Content uploaded by Barry J Milne
Author content
All content in this area was uploaded by Barry J Milne on Jan 13, 2016
Content may be subject to copyright.
THE NEW ZEALAND
MEDICAL JOURNAL
Vol 119 No 1235 ISSN 1175 8716
NZMJ 2 June 2006, Vol 119 No 1235 Page 45 of 145
URL: http://www.nzma.org.nz/journal/119-1235/2002/ © NZMA
The Dunedin Multidisciplinary Health and Development
Study: are its findings consistent with the overall
New Zealand population?
Richie Poulton, Robert Hancox, Barry Milne, Joanne Baxter, Kate Scott,
Noela Wilson
Abstract
Aims To compare the health of the Dunedin Multidisciplinary Health and
Development Study members with people of the same age in the nationally
representative New Zealand Health and National Nutrition Surveys.
Method Where similar information was obtained, means or proportions and
confidence intervals were generated for both the age 26 assessment of the Dunedin
sample and for the 25–26 year old participants in the national surveys. The
populations were considered to differ when confidence intervals did not overlap.
Results For smoking habit, body mass index, waist-hip ratio, general practitioner and
medical specialist consultations, and hospital admissions, the findings of the Dunedin
Study were not significantly different to the nationally representative surveys. The
Dunedin Study members also did not differ from their national counterparts on SF-36
subscales measuring physical functioning, bodily pain, general health, vitality, and
mental health. They had better scores on the three interference subscales of the SF-36
compared to the national sample, and men in the Dunedin Study spent a little more
time doing vigorous physical activity.
Discussion For most outcomes, the Dunedin Study members were very similar to the
nationally representative samples. There was little evidence that the repeated
assessments in the Dunedin Study had significantly altered the Study members’
health, either in terms of responses to questionnaires or on physiological measures of
health status. Findings from the Dunedin Study are likely to be generalisable to most
young New Zealanders. However, the Dunedin Study is under-representative of Māori
and Pacific peoples, so these findings need to be interpreted with caution in this
context. Implications for the proposed national Longitudinal Study of New Zealand
Children and Families are discussed.
The Dunedin Multidisciplinary Health and Development Study (“Dunedin Study”) is
a long-running cohort study of 1037 children born in Dunedin in 1972–1973. Over the
past 30 years, the study has generated more than 900 publications and reports and it is
regarded as one of the most important sources of information on the health,
development, and behaviour of young people.
1
However, concerns are sometimes raised about whether the Dunedin Study cohort is
truly representative of young New Zealanders and, more importantly, whether
findings from the cohort can be generalised to other populations of young people in
New Zealand. This report seeks to address some of these concerns by comparing the
NZMJ 2 June 2006, Vol 119 No 1235 Page 46 of 145
URL: http://www.nzma.org.nz/journal/119-1235/2002/ © NZMA
health status of the Dunedin Study members with those participating in nationally
representative surveys.
Two issues arise. First, are the Study members, who were all born in Dunedin, similar
to other New Zealanders of the same age? Second, have the health behaviours of
Dunedin Study members changed (due to being intensively studied throughout their
lives) to the point where they are no longer representative of the original population
from which they were drawn (the so-called “Hawthorne effect”
2
). These are not trivial
matters. Despite information to the contrary,
3
misperceptions about Dunedin Study
sample persist and at times they raise questions about the value of the Dunedin Study
data for policy-making in the New Zealand context.
The Dunedin Study members are now 32 years old, and they are undergoing a further
assessment as we prepare to study the positive and problematic aspects of the
transition from young adulthood to mid-life. This represents an opportune time to
revisit the question of whether the findings from the Dunedin Study are generalisable
to other New Zealanders.
In addition, there is another reason to do this review now as the New Zealand
Ministries of Social Development, Health, and Education as well as Treasury and The
Families Commission are planning to embark upon a national Longitudinal Study of
New Zealand Children and Families.
4
Because multi-site studies tend to be more
costly, logistically-demanding, and risk greater threats to internal validity (e.g.
standardisation of procedures) than single (or perhaps two) site studies, knowledge
about the generalisability of findings from regionally-based studies like the Dunedin
Study may help to plan the optimal sampling strategy for the National cohort study.
