• Open Access

The beneficial effects of preschool attendance on adult cardiovascular disease risk


Correspondence to: Dr Katina D'Onise, Sansom Institute for Health Research, University of South Australia, City East Campus, North Terrace, Adelaide, South Australia 5000; e-mail: katina.d'onise@postgrads.unisa.edu.au


Objective: To assess the effect of South Australian Kindergarten Union participation on adult cardiovascular behavioural risk factors.

Methods: Using a retrospective cohort design, this study examined the effect of attendance at a Kindergarten Union preschool from 1940 to 1972 on behavioural risk factors for cardiovascular disease in adults 34–67 years. Dichotomous outcomes were analysed using a generalised linear model (Poisson distribution) with robust variance estimates. Outcomes with more than two categories were analysed with a multinomial logistic model.

Results: There was a beneficial effect of preschool on high physical activity relative to sedentary and on ever smoking, but a negative effect on fruit consumption. Preschool attendance was not associated with alcohol risk or vegetable consumption under traditional criteria, however the point estimate for vegetable consumption was in the beneficial direction. The point estimates from the multinomial model suggested a step-wise decreasing risk for preschool attendees to have less risk of experiencing multiple behavioural risk factors (e.g. risk of five risk factors for preschool participants compared with non-participants).

Conclusions and implications: Attendance at a Kindergarten Union preschool was associated with a reduced risk of two and an indication of benefit in a third behavioural risk factor in adulthood. This study provides some evidence for the potential health benefit of interventions outside of the health sector to prevent cardiovascular diseases, which are strongly associated with lifelong social disadvantage.

Cardiovascular disease (CVD), which is responsible for 17% of the total burden of disease in Australia, remains a challenge for population health despite the many recent advances in prevention, identification and management of established disease.1 Social factors continue to play an important role in the aetiology of CVD, with social disadvantage from early infancy (and possibly from in utero or earlier) through to adulthood potentially having an important influence on known cardiovascular risk factors and CVD.2–4 While a proportion of the excess CVD risk for low socioeconomic position (SEP) groups is unexplained, there is evidence that the well known lifestyle related risk factors tend to co-occur more frequently in people who are socially disadvantaged.5–7

There is some evidence that the social, physical and biological environment in childhood influences the development of CVD, which points to the potential for interventions in childhood having beneficial effects on CVD in adulthood.8 Early childhood educational interventions (ECDIs) which involve a combination of educational, health and social services for children in addition to parenting programs or home visiting, are thought to enhance child development.9,10 The success of these interventions is thought to be due in part to being set in early childhood, which is thought to be a sensitive period in development. The benefits attained in childhood provide a turning point that sets the children on positive social trajectories either through a pathway or chains of risk model.11,12 Through this improved socioeconomic trajectory, ECDIs may in turn reduce the prevalence of various lifestyle related risk factors for CVD and possibly also reduce the probability of clustering of these risk factors. Some evidence for this potential of preschool programs comes from a small number of studies conducted in the United States,13–15 with indications of benefits into adulthood (up to age 40) on smoking and physical activity, but no benefits on fruit and vegetable consumption or binge alcohol use. These studies are however limited by being mostly resource intensive interventions in highly socially disadvantaged US populations, with relatively crude measures of health outcomes and so may be of limited generalisability.

South Australian Kindergarten Union preschools

The Kindergarten Union (KU) managed preschools in SA from 1906 to 1985, and until the 1970s, the KU managed the vast majority of preschool services in SA.17 The preschools were initially established to enhance the social, emotional, physical and cognitive development of children who were living in poverty, with an emphasis on educational services.18 In the initial years of the KU, preschools were established in suburbs with high levels of poverty and were free for socially disadvantaged children, expanding to middle class suburbs by the 1940s. Attendance was primarily through geographic proximity to an existing centre.

The preschool program enrolled children between two and five years old, for half or full days, for up to five days a week. The program involved direct educational services for children, parenting services, home visiting and health screening and referral for specialist services when required.19

The KU preschools included a number of features of high quality. All preschool directors and teachers were required to have a recognised early childhood development qualification. Standards developed by the Australian Pre-School Association were adopted across all preschools, which included a child-staff ratio limit20 and standards for building and playground design.19 As the number of preschools increased, a ‘Pre-school Adviser’ was appointed in 1945 to assist the preschools to adhere to the standards and curriculum set by the KU.19

This study aimed to assess the effect of Kindergarten Union (KU) attendance in SA on single cardiovascular behavioural risk factors and their clustering. Investigating the potential for these interventions in Australia is timely given the renewed focus of federal and state governments on early childhood education, both in terms of quality and increased access for disadvantaged groups.21 Ethics approval for the study was granted by the University of South Australia Human Research Ethics Committee.



