Journal of the New Zealand Medical Association, 05-November-2010, Vol 123 No 1325
A comparative analysis of cardiovascular disease risk profiles of five Pacific ethnic groups assessed in New Zealand primary care practice: PREDICT CVD-13
Corina Grey, Sue Wells, Tania Riddell, Romana Pylypchuk, Roger Marshall, Paul Drury, Raina Elley, Shanthi Ameratunga, Dudley Gentles, Stephanie Erick-Peleti, Fionna Bell, Andrew Kerr, Rod Jackson
Studies since the 1990s have consistently shown that Pacific peoples in New Zealand have higher incidence, mortality and case-fatality rates from cardiovascular diseases (CVD) than New Zealand Europeans.1–4 Research also suggests that these outcomes are the result of a more adverse CVD risk factor profile among Pacific peoples compared to Europeans.5–10
However, Pacific peoples in New Zealand are not a homogeneous population. The term ‘Pacific peoples’ is an umbrella term describing about 7% of the New Zealand population who identify with at least one of the ethnic groups originating from the Pacific Islands of Polynesia, Melanesia and Micronesia.11 There are over 12 nations represented in New Zealand’s Pacific community. However most Pacific peoples identify with one or more of the four main ethnic groups (Samoan, Tongan, Cook Island Māori and Niuean).12
Most health-related surveys present aggregated data from these different Pacific ethnic groups, in part because of the small samples of each contributing ethnic group. Research into CVD has been no exception. Despite recommendations to investigate each Pacific ethnic group separately,13–15 only two previous studies have attempted ethnic-specific analyses of the CVD risk profiles of Pacific peoples in New Zealand, and both studies only had the statistical power to find substantial differences between Pacific groups.13,16 Also of note, the New Zealand Census Mortality Study (NZCMS) reported much higher CVD mortality among Cook Island Māori (RR 1.66 compared with Samoans), which was reproduced in both 1991–99 and 2001–04 cohorts.15 This mortality differential is unexplained.
PREDICT is a clinical decision support programme aimed at assisting primary care practitioners with CVD risk assessment and management.17 Since 2002, it has been implemented in nine PHOs throughout Auckland and Northland, representing about 65% of the population in these two regions. Over 10,000 Pacific participants have now been recruited into the PREDICT programme, which has generated the largest Pacific cohort ever assembled in New Zealand. This cohort does not currently include a representative sample of people living in New Zealand as only about 20% of the eligible population have been risk assessed to date. However it seems unlikely that there would be any systematic differences in the selection of patients from different Pacific groups for CVD risk assessment. Therefore, we present a comparative analysis of CVD risk factor profiles in the major Pacific populations living in New Zealand.
While the ethnic-specific risk factor profiles reported here do not represent prevalence estimates for the Pacific populations in New Zealand, the PREDICT study provides an opportunity to add to our very limited knowledge on CVD risk factor differences between Pacific populations. Comparing these risk profiles could help answer the question “Is it appropriate to aggregate data on CVD risk profiles from different Pacific ethnic groups living in New Zealand?”
PREDICT is a web-based clinical decision support programme for CVD risk assessment and management in routine primary care practice. Study methods and data definitions are described in full elsewhere.17
The software programme has been integrated with several of the most commonly used primary care patient management systems. This integration allows uniform, systematically coded CVD risk data to be automatically (and anonymously) extracted from a patient’s electronic medical record. Gaps in the data required to undertake a formal CVD risk assessment are then completed by either the GP or practice nurse on the PREDICT templates, which are then automatically written back to the patient medical record. Health data captured by PREDICT, including history of CVD, family history of CVD, and total cholesterol/HDL ratio, have previously been shown to be highly consistent with data held in electronic patient records.18
Risk profiles are sent via a secure broadband internet connection to a central server at the time of the assessment. Within seconds the clinician receives the patient’s calculated 5-year CVD risk as well risk management recommendations based on New Zealand CVD risk management guidelines.19 The central server stores the CVD risk factor profiles of each patient, and with permission from participating PHOs, these are extracted anonymously and linked, via an encrypted National Health Index (NHI) number, to national hospitalisation and mortality datasets. PREDICT data can also be linked to the New Zealand Health Information Service (NZHIS) NHI dataset that holds details of date of birth, gender, ethnicity and socioeconomic status according to the NZDep01 Deprivation index.
