-Year Trends in Stroke Rates and Outcome in Auckland, New Zealand (1981-2012): A Multi-Ethnic Population-Based Series of Studies

Background Insufficient data exist on population-based trends in morbidity and mortality to determine the success of prevention strategies and improvements in health care delivery in stroke. The aim of this study was to determine trends in incidence and outcome (1-year mortality, 28-day case-fatality) in relation to management and risk factors for stroke in the multi-ethnic population of Auckland, New Zealand (NZ) over 30-years. Methods Four stroke incidence population-based register studies were undertaken in adult residents (aged ≥15 years) of Auckland NZ in 1981–1982, 1991–1992, 2002–2003 and 2011–2012. All used standard World Health Organization (WHO) diagnostic criteria and multiple overlapping sources of case-ascertainment for hospitalised and non-hospitalised, fatal and non-fatal, new stroke events. Ethnicity was consistently self-identified into four major groups. Crude and age-adjusted (WHO world population standard) annual incidence and mortality with corresponding 95% confidence intervals (CI) were calculated per 100,000 people, assuming a Poisson distribution. Results 5400 new stroke patients were registered in four 12 month recruitment phases over the 30-year study period; 79% were NZ/European, 6% Māori, 8% Pacific people, and 7% were of Asian or other origin. Overall stroke incidence and 1-year mortality decreased by 23% (95% CI 5%-31%) and 62% (95% CI 36%-86%), respectively, from 1981 to 2012. Whilst stroke incidence and mortality declined across all groups in NZ from 1991, Māori and Pacific groups had the slowest rate of decline and continue to experience stroke at a significantly younger age (mean ages 60 and 62 years, respectively) compared with NZ/Europeans (mean age 75 years). There was also a decline in 28-day stroke case fatality (overall by 14%, 95% CI 11%-17%) across all ethnic groups from 1981 to 2012. However, there were significant increases in the frequencies of pre-morbid hypertension, myocardial infarction, and diabetes mellitus, but a reduction in frequency of current smoking among stroke patients. Conclusions In this unique temporal series of studies spanning 30 years, stroke incidence, early case-fatality and 1-year mortality have declined, but ethnic disparities in risk and outcome for stroke persisted suggesting that primary stroke prevention remains crucial to reducing the burden of this disease.


Introduction
The burden of stroke is large and increasing worldwide, with notable ethnic/racial disparities. [1][2][3][4][5][6][7] Effective primary stroke prevention strategies are therefore critical in ageing populations [1,8] and where there is increasing numbers of people surviving with stroke-related disability and ongoing risk. [9] However, because of the considerable challenges to determining temporal trends in incidence and outcome of stroke, there is limited information on the impact of declines (or increases) in rates and case fatality in whole populations. This leaves uncertainties regarding the impact of public health policies and improvements in health service delivery for this important disease. [10][11][12][13] As a large proportion of the burden of stroke is borne outside the hospital sector, it is crucial that incident cases are ascertained and studied in a population-wide context. [10,11,13] Over the last two decades new primary and secondary preventative strategies have been implemented but the effect of the strategies on stroke burden have not been reliably assessed. [14] Accurate population-based data on population trends in stroke incidence and risk factors reflect the success or otherwise of prevention strategies, while case-fatality and mortality trends reflect management and natural history of stroke. All these indices are crucial for evidence-based planning and resource allocation for stroke services and preventive strategies. Yet population-based stroke incidence studies are complex, [10,13] and are limited in number when compared to those using hospital-based registers, mortality data, or studies limited to certain age groups. To address this gap, we used population-based data from the four Auckland Regional Community Stroke (ARCOS I-IV) studies (1981-1982, 1991-1992, 2002-2003 and 2011-2012) to determine 30-year trends in stroke incidence, case-fatality and mortality in the four major ethnic groups (NZ European, Asian, Māori and Pacific Islanders) in Auckland, New Zealand (NZ).
included ethnic groups such as other European, Middle Eastern, Latin American and African ethnicities) and NZ Europeans as per the NZ Census. [21] When multiple ethnicities were selfidentified, ethnicity was prioritized as a single ethnic group for the purposes of analysis in the following order: Māori, Pacific, Asian/other, NZ European. For example if a person self-identified as NZ European and Māori, they were classified as Māori, and if they self-identified as Pacific and Chinese they were classified as Pacific. Ethnicity information was not available for 2 of the ARCOS 2011-2012 study participants and 46 of the ARCOS 2002-2003 study participants, who were excluded from the ethnicity analyses. Age, sex and ethnic structure of the corresponding Auckland census data were used as the denominator in calculating incidence.
