Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Area-Level and Individual-Level Factors for Teenage Motherhood: A Multilevel Analysis in Japan

  • Sachiko Baba ,

    baba@cir.med.osaka-u.ac.jp

    Affiliations Center for International Relations, Osaka University Graduate School of Medicine, Suita, Japan, Department of Social Medicine, National Research Institute for Child Health and Development, Tokyo, Japan

  • Hiroyasu Iso,

    Affiliation Public Health, Department of Social Medicine, Osaka University Graduate School of Medicine, Suita, Japan

  • Takeo Fujiwara

    Affiliations Department of Social Medicine, National Research Institute for Child Health and Development, Tokyo, Japan, Department of Global Health Promotion, Tokyo Medical and Dental University, Tokyo, Japan

Area-Level and Individual-Level Factors for Teenage Motherhood: A Multilevel Analysis in Japan

  • Sachiko Baba, 
  • Hiroyasu Iso, 
  • Takeo Fujiwara
PLOS
x

Abstract

Background

Teenage motherhood is strongly associated with a range of disadvantages for both the mother and the child. No epidemiological studies have examined related factors for teenage motherhood at both area and individual levels among Japanese women. Therefore, we performed a multilevel analysis of nationwide data in Japan to explore the association of area- and individual-level factors with teenage motherhood.

Methods

The study population comprised 21,177 mothers living in 47 prefectures who had their first, singleton baby between 10 and 17 January or between 10 and 17 July, 2001. Information on the prefecture in which the mothers resided was linked to prefecture-level variables. Primary outcomes were area-level characteristics (single-mother households, three-generation households, college enrollment, abortions, juvenile crime, and per capita income) and individual-level characteristics, and divided into tertiles or quintiles based on their variable distributions. Multilevel logistic regression analysis was then performed.

Results

There were 440 teenage mothers (2.1%) in this study. In addition to individual low level of education [adjusted odds ratio (OR), 7.40; 95% confidence interval (CI), 5.59–9.78], low income [4.23 (2.95–6.08)], and smoking [1.65 (1.31–2.07)], high proportions of single-mother households [1.72 (1.05–2.80)] and three-generation household [1.81 (1.17–2.78)], and per capita income [2.19 (1.06–3.81)] at an area level were positively associated, and high level of college enrollment [0.46 (0.25–0.83)] and lower crime rate [0.62 (0.40–0.98)] at area level were inversely associated with teenage motherhood compared with the corresponding women living in prefectures with the lowest levels of these variables.

Conclusions

Our findings suggest that encouraging the completion of higher education and reducing the number of single-mother household at an area level may be important public health strategies to reduce teenage motherhood.

Introduction

Teenage motherhood is an important issue in terms of maternal and child health [1,2]. For example, teenage motherhood is reported to lead to adverse health outcomes, including coronary heart disease and suicide, due to lower physical and mental health conditions [35]. Moreover, teenage motherhood is strongly associated with social deprivation such as school disruption and lower educational attainment [6], unemployment [7], poverty [8], single parenthood [9], and child abuse or neglect [10]. Newborns from teenage pregnancies are associated with low birth weight and preterm delivery [11, 12], which could have a latent effect on the development of diseases during adulthood [13, 14]. Furthermore, children born to teenage mothers are reported to have a variety of adverse health and socioeconomic outcomes and are more likely to develop attention-deficit/hyperactivity disorder [15], to achieve a lower education and personal income [16], and to experience an unsafe and early initiation of sexual intercourse [17, 18]. These children themselves may eventually become teenage parents, starting the cycle all over again [2].

To respond to teenage motherhood, it is important to detect high risk groups and take measures for prevention. Several risk factors for teenage motherhood at individual level have been reported in the US, UK, and Australia, such as a history of childhood abuse, early initiation of sexual activities, being in a single-parent family, being the child of a teenage mother, and initiation of smoking during teenage years [1, 7, 1922]. However, no epidemiological investigations have been conducted concerning individual related factors for teenage motherhood in Japan or other Asian countries probably because these reported individual factors are very sensitive to be asked about. Another strategy is identifying characteristics at area level where data are regularly collected through existing frameworks as well as at individual level, regardless of causality. This approach may help to reduce the effect of adversity for teenage mothers and their children in these areas.