Ethnicity data can therefore be collected from the PREDICT template (originally from the patient’s electronic medical record, or entered manually by the practitioner) or via linkage to the NHI dataset. Both datasets have provision for up to three different ethnicities to be entered, so that each patient in the PREDICT cohort could potentially have up to six ethnicities.
Any person in the PREDICT cohort aged 35–74 years with a Pacific ethnicity on any one of the six potential ethnicity fields was included in these analyses. Pacific ethnicities were defined according to the Ministry of Health’s Ethnicity Data Protocols for the Health and Disability Sector as Level 2 codes 30 to 37.20 However, preliminary analyses revealed that the Fijian group (code 36) was an anomaly, making up over 11% of the total Pacific cohort but only 4% of Pacific peoples in official New Zealand statistics.12 Furthermore, as the CVD risk profile of the Fijian group was much closer to that of the Indian cohort than to that of the other Pacific groups,21 we suspected that some of those classified as Fijians were not ethnic Fijians, but rather Fijian Indians. Therefore this group (n=1341) was excluded from these analyses.
The Pacific groups included in this study were thus the level 2 ethnicity codes 31 (Samoan), 32 (Cook Island Māori), 33 (Tongan), 34 (Niuean), and a combined group of both codes 37 (Other Pacific peoples, including Tokelauans) and code 30 (Pacific peoples Not Further Defined, NFD).
Less than 1% of patients in the Pacific cohort identified with more than one Pacific ethnicity. Classification of ethnicity prioritised the smaller Pacific groups over the larger ones, as was done in the 2001 New Zealand Census22 and the Diabetes, Heart and Health Study 2002/03.16 This method gave first priority to Niuean, followed by Cook Island Māori, Tongan and lastly Samoan ethnicity. Only 70 people (0.7% of the Pacific cohort) identified themselves as Tokelauan (code 35), therefore they were included in the ‘Other Pacific peoples’ ethnic group (code 37).
The data extracts for these analyses included all PREDICT first risk assessments from August 2002 until January 2009. Data were analysed using Stata v10.0 statistical software. Men and women were analysed separately. Deprivation was assessed according to the NZDep01 index. NZDep01 is a census based small area index of deprivation, which assigns a relative deprivation score to each meshblock in New Zealand.23 Each individual was assigned the value according to their meshblock of residence. Risk ratios (RRs) with 95% confidence intervals (CIs) were calculated for each ethnic group with Samoans as the reference group, as they comprised the largest group, making up almost 50% of the Pacific cohort. A binomial regression model was used to calculate risk ratios (RR) adjusted for age. A post-regression test was used to test for any overall differences in proportions between groups. A linear regression model adjusting for age was used to calculate mean differences and 95% CIs for continuous CVD risk factor data, again with Samoans as the reference group.
Mean differences in Framingham risk scores were also calculated, using the original Framingham risk prediction equation24 rather than the New Zealand Guidelines Group-adjusted Framingham equation which adds a 5% 5 year risk increment to high risk ethnic groups including Pacific peoples.19 We decided not to adjust for deprivation, as there were only small differences in NZDep between Pacific groups and it is a relatively crude and indirect measure of deprivation which is most useful when there are major differences between groups.
Ethical approval—The PREDICT research project was approved by the Northern Region Ethics Committee Y in 2003 (AKY/03/12/314) and the national Multi Region Ethics Committee in 2007 (MEC/07/19/EXP).
Between 2002 and June 2009, baseline PREDICT CVD risk assessments were conducted on 10,301 people aged 35–74 identifying with at least one Pacific ethnic group, after excluding those classified as Fijian (as discussed in the Methods). Of these, 48% identified as Samoan, 17% as Tongan, 13% as Cook Island Māori, 8% as Niuean and 14% as Other Pacific or Pacific NFD.
Table 1 shows the baseline demographic characteristics of Pacific peoples in the PREDICT cohort by Level 2 ethnic group. In most Pacific groups, there were similar proportions of men and women receiving CVD risk assessments. The only exception to this was for the Tongan cohort (54% men).