Crude annual age-, sex-, and ethnic-specific stroke incidence (first-ever-in-a-lifetime events) and attack rates (all events, including recurrent strokes), 28-day case fatality (proportion [%] of people with stroke who died within 28 days of stroke onset among the total number of people with incident stroke) and mortality rates (number of people with incident stroke who died in the study population over one-year follow-up period [nominator] divided by the study population at risk [denominator]), with 95% confidence intervals (CI) per 100,000 people were calculated assuming a Poisson distribution. Sampling procedure differences between ARCOS 1981ARCOS -1982ARCOS and 1991ARCOS -1992 studies were taken into account when computing CI and standard error (SE) of all estimated rates, as described elsewhere. [12] Standardised rates were calculated using the direct method and age-standardised to WHO 'World' standard population. [22] Rate ratios (RR) were calculated to evaluate differences in age-adjusted rates between the 1981-1982 and 2011-2012 study periods, and the Wald statistic test of heterogeneity across age groups were performed. [12] For categorical variables, the Cochrane-Armitage test was used to assess whether there was evidence of a statistically significant trend. For continuous variables, analysis of variance or the Kruskal-Wallis test was used.
Completeness of case ascertainment based on the sources of notification was determined using capture-recapture techniques. [23] This involved conducting log-linear modelling assuming a Poisson distribution (unadjusted for sample procedures), [24] and used the four main sources of notification: hospital, general practitioner, death certificate and other sources. The final model with the least deviance, (deviance is an indicator of 'goodness of fit' and a lower deviance suggests a better fitting model), for all studies included the main effects of the four sources and the 3-way interaction between hospital, general practitioner and death certificate. A Bonferroni-corrected threshold of p = 0.00013 (based on 380 tests conducted 0.05/380) was used to indicate statistical significance after adjusting for multiple testing.
To compare our findings with similar population-based studies carried out in other countries we searched Medline with the terms"stroke", "epidemiology", "incidence", "ethnic/racial", "trend(s)", and "population or community based" for population-based studies of stroke carried out between 1980 and 2014 and published in English. Only population-based studies from high-income countries that reported age-specific raw numbers of first-ever-in-a-lifetime strokes (numerator and populations at risk (denominator) sufficient to calculate incidence rates age-standardised to WHO 'World' standard population [22] were included in the analysis. We also compared trends in stroke incidence and mortality rates from our series of Auckland Regional Community Stroke (ARCOS) studies (1981-1982, 1991-1992, 2002-2003, 2011-2012) with those found in the Global Burden of Disease (GBD) 2010 study. [1,25]

Results
There were 5400 new stroke patients (47% men) registered across the four studies (Tables 1  and 2). Over the 30-year study period, the proportion of patients presenting with recurrent  Table 2. Crude, age-specific and age-standardised annual stroke incidence rates (first-ever strokes) per 100,000 people-years in Auckland, New Zealand across four study periods (1981-1982, 1991-1992, 2002-2003 and 2011-2012)   We observed that risk factors prevalence increased significantly over 30 years (Table 1), with the proportion of patients with a history of high blood pressure (!160/95 mmHg; including people on antihypertensive medication) increasing between the first and last survey by 15.0% (95% CI 11.7%-18.4%), myocardial infarction by 12.6% (95% CI 12.8%-15.1%), and diabetes mellitus by 12.5% (95% CI 10.1%-14.9%), across all ethnic groups except Māori for high blood pressure, myocardial infarction and Type 2 diabetes mellitus, and Asian/other people for diabetes mellitus. However, there was a statistically significant decrease in the frequency of current smokers among stroke patients by 11.9% (95% CI 9.1%-14.8%), except for Pacific people in whom the smoking rate over the last decade increased by almost two times. There was also a noticeable reduction in the proportional frequency of strokes due to SAH (2.5% [95% CI 2.3%-2.6%]), and a noticeable decline in 28-day stroke case-fatality of 14  Age-standardised stroke mortality rates (Table 3)  Over the 30-year study period, there was a 23% (95% CI 15%-31%) reduction in stroke incidence (Fig 1) and 27% (95% CI 21%-33%) reduction in stroke attack rate overall (Table 4 and Fig 1). These reductions were largely due to decreases in NZ Europeans (decrease 20% in incidence, 26% attack rates) and Asian/others (decrease 81% [95% CI 63%-90%] in incidence, 75% [95% CI 51%-87%] in attack rates) (Figs 2 and 3). However, in Māori and Pacific people, there were non-significant increases in stroke incidence (first-ever strokes) and attack rates (incident and recurrent strokes combined) between 1981-1982 and 2011-2012 study periods. Overall, decreases in stroke incidence and attack rates were limited to people 65 years or older (Fig 3). In people aged 15-64 years, there were no significant changes in stroke incidence or attack rates, except for Pacific people where stroke incidence and attack rates doubled (Fig 1).