To address these questions about generalisability, we directly compared the Dunedin
Study members from their most recently completed assessment in 1998–1999 (when
they were all aged age 26) to 25 and 26 year-olds participants in the cross-sectional
New Zealand Health Survey in 1996/97
5
and the National Nutrition Survey in 1997.
6
Comparisons were conducted wherever the same or very similar data were collected
in the Dunedin Study and the national surveys.
Methods
Sample characteristics
Dunedin Study—This analysis involved 499 male and 481 female members who participated in the
Dunedin Study assessment at age 26 years (mean age = 26.0 years, SD = 3 months). The background to
the study and Study members are described in detail elsewhere.
1
Briefly, the Dunedin Study is a longitudinal investigation of the health, development, and behaviour of
1037 children born in Queen Mary Maternity Hospital, Dunedin between April 1972 and March 1973.
The sample has been assessed with a diverse array of medical, psychological, and sociological
measures with high rates of participation at age 3 (n = 1037), age 5 (n = 991), age 7 (n = 954), age 9 (n
= 955), age 11 (n = 925), age 13 (n = 850), age 15 (n = 976), age 18 (n = 993), age 21 (n = 992), and
age 26 (n = 980, 96% of the living cohort). Seventy-three (7.5%) Study members self-identified as
Māori and 15 (1.5%) as Pacific people at age 26.
The age-26 assessments took place at the Dunedin Unit between March 1998 and June 1999. A small
number (27/980, 3%) of participants who were unable to attend the Unit were assessed in the field. The
assessment took a full day lasting from 8.30am to 5.15pm and involved interviews and physical
examinations.
Of those who participated at the age-26 assessment, 41% (404) were still resident in Dunedin at the
time of interview, 21% (202) were resident in other parts of the South Island and 17% (168) were
NZMJ 2 June 2006, Vol 119 No 1235 Page 47 of 145
URL: http://www.nzma.org.nz/journal/119-1235/2002/ © NZMA
resident in the North Island. Hence, 774 (79%) were resident in New Zealand at the time of interview.
Of the remainder, 11% (108) were resident in Australia, 7% (66) were resident in the United Kingdom,
and 3% (32) were resident elsewhere.
New Zealand Health Survey (“Health Survey”)—The 1996/1997 Health Survey used a clustered
stratified design based on geographic areas to obtain a sample with characteristics that were
representative of the entire New Zealand civilian population. To obtain more reliable estimates for
Māori and Pacific peoples, a proportionately greater sample of these ethnic groups was included. A
total of 7862 adults (aged 15 years and over) participated, thus representing a 73.8% response rate. This
analysis included the 292 respondents who were aged 25 or 26 at the time of the survey. Of these, 64
(21.9%) identified themselves as Māori and 34 (11.6%) as Pacific people.
National Nutrition Survey (Nutrition Survey)—At the conclusion of the Health Survey, participants
were asked if they would undergo further assessment for the 1997 National Nutrition Survey. A total of
4636 adults completed the Nutrition Survey, of which 146 aged 25 or 26 years are included in this
analysis. Twenty-eight (19.2%) identified themselves as Māori and 12 (8.2%) as Pacific people.
Comparison measures
Self-reported health status—For both the Dunedin Study and Health Survey samples, self-reported
health status during the previous 12 months was measured by the Australian/New Zealand adaptation
of the SF-36 survey—a 36-item questionnaire measuring eight aspects of health.
7
These included
physical functioning, role physical (the impact of physical health on performance of everyday roles),
bodily pain, general health, vitality, social functioning, role emotional (the impact of emotional health
on performance of everyday roles), and mental health. This instrument has been shown to be a reliable
and valid measure of the health status of New Zealanders.
8
Body size measurements—The Dunedin Study and Nutrition Survey measured height without shoes
and weight in light clothing to calculate body mass index (BMI) in kg/m
2
. Waist and hip circumference
were measured to calculate the waist:hip ratio—an index of central adiposity. All body size
measurements in both the Dunedin Study and Nutrition Survey were taken twice.
Physical activity—Participants in the Dunedin Study were asked if they had done any physical
activities that caused them to “breathe hard or puff a lot” in the past 4 weeks and, if so, how much time
per week they spent doing these activities in a normal week. This was taken as the time spent per week
doing vigorous physical activities. Vigorous physical activity in the Health Survey was taken as the
time they reported that they had spent doing physical activities in the past 7 days that had made them
“breathe hard or sweat”. This question was prompted by a list of likely activities.