The North West Adelaide Health Study (NWAHS) is a longitudinal representative cohort study of adults over 17 years old, randomly selected from the northern and western metropolitan regions of Adelaide using the electronic telephone directory.22 Within each household, the person with the last birthday aged over 17 years was selected for interview. Exclusion criteria included not having the capacity to participate (intellectual, illness), living in a residential institution and being unable to communicate in English.

The sample was recruited from November 1999 to July 2003. The 4060 participants represented 49.4% of those who were eligible to participate. Data were collected by questionnaire, Computer Aided Telephone Interview (CATI) and clinic attendance in stage 1 (years 1999–2003) and stage 2 (2004), and a telephone follow up CATI was conducted in 2007 when details of preschool attendance were collected.

Study population

Figure 1 outlines the process for selecting the study population. Participants in the 2007 telephone follow up survey in the NWAHS (n=2996, 74% of baseline population) who lived in SA as children and were born during the years 1937–1969 were included in the study. Application of the inclusion criteria led to a reduction in sample size from 2,996 to 1,490, and “don't know” responses on preschool attendance reduced the sample size further to 1,395. Retired people were excluded from the income variable (163, explained below) and additional missing data (169) led to a final analytic sample of 1,063.

Figure 1.

Selection of study population.

Kindergarten Union attendance

Participants were asked in the 2007 telephone follow up study to recall if they had attended preschool (Did you attend kindergarten or preschool?), and the age at which they attended (How old were you when you first started kindergarten or preschool?). People who reported attending preschool at age five (n=147) and six (n=12) were re-categorised as not having attended preschool as school entry generally occurred by five years in SA and also to reflect the evidence that suggests that intervention before age five is important for long term effects. People who indicated that they did not know if they went to preschool were considered to be missing (n=95 after application of the inclusion criteria). This group was less likely to have gained a Bachelor's degree (prevalence ratio (PR) 0.44, 95% CI 0.18–1.06), and more likely to be in the lowest income category (PR 1.7, 95% CI 0.83–3.54) than the non-preschool group.

Behavioural risk factors

Physical activity questions were taken from the National Health Survey (NHS) collected in stage 2 (2004), which asked about the intensity, frequency and length of leisure time physical activity in the previous two weeks. Categories for sedentary, low, moderate and high exercise level were constructed as described in the NHS.23,24 Missing data on physical activity were replaced with data collected at stage 1 (9.2%). Alcohol intake was measured using self report of usual frequency of intake and usual number of standard drinks from stage 2 data collection with missing data replaced with data collected in stage 1 (0.1%). The different risk categories were based on the NHMRC recommendations in 200125 which was current at the time of the NWAHS data collection (Table 1). Low risk drinkers were the reference category given non-drinkers represent a potentially heterogeneous group of ex-drinkers and never-drinkers. Ever smoking was measured using a self report of ever having smoked and parental smoking was taken from report of smoking in a parent or guardian when the respondent was 4 years old. Fruit and vegetable intake collected at the 2007 interview were analysed to indicate the general quality of the diet, given a diet high in vegetables and fruit is thought to reduce the risk of CVD.26 Questions were taken from the National Nutrition Survey, on how many serves of either fruit or vegetable are usually eaten a day. These two questions were found to be a reliable indicator of the results obtained from the 24-hour recall.23

Table 1.  Alcohol risk levels and classification* (standard drinks).
Risk categoryMenWomen
Average per dayAmount per weekAverage per dayAmount per week
  1. *NHMRC 2001 guidelines, numbers indicate the upper limit of the category

Low risk drinker428214
Moderate risk drinker5–629–423–415–28
High risk drinker≥7≥43≥5≥29

To determine whether preschool attendance could reduce the total number of co-occurring risk factors, an index was created that summed each of the behavioural risk factors. A score of one was assigned for an alcohol intake of moderate or high risk, physical activity of sedentary or low, being an ever smoker, and less than two serves a day of fruit or less than five serves a day of vegetables, such that a score of five indicated high risk and zero low behavioural risk. Due to small numbers, a risk factor index of zero or one was collapsed into one low risk category.