In 2011-2012, 54% (95% CI 48%-61%) of first ever incident and 51% (95% CI 46%-57%) of recurrent strokes occurred in people younger than 75 years ( Table 2). The proportion of incident and recurrent strokes in Māori and Pacific people younger than 75  . The temporal trends in declining stroke incidence and attack rates were consistent in both males and females aged 65 years and older (Tables 2 and  4), with the most noticeable decline observed in males (Tables 2 and 4) and NZ Europeans (Tables 2 and 4, Fig 1).

Discussion
The main findings of this study include: (a) a clear trend towards reducing stroke incidence, 28-day case-fatality and 1-year mortality rates; (b) an increase in the frequency of elevated blood pressure, myocardial infarction, Type 2 diabetes mellitus and AF among stroke patients, but decrease in the frequency of current smoking; (c) better survival of acute stroke patients across all ethnic groups; and (d) persistent differences in stroke incidence, mortality rates and risk factor prevalence across the four major ethnic groups in NZ. Overall, the ethnic disparities have lessened overtime, especially over the last study decade. Our findings of decrease in stroke Stroke Incidence and Outcomes Trends in a Multi-Ethnic Community incidence and mortality rates overall in the study population but an increase, although not significant, in stroke rates in people younger than 64 years old (particularly females, Māori and Pacific Island people) are concordant with those of the GBD 2010 Study which identified a continuous decline in stroke incidence and mortality rates in high-income countries, [1] and a trend towards a stable or increasing stroke burden in people younger than 65 years. [1] The latter is of particular concern as our data showed an increased frequency of several major vascular risk factors in people with stroke that was also observed in the general NZ population and other countries. [14,[33][34][35] This includes an epidemic of obesity and Type 2 diabetes mellitus in children and young adults, [36,37] as well as behavioural and dietary risk factors. [38][39][40] An increase (or no reduction) in the incidence of stroke in middle-age populations has been observed in other studies, [41,42] thus re-emphasising the need for intensification of appropriate primary stroke prevention strategies, especially on at a population level. Although the overall decline in stroke incidence rates is in line with some population-based studies, [43][44][45][46][47] the decline in stroke incidence in NZ over the last 30 years is almost 20% less than that of comparable studies for a similar study period (Fig 4). [13,14,[26][27][28][29][30][31][32][48][49][50][51] Further, this study has highlighted that the relatively high level of stroke incidence in NZ is largely attributed to high rates in Māori and Pacific people. Moreover, unlike similar studies, [14,30,44] we found a significant trend towards increase in the prevalence of elevated blood pressure, history of myocardial infarction and diabetes mellitus among stroke patients and this  Table 4. Crude, age-specific and age-standardised annual attack (first-ever and recurrent strokes combined) rates per 100,000 people-years in Auckland, New Zealand across four study periods (1981-1982, 1991-1992, 2002-2003 and 2011-2012) Stroke Incidence and Outcomes Trends in a Multi-Ethnic Community may be a contributing factor to the observed differences in NZ compared to other high-income countries. Indeed, the 2008/2009 NZ Adult Nutrition Survey showed that there was an increase in mean systolic blood pressure in the overall population since 2002, but in particular in Māori men aged 35-74. The prevalence of hypertension in the NZ population in 2008/2009 was estimated at 31%, with only 15% on anti-hypertensive medication. [39] However, the increasing prevalence of elevated blood pressure, MI and Type 2 diabetes in stroke patients may also be related to increased detection of undiagnosed hypertension, diabetes and other health conditions over the study period, therefore these data should be interpreted with caution. The proportion of recurrent strokes has shown a statistically significant decrease over the study period overall and NZ European and Maori people in particularly, although the level of recurrent strokes (21.4%) in NZ remains relatively high compared to some other developed countries. [52] There