Smoking status—In the Dunedin Study, those who currently smoked one or more cigarettes per day
and had smoked daily for at least 1 month in the last year were deemed to be current smokers. Study
members who were not current smokers but had smoked daily for as long as a year at some time in
their lives were deemed to be ex-smokers. In the Health Survey, those who reported that they smoked
one or more cigarettes daily were to be deemed current smokers. Ex-smokers were those who had
smoked in the past but were not current smokers.
Health service utilisation—Dunedin Study members were asked whether and how many times they
had used a general practitioner (GP), or a medical specialist (e.g., cardiologist, gastroenterologist,
obstetrician/gynaecologist, urologist, orthopaedic surgeon, nephrologist, dermatologist, neurologist,
ear, nose & throat specialist, ophthalmologist, respiratory specialist, oncologist, endocrinologist,
rheumatologist) in the past year.
Study members were also asked whether they had spent any time in hospital in the past year for a
physical health (not mental health) problem. Participants in the NZ Health Survey were asked how
many times they had visited a general practitioner in the past year, and were also asked if they had seen
a medical specialist but were not prompted by a list of possible specialists.
Statistical methods—Comparisons were conducted between the Dunedin Study members and 25–26
year old participants in the Health and Nutrition Surveys. For all measures, either means (e.g. SF-36
scale scores) or prevalences (e.g. current smokers) are presented together with 95% confidence
intervals (CIs).
Sample survey weights were applied for the Health Survey based on each individual’s probability of
being selected for the survey to provide estimates consistent with the New Zealand population. The
Dunedin Study was considered to be significantly different from either of the national samples on a
measure if the 95% CIs of the samples did not overlap.
NZMJ 2 June 2006, Vol 119 No 1235 Page 48 of 145
URL: http://www.nzma.org.nz/journal/119-1235/2002/ © NZMA
Results
Comparisons between the Dunedin Study and the Health and Nutrition Surveys are
shown in the following Tables. Because not all participants in the studies consented to
every assessment, the numbers included in the tables vary slightly.
Self-reported health status—There were no significant differences between the
Dunedin Study and the Health Survey on SF36 subscales measuring physical
functioning; bodily pain; general health; vitality; and mental health (Table 1). On the
subscales measuring interference with physical and emotional task roles, members of
the Dunedin Study scored better than their Health Survey counterparts. They also
reported higher social functioning scores, indicating that they experienced less
interference in social activities as a result of a physical or emotional problem.
Body size measures—The Dunedin Study and the Nutrition Survey participants were
very similar on measures of Body Mass Index and waist:hip ratio (Table 2). These
measures were also similar if the comparison was restricted to Māori [Dunedin Study:
mean (95%CI) BMI 25.5 (24.6–26.6), mean waist:hip ratio 0.799 (0.784–81.5);
Nutrition Survey: mean BMI 27.5 (23.7–31.3), mean waist:hip ratio 0.808 (0.756–
0.861)]
Physical activity—Overall, there were no significant differences in the time spent
doing vigorous activity in the participants in the Dunedin Study and the Health
Survey (Table 3). However, more men in the Dunedin Study spent more than 300
minutes per week doing vigorous activity and fewer of the Dunedin Study men did no
vigorous activity.
Smoking status—The Dunedin Study had a slightly greater proportion of current
smokers (37.1% vs 33.0%), and a slightly lower proportion of ex-smokers (11.5% vs
16.6%) than the Health Survey, although neither of these differences were significant
(Table 4). Approximately half of both samples had never smoked.
Health service use—Similar proportions of the Dunedin Study and the Health Survey
had used a GP in the previous 12 months (78.6% and 76.5%, respectively) and were
admitted as an inpatient in the previous 12 months (9.7% and 7.8%, respectively)
(Table 5). A slightly, though not significantly, greater proportion of the Health Survey
participants had used a medical specialist, as compared to the Dunedin Study (29.2%
and 20.5%, respectively). There were no differences between the samples in terms of
frequency of GP use.
NZMJ 2 June 2006, Vol 119 No 1235 Page 49 of 145
URL: http://www.nzma.org.nz/journal/119-1235/2002/ © NZMA
Table 1. SF-36 Health Survey results. The mean and 95% confidence intervals are presented for men and women in the Dunedin Study
at age 26 and 25-26 year-olds in the New Zealand Health Survey for each of the eight subscales. Higher scores represent better health.