Indicators of childhood socioeconomic position (SEP)

Childhood SEP was measured using report of father's main lifetime occupation27–29 (substituted for mother's main lifetime occupation if brought up in a maternal single parent household), coded as manual or non-manual,30 report of periods of at least six months of parental unemployment, or being brought up in a sole parent household. An index was created that summed these variables such that zero indicated no marker of disadvantage and three indicated a maximal marker of disadvantage (category two or three were collapsed to one category due to small numbers). Adult height which reflects aspects of the early nutritional and socioeconomic environment31 and has the advantage of being precisely measurable, was used as an additional indicator of childhood disadvantage.

Indicators of adult SEP

Education was categorised into four mutually exclusive categories (leaving school up to 15 years, leaving school after 15 years, attainment of a trade or diploma and attainment of a Bachelor's Degree or higher) using self reported educational attainment from stage 2. Three gross household income categories were constructed from the six collected in stage 2 (0–$40,000, $40,001–80,000, $80,001-over $100,000), excluding people who were retired.

Statistical analysis

Dichotomous outcome variables were analysed using a generalised linear model (Poisson distribution) with robust variance estimates, with resulting prevalence ratios (PR, prevalence of disease in exposed versus prevalence in unexposed) for the effect estimate. This model was chosen over a log binomial generalised linear model as the latter failed to iterate to a solution in many instances. The commonly used logistic regression model for dichotomous outcomes was not considered as the outcomes were relatively common (i.e., > 10–20%) and a measure approximating the relative risk was preferred to an odds ratio to enhance the interpretability of the results.32 Similarly, ordinal variables (physical activity, alcohol risk and the behavioural risk factor index) were analysed using multinomial logistic regression so that the outcome (a relative risk ratio, RRR) would approximate a relative risk estimate.

The association between preschool attendance and each of the outcomes was assessed in sequential regression models. Model 1 adjusted for age at stage 2 clinic follow up and gender, and model 2 further adjusted for child SEP and adult height, factors which may have influenced the chance of preschool participation. For ever smoking analyses, parental smoking was included in model 2. Model 3 further adjusted for educational attainment and adult income which were hypothesised to mediate the effect of preschool on adult health behaviours.


There were 476 people who reported attending preschool and 587 who did not attend preschool. The average age of preschool attendees was younger (45.3 years) than non-attendees (51.1 years, Table 2). There was a similar distribution of males and females across the comparison groups, however the preschool group came from a slightly more advantaged childhood SEP (40.3% compared with 32.4%). A higher proportion of preschool attendees had a Bachelor's degree and were in the higher income groups than the non-attendees.

Table 2.  Descriptive analysis of participant demographic characteristics in the North West Adelaide Health study, 1999–2007.
 Preschool attendeesNo preschool attendance
 n=476% or s.d.n=587% or s.d.
  1. s.d.- standard deviation

  2. *Child SEP: Number of markers of disadvantage from manual paternal occupation, 6+ months of parental unemployment, sole parent household

Age (years, s.d.)45.37.651.17.7
Female (%)26255.031854.2
Year of birth
1937–1949 (%)6914.520234.4
1950–1959 (%)15833.224541.7
1960–1969 (%)24952.314023.9
Child SEP*
0 (%)19240.319032.4
1 (%)26255.034358.4
2/3 (%)224.6549.2
Adult height (cm, s.d.)169.89.0168.89.5
Left school ≤15 years (%)367.69115.5
Left school >15 years (%)15733.019933.9
Trade/certificate/diploma (%)18538.922237.8
Bachelor's degree (%)9820.67512.8
0–$20 000 (%)4910.37813.3
$20,001–40,000 (%)9520.012821.8
$40,001–60,000 (%)11323.717429.6
$60,001–80,000 (%)9920.811018.7
$80,001–100,000 (%)5712.05910.1
>$100,000 (%)6313.2386.5

Multivariable analyses

Results for the multivariable analysis are presented in Table 3. There was an effect of preschool on physical activity, with a greater probability of being in any physical activity group relative to sedentary. The effect was greatest for being in the high physical activity group, which was the only category in which the 95% confidence interval (CI) did not include the null (PR 1.99, CI 1.19–3.35). Preschool attendance appeared to be associated with a reduced risk of ever smoking (PR 0.86, CI 0.77–0.97) in the fully adjusted model, but a negative effect on fruit consumption (PR 0.85, CI 0.73–0.99). The effect of preschool attendance on vegetable consumption was in the positive direction but the 95% confidence interval included the null (PR 1.41, CI 0.90–2.19). For all but the smoking outcome, addition of the child and adult SEP variables attenuated the association between preschool and the outcome slightly.