was no statistically significantly change in the proportion of recurrent strokes in Pacific, Asian and other ethnic groups in NZ from 1981 to 2012 and but showed a trend towards increasing. This and the overall high rate of recurrent strokes in NZ (especially in Maori and Pacific people younger than 75 years) signify the need for better, culturally appropriate secondary stroke prevention strategies. The younger age of stroke onset in Māori is also reflected in the lower life expectancy at birth of Māori, which in 2006 was at least eight years less than that for non-Māori for both genders. [53] Of note is the increase in the prevalence of atrial fibrillation and low levels of premorbid anticoagulation of patients with AF over the last decade. Despite 36.1% of ischaemic stroke patients having AF in the 2011-2012 study, only 26.5% of these patients were being treated with anticoagulation therapy prior to stroke. A similar increase in the prevalence of AF in stroke patients was recently observed in another population-based study. [54] This, together with the global epidemic of AF, [55] suggests that it is now becoming the leading risk factor for stroke.
Similar diverging trends and ethnic disparities in stroke incidence have been shown in other comparable population-based studies in the US [43] and UK, [44] suggesting unfavorable trends in the prevalence of risk factors. [56] In the Atherosclerosis Risk in Communities cohort there Stroke Incidence and Outcomes Trends in a Multi-Ethnic Community was a 40% increase in the proportion of hypertension and diabetes from 1987 to 2011. [46] Ethnic differences in stroke risk have been attributed to differences in socioeconomic status (SES) and exposure to risk factors. [57] Lower SES groups have greater exposure to cardiovascular risk factors, including hypertension, smoking, poor diet, physical inactivity, diabetes and excess alcohol use. [58] Also, lower SES groups may have limited access to, or make less effective use of, services to manage these risk factors (e.g., early hypertension detection and control). [59] However, there is evidence that not all differences among ethnic groups are explained by differences in cardiovascular risk factors, suggesting genetic [60] and other factors such as socioeconomic disparities and the experience of discrimination [61] may be important. There remains uncertainty about the relative importance of stroke risk factor management and other factors in causing these inequalities. The impact of SES, changing family dynamics and cultural values on stroke risk and recovery in Māori and Pacific Island people is uncertain. [62] Ethnic-specific differences in stroke risk factors have also been shown in some other multi-ethnic populations [63][64][65][66][67][68][69][70] suggesting the need for culturally tailored primary and secondary preventative strategies. If the current temporal trend in the distribution of stroke risk factors in the population continues, we are likely to observe an increase in the incidence of stroke in the near future. Primary stroke prevention is a mainstream strategy to reduce stroke burden in a population. [71][72][73] However, the relative significance of various determinants of stroke occurrence as well as cultural appropriateness of stroke prevention strategies, especially among Māori and Pacific people, [3] are not known, hindering the development of effective, evidence-based primary stroke prevention strategies on both individual and population levels. [4,[74][75][76][77][78][79] Furthermore, a detailed analysis of recent changes in stroke risk factors is needed to explain the diverging trends in stroke incidence and mortality rates in the different ethnic groups in NZ and to further inform culturally appropriate primary stroke prevention strategies. [80] Better management (including enhanced admission rates to acute stroke units) and resulting improved 28-day and 1-year survival of acute stroke patients have been observed in other studies. [13,32,81,82] Although hospital admission has improved in this study over the 30 year period, there is still an underutilisation of acute stroke units (only 51% of stroke patients were treated in acute stroke units) and thrombolytic therapy for ischaemic stroke patients (only 5% of ischaemic stroke patients received alteplase [data not shown]) against the recommended Stroke Incidence and Outcomes