Significant differences between the Dunedin Study and Health Survey are highlighted in bold.
Dunedin Study New Zealand Health Survey
SF36 scales Male (n=499) Female (n=480) All (n=979) Male (n=98) Female (n=194) All (n=292)
Physical functioning
Role physical
Bodily pain
General health
Vitality
Social functioning
Role emotional
Mental health
94.4 (93.5–95.4)
91.5 (89.3–93.6)
80.1 (78.4–81.9)
77.5 (76.1–78.9)
68.8 (67.5–70.1)
90.8 (89.4–92.1)
93.5 (91.6–95.3)
80.4 (79.3–81.5)
91.1 (89.9–92.3)
87.3 (84.7–89.9)
77.7 (75.8–79.6)
77.2 (75.7–78.8)
62.0 (60.5–63.6)
87.7 (86.1–89.3)
89.1 (86.8–91.4)
77.0 (75.7–78.4)
92.8 (92.0–93.6)
89.4 (87.7–91.1)
78.9 (77.6–80.2)
77.4 (76.3–78.4)
65.5 (64.5–66.5)
89.3 (88.2–90.3)
91.3 (89.8–92.8)
78.8 (77.9–79.6)
89.9 (85.5–94.3)
79.6 (65.8–93.4)
79.0 (72.0–86.1)
75.0 (69.0–80.9)
68.5 (64.2–72.7)
88.0 (83.3–92.6)
85.6 (78.3–92.9)
78.6 (74.7–82.5)
92.4 (89.8–94.9)
80.0 (71.9–88.1)
78.7 (73.6–83.7)
75.2 (71.8–78.7)
61.2 (57.7–64.7)
81.9 (77.0–86.9)
71.6 (61.7–81.5)
74.2 (71.0–77.4)
91.2 (88.8–93.7)
79.8 (72.2–87.3)
78.8 (74.6–82.9)
75.1 (71.9–78.3)
64.6 (61.9–67.2)
84.7 (81.3–88.2)
78.1 (71.6–84.7)
76.3 (73.8–78.8)
Table 2. Body Mass Index (BMI) and waist:hip ratio means (95%CI) for the Dunedin Study members at age 26, and for 25 & 26 year
olds from the National Nutrition Survey. Data from pregnant women are excluded.
Dunedin Study National Nutrition Survey
Variable Male Female All Male Female All
BMI 25.2 (24.8–25.5)
n=494
24.9 (24.4–25.4)
n=445
25.0 (24.8–25.3)
n=939
25.7 (23.9–27.4)
n=49
25.0 (23.9–26.1)
n=97
25.3 (24.3–26.3)
n=146
Waist:hip ratio 0.849 (0.846–0.853)
n=489
0.745 (0.741–0.750)
n=438
0.800 (0.796–0.805)
n=927
0.860 (0.832–0.888)
n=47
0.761 (0.740–0.781)
n=95
0.806 (0.783–0.829)
n=142
NZMJ 2 June 2006, Vol 119 No 1235 Page 50 of 145
URL: http://www.nzma.org.nz/journal/119-1235/2002/ © NZMA
Table 3. Time spent in vigorous activity during a typical week by the Dunedin Study members at age 26 and the 25–26 year olds in the
New Zealand Health Survey (percentage and 95%CI of sample). Significant differences between the Dunedin Study and Health Survey
are highlighted in bold.
Dunedin Study New Zealand Health Survey
Male (N=496) Female (N=476) All (N=972) Male (N=98) Female (N=194) All (N=292)
Time interval
n % n % n % n % n % n %
0 mins 138 27.8
(23.9–32.0) 183 38.4
(34.1–42.9) 321 33.0
(30.1–36.1) 42 47.7
(34.7–60.7) 80 32.9
(24.4–41.4) 122 39.9
(31.9–47.9)
<150 mins 95 19.2
(15.8–22.9) 109 22.9
(19.2–26.9) 204 21.0
(18.4–23.7) 14 12.6
(5.2–20.0) 54 29.8
(20.5–39.2) 68 21.7
(15.4–27.9)
150–300 mins 110 22.2
(18.6–26.1) 128 26.9
(23.0–31.1) 238 24.5
(21.8–27.3) 19 21.5
(10.6–32.5) 30 19.8
(9.8–29.8) 49 20.6
(13.3–28.0)
>300 mins 153 30.8
(26.8–35.1) 56 11.8
(9.0–15.0) 209 21.5
(19.0–24.2) 23 18.2
(10.0–26.3) 30 17.5
(9.8–25.2) 53 17.8
(12.2–23.4)
Table 4. Smoking status (percentage and 95% CI) of the Dunedin Study members at age 26 and the 25-26 year olds in the New Zealand
Health Survey sample.