Some studies have already reported that area-level factors influence teenage motherhood; however, these studies included no individual-level data and were limited to investigating the association of teenage motherhood with social deprivation pertaining to neighborhood levels, which may reflect aggregated individual deprivations [8, 2325]. Since it is difficult to differentiate the impact of social deprivation at neighborhood level from that at individual level, further studies with both area-level and individual-level factors might be also be helpful.

A Japanese previous study using only area-level variables in Japan showed a strong ecological correlation between early adolescent (i.e., 13–15 years) pregnancy and juvenile delinquency [26]. Based on this finding, we hypothesized that area-level security may have an impact on teenage motherhood. At the same time, we also hypothesized that an area where single-mother households are more common may be associated with higher proportion of teenage motherhood because girls may have more opportunities to consider diverse family structures as their future choices. Similarly, living in a society where three-generation households account for a large proportion of households may provide a socially supportive environment which leads a sense of relief for young women [27, 28].

The purpose of this study was to explore the association of area-level and individual-level factors relating to teenage motherhood, using nationwide data in Japan.

Materials and Methods

Study sample

Data for the individual level were obtained from the Japanese panel study entitled “The Longitudinal Survey on Babies Born in the 21st Century.” Based on the birth record lists of Vital Statistics, the study sample included all babies born in Japan between January 10 and 17, 2001, or between July 10 and 17, 2001 (n = 53,575). Selected subjects were recruited by mail questionnaires when the children were 6 months old. The subjects were considered to have agreed to participate in the study if they returned the questionnaire to the Ministry of Health, Labour, and Welfare. There were a total of 47,015 respondents, giving a response rate of 87.8%. Participants who had not provided information on the mother’s age, maternal education, or maternal smoking status were excluded (n = 3,612). Among the remaining 43,403 mothers, those with multiple births and more than one parity, and mothers who did not live with the child were excluded, which lead to a final sample size of 21,159 mothers. A teenage mother was defined as a mother who gave birth when she was less than 20 years old. Mother’s age at delivery was retrieved from the birth record list.

We obtained permission to use these panel data from the Ministry of Health, Labour and Welfare under the Statistics Act in Japan (No.1127-4).

This study was approved by the Ethical Review Board of National Research Institute for Child Health and Development (No.1007).

Predictor variables

Area-level characteristics.

Several characteristics of prefectural conditions were taken into consideration. We chose (a) the proportion of single-mother households (%), (b) the proportion of three-generation households (%), (c) the proportion of college enrollment (%), (d) abortion rate (‰), (e) juvenile crime rate (‰), and (f) per capita annual income. Proportions of single-mother and three-generation households were obtained from Vital Statistics (special issue) of 2000 [29]. We preferred to use college enrollment rather than high school enrollment because average high school enrollments in Japan are greater than 95%, with very little variance. The proportion of college enrollment is defined as the number of students who enroll in either university or college per 1,000 students who graduate from high school, and was obtained from the School Basic Survey for 2000 [30]. Abortion rate was defined as the number of terminations of pregnancy before 22 weeks’ gestation per 1,000 women aged 15–49 years old [31], and data were obtained from the Report on Public Health Administration and Service for 2000 [32]. Data for per capita annual income were obtained from the Annual Report on Prefectural Accounts for fiscal year 2000 [33]. All these data were accessible from the database of the Ministry of Internal Affairs and Communications [34]. The juvenile crime rate was defined as the number of persons aged less than 20 years who committed a crime per 1,000 junior-high and high school students. These data were retrieved from the Report on Minority Protection and Guidance for 2000 published by the National Police Agency [35]. These area-level variables were divided either into tertiles or quintiles based on their distribution, not weighted by the population of prefectures, and linked to individual data according to living prefecture at the time of delivery.