Trends in a Multi-Ethnic Community corresponding figures in the NZ guidelines. [83] A recent individual-participant data metaanalysis of alteplase [84] demonstrated that irrespective of age or stroke severity, and despite an increased risk of fatal intracranial haemorrhage during the first few days after treatment, alteplase significantly improves the overall odds of a good stroke outcome when delivered within 4Á5 h of stroke onset, with earlier treatment associated with bigger proportional benefits. Similar to some previous observations [85,86] but contrary to others [87][88][89] we found a moderate, though non-statistically significant, decrease in the incidence of subarachnoid haemorrhage that may be related to the reduced smoking rates [90] in NZ. [91] The observed level and magnitude in stroke mortality decline in Auckland, NZ (3-fold decline [from 98/100,000 person-years in 1981-1982 to 37/100,000 person-years in 2011-2012]) is similar to official NZ statistics on stroke mortality rates, [92] supporting the validity of our data. However, our findings of faster declines in stroke mortality rates than stroke incidence rates, in line with our previous estimates, [93] suggest a further rise of stroke related disability over the next decade due to population growth and aging in NZ. This is of particular importance for Māori and Pacific groups where there has been a substantial increase in stroke in younger age groups, thus emphasizing the importance and priority of primary stroke prevention. A mismatch in temporal trends between stroke incidence and mortality rates has been shown in some other populations. [94] Novelty and Translation To the best of our knowledge, this is the first, longest and largest population-based survey to compare long-term trends and 1-year outcomes of stroke. There are several important new findings reported in this paper. The first is the diverging and changing trends in stroke incidence and stroke risk factors in different ethnic groups of the population. The trends in stroke incidence for Māori and Pacific Island people behave are similar to those observed in low to middle-income countries, while in NZ/ Europeans, trends in stroke incidence are like those in high-income countries with ethnically more homogeneous populations. Ethnicity/race is a proxy for disparity [95] and this is relevant internationally because other ethnically diverse societies will need to make provisions for population sub-groups prone to disparity to ensure that they have access to appropriate health care. Second, our findings were obtained before and during contemporary changes in prevention (primary and secondary) and service delivery (acute stroke unit care and thrombolysis) in a whole population context. This provides, for the first time, direct feedback to health-policy decision-makers on the relative success of these changes in medical practice. Specifically, the study demonstrated that current stroke preventative and management strategies are far more effective in reducing stroke risk and mortality on a population level than those used before mid-1990s. This should encourage health systems worldwide to seriously consider implementing modern stroke preventive and treatment strategies if they have not already done so. Third, this is the first documentation, to the best of our knowledge, of stroke mortality rates falling faster than stroke incidence rates. If this trend is also borne out in other populations then health systems will need to adapt in order to accommodate the likely increase in rehabilitation services needs in order to cater for this increase in prevalence and, therefore, make plans to implement more effective and culturally appropriate primary stroke prevention strategies at a population level. The differences in the stroke incidence trends by ethnicity are also important to take into account when projecting stroke burden in multi-ethnic communities on a national and global scale (including Global Burden of Disease estimates) and further reinforce the need for culturally appropriate prevention and health care strategies in multi-ethnic communities. The diverging trends in stroke incidence in the different ethnic groups and different rates of reduction in stroke mortality and incidence rates should be generalizable to other ethnically diverse health-care systems that have achieved similar risk factor modification.