Dunedin Study New Zealand Health Survey
Male (N=499) Female (N=481) All (N=980) Male (N=98) Female (N=194) All (N=292)
Smoking status
n % n % n % n % n % n %
Never smoked 266 53.3
(48.8–57.8) 241 50.1
(45.5–54.7) 507 51.7
(48.6–54.9) 47 51.9
(38.3–65.5) 81 49.0
(38.3–59.7) 128 50.4
(41.8–58.9)
Ex-smoker 54 10.8
(8.2–13.9) 64 13.3
(10.4–16.7) 118 12.0
(10.0–14.2) 15 13.1
(5.8–16.9) 31 19.7
(10.6–28.8) 46 16.6
(10.5–22.7)
Current smoker 179 35.9
(31.7–40.3) 176 36.6
(32.3–41.1) 355 36.2
(33.2–39.3) 36 35.0
(21.5–48.5) 82 31.3
(22.2–40.4) 118 33.0
(25.0–41.0)
NZMJ 2 June 2006, Vol 119 No 1235 Page 51 of 145
URL: http://www.nzma.org.nz/journal/119-1235/2002/ © NZMA
Table 5. Twelve-month health service use (percentage and 95% CI of sample) by the Dunedin Study members at age 26 and the 25–26
year olds in the New Zealand Health Survey
Dunedin Study New Zealand Health Survey
Male (N=498) Female (N=479) All (N=977) Male (N=98) Female (N=194) All (N=292)
Health service use
n % n % n % n % n % n %
Used GP 331 66.5
(62.1–70.6) 429 89.6
(86.5–92.3) 760 77.8
(75.2–80.4) 64 66.9
(54.9–78.9) 163 85.1
(77.2–92.9) 227 76.5
(69.5–83.5)
Used specialist 73 14.7
(11.7–18.1) 122 25.5
(21.6–29.6) 195 20.0
(17.5–22.5) 25 29.9
(16.7–43.0) 51* 28.5
(18.8–38.2) 76* 29.2
(21.1–37.2)
Admitted as inpatient 39 7.8
(5.6–10.6) 55 11.5
(8.8–14.8) 94 9.6
(7.8–11.4) 9 9.0
(5.2–12.7) 29 10.1
(5.6–14.5) 38 7.8
(1.5–14.0)
Male (N=331) Female (N=429) All (N=760) Male (N=62) * Female (N=163) All (N=225) *
n % n % n % n % n % n %
Used GP once 136 41.1
(35.7–46.4) 66 15.4
(12.1–19.2) 202 26.6
(23.5–29.9) 16 33.4
(15.2–51.7) 23 11.7
(6.3–17.0) 39 20.5
(11.6–29.4)
Used GP twice 74 22.4
(18.0–27.2) 104 24.2
(20.3–28.6) 178 23.4
(20.5–26.6) 18 29.9
(15.4–44.4) 37 24.0
(15.2–32.9) 55 26.4
(18.5–34.2)
Used GP 3–5 times 81 24.5
(19.9–29.5) 149 34.7
(30.2–39.4) 230 30.3
(27.0–33.7) 17 24.3
(10.8–37.8) 63 41.0
(29.3–52.7) 80 34.3
(25.1–43.4)
Used GP 6–11 times 27 8.2
(5.4–11.6) 72 16.8
(13.4–20.7) 99 13.0
(10.7–15.6) 7 9.6
(1.9–17.4) 22 14.0
(5.5–22.5) 29 12.2
(6.2–18.2)
Used GP >11 times 13 3.9
(2.1–6.6) 38 8.9
(6.3–12.0) 51 6.7
(5.0–8.7) 4 2.7
(0.0–5.9) 18 9.3
(3.5–15.1) 22 6.6
(2.9–10.3)
* Data from two members of the Health Survey sample not available
NZMJ 2 June 2006, Vol 119 No 1235 Page 52 of 145
URL: http://www.nzma.org.nz/journal/119-1235/2002/ © NZMA
Discussion
Dunedin Study members were similar to their age matched peers in the national
Health and Nutrition samples on most of the health measures we compared. This
included five of the eight subscales of self-reported health status from the SF36;
smoking behaviour; physical activity; two physical measurements (BMI and waist:hip
ratio); and use of general practice and specialist health services.