On average, Pacific women were approximately 3 years older than their male counterparts. The distribution of age groups at first CVD risk assessment was similar across Pacific groups, with approximately 25% aged 35–44, 35% aged 45–54 and 25% aged 55–64. All Pacific groups were over-represented in areas of high deprivation, with approximately 75% residing in the two most deprived NZDep01 quintiles (deciles 7–10).
Table 2 presents the gender-specific age-adjusted risk ratios for three CVD risk factors (smoking, diabetes and a prior history of CVD) for Pacific groups, using Samoans as the reference group. There were no overall differences in the proportion of smokers among Pacific men (p=0.16). Cook Island women had the highest proportion of smokers among women (21%) and were almost 60% more likely than Samoan women to smoke. Niuean men had the highest burden of diabetes (almost 40%) among Pacific males.
Pacific women in all groups had a higher prevalence of diabetes than their male counterparts, with over one-third of all Pacific women having a diagnosis of diabetes. Tongan women were 26% more likely, and Niuean women 17% more likely, than Samoan women to have diabetes. Overall differences between Pacific groups in the proportion with a prior history of CVD were small.
Table 1. Baseline demographic characteristics of people in Pacific ethnic groups in the PREDICT cohort
Table 2. Age-adjusted estimates and risk ratios, with 95% confidence intervals, for smoking, diabetes and prior history of CVD for Pacific groups in PREDICT cohort by gender (reference group is Samoan)
Table 3 shows the age-adjusted mean systolic and diastolic blood pressures (BPs), total cholesterol/HDL ratio and Framingham 5-year CVD risk scores of Pacific groups, by gender. Niuean men and women had the lowest mean systolic and diastolic BPs of all groups (mean systolic and diastolic BPs 2 to 3 mmHg lower than Samoans). Tongan men had the highest mean total cholesterol/HDL ratio (0.26 higher than Samoan men), and Niuean men the lowest (0.16 lower than Samoan men). Tongan women had the highest mean total cholesterol/HDL ratio (0.14 higher than Samoan women after adjusting for age and deprivation).
Tongan men and women had the highest 5-year Framingham CVD risk scores and while these were statistically significantly higher than the Samoan reference categories, they only represented about a 10% relative difference in risk.
Table 3. Age-adjusted mean values and mean differences for systolic and diastolic blood pressures, total cholesterol/HDL ratio and Framingham 5-year CVD risk score for Pacific groups in PREDICT cohort by gender (Reference group is Samoan)
Body mass index (BMI) is not used in the Framingham risk equation to calculate CVD risk and it is therefore not part of the risk assessment template. BMI is only mandatory if clinicians wish to receive CVD management recommendations based on New Zealand guidelines. Therefore, BMI was only available for the 30% of this cohort whose general practitioner completed both a risk assessment and risk management template. Mean BMI for this subset did not differ significantly between different ethnic groups, with BMI ranging from 32.6 to 33.9 in men and 34.3 to 36.7 in women.
This study presents a comparative analysis of CVD risk profiles by the major Pacific ethnic groups living in New Zealand. The data were generated from a web-based CVD risk assessment and management decision supported system used in the majority of PHOs in the Northland and Auckland regions and represents the largest cohort of Pacific people ever assembled in New Zealand. As only about 20% of the eligible population has been risk assessed to date, these findings cannot be used as population prevalence estimates. However we assumed that there were unlikely to be important systematic differences in CVD risk factor screening between the different Pacific populations, so that it was reasonable to undertake comparative analyses. Interestingly, the age, gender and NZDep01 profiles of the individual Pacific groups were remarkably similar.
We found relatively small differences in overall CVD risk factor profiles between the five Pacific groups assessed, suggesting that it is reasonable to combine these data when describing the overall CVD risk profiles of Pacific peoples living in New Zealand. However there were substantial differences in the prevalence of some individual CVD risk factors. The largest relative difference for a single risk factor was the almost 60% higher proportion of smokers among Cook Island women compared to Samoan women, which is consistent with results from previous studies.16,25,26
Of note, smoking rates among Pacific groups in the PREDICT cohort were lower than those reported by other studies, including the 2006 New Zealand Census27 and the Pacific Drug and Alcohol Survey (PDAS).26 However, the reported Census data included all Pacific peoples aged 15 years and over, while our study included those aged between 35 and 74 years, and the prevalence of smoking decreases significantly with age.28 The PDAS reported smoking rates for Pacific peoples aged 13 to 65 years, and had only about one tenth the sample size of out study (n=338 Samoans, 228 Cook Island Māori, 232 Tongans and 207 Niueans aged 13–65 years).26 The PDAS did not have the statistical power and precision to provide meaningful comparisons between Pacific groups.