Table 3.  Multivariable analysis of the effect of Kindergarten Union preschool attendance on behavioural risk factors in the North West Adelaide Health study, 1999–2007.
 nModel 1Model 2Model 3
  Effect estimate95% CIEffect estimate95% CIEffect estimate95% CI
  1. Model 1: adjusted for age, gender. Model 2: model 1 + child SEP, adult height (for ever smoking also included parental smoking). Model 3: model 2 + educational attainment, adult income.

  2. Behavioural risk factor index: Score of one for each of an alcohol intake of moderate or high risk, physical activity of sedentary or low, being an ever smoker, and less than two serves a day or less than five serves a day of vegetables

  3. Effect estimates: RRR – relative risk ratio, PR – prevalence ratio, 95% CI – 95% confidence interval

  4. Missing data: physical activity 11, fruit consumption 1, vegetable consumption 3, smoking 23, alcohol 36, behavioural risk factor index 51.

Physical Activity n=1052 (RRR)
Sedentary2981.0 1.0 1.0 
Low physical activity3881.290.93–1.801.260.91–1.761.240.89–1.74
Moderate physical activity2731.370.96–1.961.320.92–1.901.260.87–1.81
High physical activity932.221.34–3.672.071.24–3.451.991.19–3.35
Fruit n=1,062 (PR)
<2 serves a day6211.0 1.0 1.0 
≥2 serves a day4410.880.75–1.020.860.74–1.010.850.73–0.99
Vegetable n=1060 (PR)
<5 serves a day9761.0 1.0 1.0 
≥5 serves a day841.510.96–2.371.460.93–2.291.410.90–2.19
Smoking n=1040 (PR)
Never smoker4641.0 1.0 1.0 
Ever smoker5760.850.76–0.950.860.76–0.970.860.77–0.97
Alcohol risk of harm n=1027 (RRR)
Low risk6791.0 1.0 1.0 
Moderate risk1670.850.59–1.230.830.57–1.210.870.60–1.27
High risk661.000.58–1.720.970.56–1.691.010.58–1.77
Behavioural risk factor index n=1012 (RRR)
0/1 risk factors1081.0 1.0 1.0 
2 risk factors2370.780.48–1.280.810.50–1.330.820.50–1.35
3 risk factors3540.750.47–1.200.780.49–1.250.830.52–1.34
4 risk factors2270.670.41–1.100.710.43–1.170.750.45–1.26
5 risk factors860.500.27–0.930.530.28–0.990.570.30–1.08

The effect sizes for alcohol drinking risk were small and low in precision. There was a null effect (small effect size with a wide CI) on being a moderate or high risk drinker compared with a low risk drinker and possibly a greater probability of being a non-drinker (PR 1.26, CI 0.81–1.96). Addition of adult SEP increased the magnitude of the effect estimates slightly across high risk and non-drinker categories.

While all of the 95% confidence intervals for the behavioural risk factor index crossed the null, the point estimates for the BRF index suggested an increased protective effect against an increasing number of risk factors for preschool attendees (e.g. risk of five risk factors PR 0.57, CI 0.30–1.08). Addition of the child and adult SEP in the models slightly attenuated the effect sizes.


This study found that preschool attendance resulted in a more favourable cardiovascular behavioural risk factor profile, for three of five examined risk factors examined individually but also in an index independent of age, gender, and SEP in childhood and adulthood. These results extend the evidence on the effects of preschool programs, finding some benefits into late adulthood of attendance at a multi-site, universal, community intervention in a country outside of the US. While the benefits seen were modest, it is noteworthy that such benefits were demonstrated over the follow up period spanning up to 65 years.