There were significant differences between the Dunedin Study members and the
nationally representative samples on three of the eight subscales of the SF36. The
SF36 is a widely-used, validated, and reliable instrument that provides a
multidimensional assessment of health. The 8 subscales measure physical, emotional,
and social factors and the SF36 is used to provide a reasonable overall assessment of a
person’s health in the context of large Health Surveys. However, despite its
usefulness, the SF36 remains a self-report measure and has the accompanying
limitations. For example, reports can be confounded by mood or certain personality
traits.
9
The three subscales on which the Dunedin Study members differed from their
peers in the Health Survey were “role physical”, “social functioning”, and “role
emotional” (only differed in women).
Interestingly, the Dunedin Study members tended to score higher (better health) on
these scales than the participants in the Health Survey. If the repeated interviews of
the Dunedin cohort had altered their perception of their health (the “Hawthorne
effect”), we might have predicted that they would become more sensitised to their
health problems. In fact, the Dunedin Study members reported less interference in
their roles than participants in the national studies. It is possible that these minor
differences arose because of a selection bias of more health-focussed individuals
among the 74% of people who agreed to participate in the Health Survey.
For some measures, slightly different methodologies were used by the Dunedin Study
and the national surveys. For example, members of the Dunedin sample were
considered smokers if they had smoked one cigarette a day for at least a month of the
previous year AND they currently smoked at least one cigarette a day—whereas
members of the Health Survey were only asked if they currently smoked at least one
cigarette a day (they need not have smoked for a month). Also, members of the Health
Survey weren’t required to have smoked for a year to be considered ex-smokers.
Despite this, the proportions of smokers and ex-smokers in the samples were quite
similar. For the health service utilisation measures, the methods used by both studies
seem comparable except that members of the Dunedin sample were given a checklist
of specialists who they may have visited in the previous year—whereas members of
the Health Survey were not. Similarly the physical activity measures were similar,
although worded slightly differently and the Health Survey respondents were shown a
checklist of activities. Taken together these measures indicate that the samples are
broadly comparable in terms of health-risk behaviours and lifestyle factors (smoking
and physical activity) and health problems indexed by health service use. It is notable
that on the comparison of the two objective physical health markers (body mass index
and waist hip ratio), the Dunedin Study members and the National Nutrition sample
were almost identical.
Have we changed people? It would appear not. Participants in the Dunedin Study look
the same as research-naïve participants in the national studies in almost all respects.
NZMJ 2 June 2006, Vol 119 No 1235 Page 53 of 145
URL: http://www.nzma.org.nz/journal/119-1235/2002/ © NZMA
These findings are consistent with an earlier study comparing respiratory symptoms
among Dunedin Study participants at age 21 with those of 20–22-year-old participants
in the New Zealand section of the European Community Respiratory Health Study.
10
This was a once-only postal questionnaire conducted in Auckland, Wellington,
Christchurch, and Hawke’s Bay which used virtually identical questions to those used
in the Dunedin Study. There was no difference in the prevalence rates of any of the
reported symptoms or asthma medication-use between the samples. Thus, in two
comparisons we have found little evidence that the health status of our study members
has been altered by virtue of their involvement in a longitudinal study.
These finding have several implications for planning of future cohort studies,
including the New Zealand Longitudinal Study of Children and Families. The first
relates to the need to distinguish between ‘representativeness’ and ‘generalisability’.