In our study, Tongan women and Niuean men and women had the highest burden of diabetes (approximately 40%), making them 20-30% more likely to have diabetes than their Samoan counterparts. Tongan women have previously been reported to have the highest prevalence of diabetes among Pacific groups in the Diabetes Heart and Health Study (DHAHS) 2002/03,16 and Niuean men the highest prevalence of diabetes among Pacific men in the Workforce Diabetes Survey (WDS) 1988-1990.13 However, our finding of a high burden of diabetes among Niuean women has not been described previously—possibly because of the small numbers of Niuean women included in earlier studies (only 13 in the WDS and 60 in the DHAHS).13, 16
Despite their high burden of diabetes, Niuean men and women in our study had the lowest mean total cholesterol/HDL ratios and systolic and diastolic blood pressures of the five Pacific groups (approximately 2-3 mmHg lower than Samoan men and women). Our finding that Niueans had lower blood pressure levels than other Pacific groups has not been reported previously. However, Niuean men and women were reported to have significantly lower mean total cholesterol:HDL ratios compared to other Pacific peoples in both the WDS and the DHAHS, suggesting that the differences in this particular CVD risk factor between Niueans and other Pacific groups are real.
The lower mean levels of systolic and diastolic BP and total cholesterol:HDL ratio in Niueans compared to other Pacific groups in our study may possibly be due to targeted risk management in people with diabetes, given they also had the highest burden of diabetes in these analyses. We intend to explore this hypothesis further once we have linked these data to national pharmaceutical dispensing records. Comparing our data with all other New Zealand surveys (except the Census) that have attempted to make comparisons of CVD risk factors between the different Pacific populations living in New Zealand is problematic because the largest of these other surveys was about one-tenth the size of the PREDICT cohort. Therefore random error is likely to be a major problem when considering the validity of comparative differences.
Unfortunately we were unable to provide any insights into the observation by Blakely et al.15 of an increased risk of CVD death among Cook Island Māori compared to other Pacific peoples. While we identified some differences between groups for individual risk factors, our summary measure of overall risk, using the Framingham score, suggested that Tongan men and women had a small increased risk compared to other Pacific groups, which had similar Framingham estimated risks. However the Framingham risk score was developed to help improve the targeting of risk factor management in individual patients and at best only explains about one third the variability in CVD event rates in a population.
While it is possible there were systematic differences in the selection criteria for risk assessing different Pacific groups which might explain our inability to detect a higher overall risk among Cook Island Māori, it is more likely that the tool was simply not sufficiently sensitive. We plan to investigate the predicted risk versus the actual observed event rate in this cohort, with a view to developing more accurate risk prediction equations.
This study highlights the ongoing problems with the collection of ethnicity data in the health sector. Prior to 2004, when the Ministry of Health introduced Ethnicity Protocols for the Health and Disability Sector,20 there were no standards for the coding and recording of ethnicity data in primary care. Consequently, inconsistencies in ethnicity recording have been noted in hospital29 and primary care records.30 A study by Riddell et al. involving a sample (n=665) of people who had previously been risk assessed using PREDICT also found that self-identified ethnicity was the same as that recorded in the primary care record for only two-thirds of the sample.31
Notably, this study found that only 37.5% of those recorded as Fijian on the primary care record agreed with this when asked to self-identify their ethnicity. This finding further supports our exclusion of PREDICT participants classified as Fijian from the analyses. Furthermore, Riddell’s study found zero agreement among those recorded as Pacific Not Further Defined (NFD) on the primary care record with their self-identified ethnicity—presumably because the majority of these people identify with a specific Pacific ethnic group (for example, Samoan, Tongan etc).31 Our finding that 959 people (9.3% of our cohort) were only classified as Pacific NFD, rather than as a particular Pacific ethnic group, suggests that there are still significant gaps in the accuracy of recording of ethnicity in primary care.