These findings are generally consistent with those from other studies. Comprehensive ECDIs have been shown to enhance exercise participation (measured dichotomously)13,33 and reduce the risk of ever smoking16,33,34 as was seen in participants of the KU. The results of this study, however, differ from those of the small randomised US studies Project CARE and Abecedarian that found no difference between participants and the control group in dietary factors (a ‘good’ diet was defined as consuming fruit and/or vegetables once or twice within the past 24 hours).13 This study found a negative effect of preschool attendance on fruit consumption but a suggestion of benefit for vegetable consumption, which was an unexpected finding. A possible explanation is that the SEP variables measured in this study did not predict fruit consumption but did predict vegetable consumption and so any SEP effect through preschool attendance may not have an effect on fruit consumption. It may also reflect error in the measurement and categorisation of fruit consumption. The results on alcohol consumption in this study did not follow a clear pattern of benefit or risk, which is in contrast to the Perry Preschool project and Abecedarian studies that both found preschool increased the risk of alcohol binge drinking, but measured differently than here.

Addition of adult SEP variables resulted in slight attenuation of all effect sizes (except for alcohol) suggesting that these variables mediated only a minor component of the association between preschool and the outcome assessed. This suggests either measurement error with these SEP factors or that they do not adequately index the mediating factors between preschool and health. For example, it may be that cognitive or non-cognitive factors not indexed by educational attainment and income may further explain the association between preschool and behavioural outcomes, given preschool programs are thought to improve long term outcomes mostly through cognitive gains34–37 and cognitive factors are thought to influence behavioural risk factors in adulthood independently of adult SEP.38

As a retrospective cohort study, there are a number of limitations that should be considered in the interpretation of the results. There is a potential for residual confounding by unmeasured and/or poorly measured background characteristics related to family environment, which may not have been indexed by the self-reported childhood factors measured in this study. Measurement error was possibly introduced by the use of adult recall of preschool attendance and self report of the behavioural outcomes, however this approach has been used in a number of studies33,39,40 and was found by one study to have reasonable validity.40 Furthermore, the results in this study are consistent with the small amount of evidence on preschool programs reported elsewhere suggesting reasonable validity of recall of preschool attendance. The self report of behavioural risk factors introduced measurement error, despite the use of reliable, validated self-reported assessment tools. While the effects observed in this study were generally in the positive direction they were unable to be estimated with great precision.

These results may not be generalisable to all people who attended a KU preschool. There are no historical records of KU attendance that would allow a comparison with the current sample. Further, the retrospective design has lead to exclusion of those KU attendees who did not remain in SA, with an unknown effect on the results and generalisability of the study findings. The combination of attrition (24%), missing data and selecting a sub-sample of age eligible participants may have introduced selection bias into the study which may also reduce generalisability of the results. However our purpose was to examine the associations between preschool attendance and behavioural risk factors, not estimate prevalence. Thus, it is not the case that selection processes, which operate in every cohort study, necessarily bias observed associations because the selection process would need to effect both the exposure factor and the outcome differentially by preschool attendance for bias to be introduced.41 Details regarding how the KU services changed over time and in each site are not available which also limits the ability to explore the precise mechanisms by which the KU may have had an effect on health outcomes. This is a limitation of any exposure such as education, which changes its content and meaning over time, however this is likely to be non-differential with respect to the outcomes examined here.

Under the assumption that the findings reported here are causal, the features of the preschool that are likely to have contributed to these findings are important to consider given the planned Australia-wide expansion of ECDIs. Most of the evidence regarding beneficial long term social outcomes is in favour of interventions that focus on high-quality service provision and that provide comprehensive services directly to children as well as their families9,42 both of which the KU preschools were able to achieve according to historical records. This KU study provides further evidence that high quality comprehensive services to children and their families that focus on optimal child development can also lead to health benefits.

In conclusion, attendance at a KU preschool was associated with modest effects on behavioural CVD risk factors in adulthood in the positive direction although generally with low precision. This study provides some evidence for the potential benefit of the health sector engaging in interventions outside of health services to prevent diseases such as CVD, which are strongly associated with lifelong social disadvantage. To this end, health professionals should collaborate in the planning, implementation and evaluation processes of the new federal government agenda for early childhood education to maximise the social and health gains from these interventions.


KD was supported by the National Health and Medical Research Council of Australia and the National Heart Foundation. JL was supported by the National Health and Medical Research Council of Australia. The authors would like to acknowledge the staff and participants of the North West Adelaide Health Study.