Classically, representativeness refers to sampling methods that faithfully represent all
members of the target population (in a New Zealand nationally representative study
this would mean the whole of the country), whereas generalisability refers to the
ability to extrapolate findings to the wider population, despite imperfect
representativeness. Deriving a sample that is perfectly representative of the major
population groups of interest (in terms of socioeconomic status and geographic
location for example) is resource intensive and costly. Moreover, for a longitudinal
study, generalisablility is more important than representativeness. By its nature, a
cohort study cannot remain truly representative of the population of interest. Thus,
although the Dunedin Study sample appears to be broadly representative of New
Zealand children born in 1972/1973, they will not necessarily be representative of
children born in 1992 or 2002. Nevertheless, it is today’s New Zealand children that
are most likely to benefit from the lessons that we have learned from the Dunedin
Study. The value of a cohort study is that it provides a means of testing hypotheses
about the importance of early influences and the sequence of events in growth and
development. Unless there are good reasons to suspect otherwise, the findings are
likely to be generalisable to other people in similar circumstances.
Second, a potentially greater threat to study validity is from non-random loss to
follow-up. In the context of a new national cohort study, a strong argument can be
made for resources being spent on maintaining cohort retention, and ensuring high
quality measurements, especially if the generalisability of findings to the wider
population from a single site can be demonstrated, as appears to be the case here. In
support of this argument, Youth 2000, a nationwide survey of health and wellbeing
amongst New Zealand secondary school students has been analysed by region.
Although there were minor differences between the 15 regions, the conclusions drawn
about health, risk behaviour, and health service needs for each region were identical.
11
The findings provide broad support for the generalisability of findings from the
Dunedin Study (and by implication similar studies such as the Christchurch Health &
Development Study
12
) to other New Zealanders. However, the cross-sectional
comparisons presented here risk underplaying the ways in which longitudinal designs
enhance generalisability compared to cross-sectional studies. For example,
prospective-longitudinal studies provide better estimates of lifetime exposures than
cross-sectional, retrospective studies.
13
Longitudinal studies also permit casual inferences about a range of exposures and
outcomes, and it is this information that is most useful for policy-making. In this
NZMJ 2 June 2006, Vol 119 No 1235 Page 54 of 145
URL: http://www.nzma.org.nz/journal/119-1235/2002/ © NZMA
regard, it is noteworthy that both the Dunedin and Christchurch longitudinal studies
produce highly replicable findings in the international context, which given the
similarities between countries such as Australia, USA, Canada, and the UK
14
is
perhaps not surprising. Indeed it is precisely this generalisability that has resulted in
significant investment in the Dunedin Study by the U.S. National Institutes of Health,
and more recently by the UK Medical Research Council.
However, there are some limitations to our findings, particularly in the capacity of
this analysis to inform on the generalisability of findings to specific ethnic groups
(Māori, Pacific, and European/Other). We have not been able to examine whether
health outcomes within specific ethnic groups (Māori, Pacific, European/Other) are
comparable between the Dunedin Study and the national surveys. Summary data for
the individual ethnic groups in the New Zealand Health Survey for this age-group
were not available. There may also be limitations due to small numbers of Māori and
Pacific ethnic groups in both the Dunedin Study and in the national surveys, which
may reduce the precision of estimates making statistical comparison difficult.
Finally, Māori, and Pacific people are under-represented in the Dunedin Study when
compared with the National Surveys (where data is weighted to match the census
population). Given differences in health status between ethnic groups, this may
impact on comparisons between the total Dunedin cohort and the New Zealand
Survey populations. For these reasons we need to be cautious about concluding that
the Dunedin cohort findings are able to be generalised on the basis of ethnicity and
this issue needs to be investigated further. Nevertheless, with regards to Māori, it is
noted that 73 Dunedin Study members self-identified as Māori at age 26. By
comparison, there were 64 Māori participants in the 25 and 26 year-old age range in
the New Zealand Health Survey. Hence, although the Dunedin Study may under-
represent Māori as a proportion of its total sample, it actually has a larger number of
Māori participants of this age than the nationally representative sample.
In conclusion, there appear to be few important differences in self-reported and
objectively measured health between participants in the long-running Dunedin Study,
and participants of similar ages in nationally-representative surveys. This suggests
that the Dunedin Study members have not been changed by undergoing repeated
assessments throughout their lives, and that findings from the Dunedin Study are
likely to be broadly generalisable to the wider New Zealand population. These
findings may be relevant to the design of future New Zealand cohort studies.