In conclusion, this study has highlighted several differences in CVD risk factors between five Pacific ethnic groups assessed in routine primary care, including higher proportions of smoking in Cook Island women, a higher burden of diabetes among Tongan women and Niuean men and women, and slightly higher estimates of 5-year absolute CVD risk in Tongan men and women. These differences suggest that Pacific peoples should not necessarily be treated as a single entity when designing community-based health promotion and disease prevention programmes, and that a targeted ethnic-specific approach to the reduction of some CVD risk factors may be appropriate.
Research is currently limited in this area, however, and studies will need to be conducted to evaluate whether ethnic-specific Pacific interventions are more effective than general Pacific interventions. In contrast, overall (absolute) CVD risk, as estimated by the Framingham equation, is remarkably similar between the Pacific ethnic groups, suggesting that the current practice of aggregating CVD risk data is reasonable for describing the overall risk factor burden for Pacific peoples living in New Zealand.
Competing interests: None.
Author information: Corina Grey, Public Health Medicine Registrar, Section of Epidemiology & Biostatistics, School of Population Health, University of Auckland; Sue Wells, Senior Lecturer Clinical Epidemiology, Section of Epidemiology & Biostatistics, School of Population Health, University of Auckland; Tania Riddell, Senior Research Fellow, Section of Epidemiology & Biostatistics, School of Population Health, University of Auckland; Romana Pylypchuk, Research Analyst, Section of Epidemiology & Biostatistics, School of Population Health, University of Auckland; Roger Marshall, Associate Professor in Epidemiology & Biostatistics, Section of Epidemiology & Biostatistics, School of Population Health, University of Auckland; Paul Drury, Clinical Director, Auckland Diabetes Centre; Raina Elley, Senior Lecturer, Clinical General Practice, School of Population Health, University of Auckland; Shanthi Ameratunga, Deputy Head – School of Population Health, Section of Epidemiology & Biostatistics, School of Population Health, University of Auckland; Dudley Gentles, Research Fellow, Section of Epidemiology & Biostatistics, School of Population Health, University of Auckland; Stephanie Erick-Peleti, National Pacific Tobacco Control Coordinator, National Heart Foundation; Fionna Bell, Clinical Director, Ta Pasefika Pacific PHO; Andrew Kerr, Clinical Head of Cardiology, Middlemore Hospital and Clinical Senior Lecturer, University of Auckland; Rod Jackson, Professor of Epidemiology, Section of Epidemiology & Biostatistics, School of Population Health, University of Auckland
Acknowledgements: The authors would like to thank the following PHOs, as well as their affiliated general practitioners, practice nurses and patients, for providing the data for analysis: ProCare Network North, Auckland and Manukau, HealthWest, Te Tai Tokerau, Manaia, Kaipara Care, Tihewa Mauriora and Whangaroa PHOs.
PREDICT was developed by a collaboration of clinical epidemiologists at the University of Auckland, IT specialists at Enigma Publishing Ltd (a private provider of online health knowledge systems), primary health care organisations, non-governmental organisations (New Zealand Guidelines Group, National Heart Foundation, Diabetes New Zealand, Diabetes Auckland), several district health boards and the Ministry of Health.
PREDICT software platform is owned by Enigma Publishing Ltd (PREDICT is a trademark of Enigma Publishing Ltd).
Funding: The PREDICT research project has support by HRC grants 03/183 and 08/121 from the Health Research Council. CG received a training endowment from the New Zealand College of Public Health Medicine. SW has been, and TR is currently, a recipient of a National Heart Foundation Research Fellowship.
Correspondence: Corina Grey, Section of Epidemiology & Biostatistics, School of Population Health, University of Auckland, PO Box 92019, Auckland, New Zealand. Fax: +64 (0)9 3737624; email: firstname.lastname@example.org
issue | Search journal |
Archived issues | Classifieds
| Hotline (free ads)
Subscribe | Contribute | Advertise | Contact Us | Copyright | Other Journals