Strength and limitations
The strength of our study is that it includes four large prospective population-based stroke registers (5400 stroke patients; 3,335,112 person-years of observation) across a long timespan (over 30 years: 1981-2012). The major limitation of the study was a relatively low number of strokes in some ethnic groups, especially in the Asian/other ethnic group making it difficult to determine whether trends in incidence, attack and mortality rates were statistically significant. Although the numbers were small, one possible explanation for the apparent large reduction in stroke incidence and mortality rates in Asian/other ethnic group may be related to the large influx of younger, presumably generally healthy, Asian immigrants in NZ that has been particularly marked over the last two decades. [21] The NZ Health and Disability sector uses the NZ Health Statistics definition for ethnicity 'as a social construct of group affiliation and identity' where ''ethnicity is the ethnic group or groups that people identify with or feel they belong to'. [96] In the 1981-1982 and 2002-2003 studies, only one ethnicity was collected for participants in each study, preventing us from applying prioritised ethnicity classification in these studies. Although this may have introduced a classification bias by ethnicity, we believe the effect of the bias was not large and was likely to be non-differential as most of the ethnic priority classifications in the 1991-1992 and 2011-2012 studies were also based on the first ethnicity self-identified by the study participants. In addition, less than 10% of the last surveys study participants indicated more than one ethnicity. We lacked statistical power to stratify our analyses by ethnicity to generate specific ethnicity-covariate estimates for risk of stroke onset due to having small numbers of stroke patients per ethnic minority, particularly the Asian subgroup. Given the continuous population increase in Auckland and in New Zealand in general, it is likely that we will have improved statistical power to investigate ethnicity-covariate interactions in a future ARCOS study. Another limitation of our series of studies is that the first two ARCOS studies (1981-1982 and 1991-1992) used a cluster sampling scheme approach for case-ascertainment [97] and, therefore, may have underestimated the number of stroke cases in these study periods. Although we applied the same criteria for cardiovascular risk factors recorded across all studies, we acknowledge that the accuracy of information about these risk factors was prone to some degree of variation and, therefore, should be interpreted with caution. However, the consistency of the assessment procedures, diagnostic criteria, the implementation of quality control procedures during the four studies, together with the decreased estimated number of missing stroke cases (as determined by capture-recapture analyses), reassure us that case-ascertainment was high across all these studies.

Conclusions
A continuous and recently accelerated reduction in total stroke incidence and mortality has been found over the last three decades in the Auckland region of NZ. Despite a trend towards closing the gap in reducing the difference in stroke burden between NZ European and non-European ethnic groups in NZ, the 15-year differences in the mean age of stroke onset in Māori and Pacific compared to NZ/ Europeans has persisted. Some positive changes in the profile of risk factors in people with stroke, such as overall decline in the proportion of current smokers and recurrent strokes, and increased use of antiplatelet, blood pressure and lipid lowering medications, were counterbalanced by increasing prevalence of high blood pressure, myocardial infarction, diabetes mellitus, AF, and low use of anticoagulants in people with AF. This, together with the ongoing discordance between trends in stroke incidence and mortality rates and persistent ethnic disparities in stroke burden, emphasises the need for better understanding of causes of ethnic disparities in stroke burden. Culturally appropriate primary stroke prevention strategies need to be prioritised at a population level. [98] Some of these strategies need government intervention in terms of increases in taxes on cigarettes and alcohol for example, reduction in the levels of salt and sugar in the population, [99,100] the use of advertising campaigns informed by behavioural change methods to reduce smoking, alcohol, increase physical activity, or evidence-based interventions. Government led health interventions should include stroke programmes ideally already at the school age in the regular teaching plan among other health education programs. To inform future stroke prevention strategies and reduce stroke burden, a better understanding (including modelling) [101] of the interplay between trends in stroke (and stroke sub-types) risk factors, demographic (e.g., aging of the population, ethnic/racial disparities) and health care (e.g., introduction of acute stroke units, thrombolysis for ischaemic stroke), morbidity (e.g. incidence and disability) and mortality rate changes in different populations is required. the support of many doctors, nurses, stroke service providers within and outside Auckland; and of course the ARCOS participants and their families and friends. We would also like to thank our International Advisory Committee members (Prof. Craig Anderson, Prof. Maurice Giroud) for their support throughout the study and comments on the draft of the manuscript. We also thank Helen McDonald for secretarial support for the study. Kathryn McPherson holds the Laura Fergusson Trust Chair.
The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. Other authors who had full access to the data include Rita Krishnamurthi, Kathryn McPherson, Suzanne Barker-Collo, P. Alan Barber, Varsha Parag, Emma Witt, Amy Jones, Priya Parmar, Rohit Bhattacharjee, and Braden Te Ao. All other authors (Bruce Arroll, Elaine Rush, Yogini Rathnasabapathy, Derrick Bennett, Martin Tobias, Craig Anderson, Alice Theadom, Max Abbott, Ruth Bonita) had full access to the aggregated data used in the manuscript.