Author information: Richie Poulton, Director, Robert J Hancox, Deputy Director,
Dunedin Multidisciplinary Health and Development Research Unit, Department of
Preventive and Social Medicine, Dunedin School of Medicine, University of Otago,
Dunedin; Barry J Milne, Data Manager, Social, Genetic and Developmental
Psychiatry Centre, Institute of Psychiatry Kings College, London, UK; Joanne M
Baxter, Senior Lecturer; Department of Preventive and Social Medicine, Dunedin
School of Medicine, University of Otago, Dunedin; Kate Scott, Senior Lecturer and
Senior Clinical Psychologist, Department of Psychological Medicine, Wellington
School of Medicine and Health Sciences, Wellington South; Noela Wilson, Director,
Life in New Zealand Activity and Health Research Unit, University of Otago,
Dunedin
NZMJ 2 June 2006, Vol 119 No 1235 Page 55 of 145
URL: http://www.nzma.org.nz/journal/119-1235/2002/ © NZMA
Acknowledgements: The Dunedin Multidisciplinary Health and Development
Research Unit is funded by the Health Research Council of New Zealand. Collection
of data used in this report was also funded by the National Heart Foundation (NZ) and
the US National Institute of Mental Health grant MH45070. The New Zealand Health
Survey and National Nutrition Survey were funded by the Ministry of Health.
We are grateful to the Dunedin Study members and their parents for their continued
support and to the participants in the New Zealand Health and National Nutrition
Surveys. We thank Karen Blakey for help in compiling data from the National Health
Survey, and Professor David Fergusson for helpful comments. We also wish to thank
Dr Phil A Silva, the Dunedin Study founder.
Correspondence: Richie Poulton, Dunedin Multidisciplinary Health and
Development Research Unit, Department of Preventive and Social Medicine,
University of Otago, P O Box 913, Dunedin. Fax: (03) 479 5487; email:
richie.poulton@otago.ac.nz
References:
1. Silva PA, Stanton WR, eds. From child to adult: The Dunedin Multidisciplinary Health and
Development Study. Auckland: Oxford University Press; 1996.
2. Gale EA. The Hawthorne studies–a fable for our times? QJM. 2004;97:439–49.
3. Dunedin Multidisciplinary Health and Development Study [webpage]. Available online.
URL http://dunedinstudy.otago.ac.nz/ Accessed May 2006.
4. Health Research Council of New Zealand. Socio-Economic Determinants of Health Joint
Research Portfolio [webpage]. Available online. URL:
http://www.hrc.govt.nz/root/pages_policy/Socio-
Economic_Determinants_of_Health_Joint_Research_Portfolio.html Accessed May 2006.
5. Taking the Pulse – the 1996/97 New Zealand Health Survey. Wellington: Ministry of
Health; 1999.
6. Russell DG, Parnell WR, Wilson NC, and the principal investigators of the 1997 National
Nutrition Survey. NZ Food: NZ People. Key results of the 1997 National Nutrition Survey.
Wellington: Ministry of Health; 1999.
7. Ware JE Jr, Sherbourne CD. The MOS 36-item short-form Health Survey (SF-36). I.
Conceptual framework and item selection. Med Care. 1992;30:473–83.
8. Scott KM, Tobias MI, Sarfati D, Haslett SJ. SF-36 Health Survey reliability, validity and
norms for New Zealand. Aust N Z J Public Health. 1999;23:401–6.
9. Watson D, Pennebaker JW. Health complaints, stress, and distress – exploring the central
role of negative affectivity. Psychol Rev. 1989;96:234–54.
10. Sears MR, Lewis S, Herbison GP, et al. Comparison of reported prevalences of recent
asthma in longitudinal and cross-sectional studies. Eur Respir J. 1997;10:51–4.
11. Youth 2000: A national secondary school youth health survey [webpage]. Available online.
URL: http://www.youth2000.ac.nz/ Accessed May 2006.
12. Christchurch Health and Development Study [webpage]. Available online. URL:
http://www.chmeds.ac.nz/research/chds/index.htm Accessed May 2006.
13. Andrews G, Poulton R, Skoog I. Lifetime risk of depression: restricted to a minority or
waiting for most? Br J Psychiatry. 2005;187:495-6.
14. Moffitt TE, Caspi A, Rutter M, Silva P. Sex differences in antisocial behaviour: conduct
disorder, delinquency, and violence in the Dunedin longitudinal study. Cambridge:
Cambridge University Press; 2001.