Individual-level characteristics.

Individual variables used in this study were retrieved from data gathered from the first and the second panel survey. Information on education, couple’s income, cohabitation with the child’s father, and living in a three-generation household were obtained from the second survey (response rate: 93.5%) in 2002. Education was categorized as junior high school or high school or over and couple’s income was categorized as “less than 2 million Japanese Yen” and “2 million Japanese Yen and over”. Cohabitation with the child’s father is categorized as “Yes” if the father live with the family, or father who lives separately but periodically come back to the family, and “No” if the father did not live with for more than 3 months. Information on smoking habits were obtained from questionnaires used in the first survey in 2001, and respondents were classified as nonsmokers or smokers.

Statistical analyses

Multilevel analyses were performed with 21,159 mothers nested within 47 prefectures. Associations with teenage motherhood were modeled using multilevel logistic regression of the above structure using STATA MP statistical package version 14 (StataCorp L, College Station, Texas, USA). We included similar area-level and individual-level characteristics such as education level, income, and three-generation household to differentiate the impacts of area-level from individual-level characteristics. We first included individual-level variables nested within 47 prefectures (model 1). Next, we added all individual-level and area-level variables nested within the 47 prefectures (model 2). Odds ratios (OR), presented with 95% confidence intervals (CI), were calculated.

Results

Table 1 presents the correlation coefficients for area-level variables. Among these indicators, college enrollment and per capita income were closely correlated. In addition, both were inversely correlated with the proportion of single-mother households and abortion rate.

thumbnail
Table 1. Spearman Correlation Coefficients for Area-Level Variables (n = 47).

https://doi.org/10.1371/journal.pone.0166345.t001

Of the 21,159 Japanese women in this study, of those who gave birth to their first, singleton baby between January 10 and 17, 2001 or between July 10 and 17, 2001, 2.1% (440 mothers) were teenagers (Table 2). Among these, 82.6% were 18 or 19 years of age (data not shown). Teenage mothers were more common in prefectures with higher proportions of single-mother households, three-generation households, and lower college enrollment, abortion rates, and lower per capita annual incomes. At the individual level, teenage mothers were less-well educated, had lower incomes, and less co-reside with the child’s father (Table 3). Teenage mothers were more likely to reside in a three-generation household, and to smoke.

thumbnail
Table 2. Characteristics of the Study Areas and Prevalence of Teenage Motherhood, in Japan, 2001.

https://doi.org/10.1371/journal.pone.0166345.t002

The results of the multilevel logistic regression analysis are shown in Table 4. Variances of the random intercept were significantly larger than zero for the null models (results not shown in Tables) and for model 1, while in model 2 (full models), they became zero and non-significant, indicating that no residual area-level variance existed. In model 1, we found that low level of education, low income, smoking, not living with the father, and living in a three-generation household at an individual level were associated with teenage motherhood. In model 2, a higher proportion of single-mother households (Q4), middle and highest proportions of three-generation households (Q3 and Q5), and highest per capita income (T3) were associated with teenage motherhood. Also, higher college enrollment rate (T2 and T3) and lower juvenile crime rate (Q2) were inversely associated with teenage motherhood. The abortion rate at an area level was not associated with teenage motherhood.

thumbnail
Table 4. Result of Multilevel Analyses for Teenage Motherhood: Adjusted Odds Ratios (and 95% CI).

https://doi.org/10.1371/journal.pone.0166345.t004

Discussion

In this study using a nationally representative sample of Japanese mothers, we found that teenage motherhood was associated primarily with individual-level factors, especially low education, but also with area-level factors.

To the best of our knowledge, this is the first study to demonstrate that teenage motherhood is positively associated with high area-level proportions of single-mother households and three-generation households, as well as high per capita income, and inversely related to high rates of college enrollment and low juvenile crime rates. We also demonstrated that low level of education and low income at individual level were associated with teenage motherhood; however, this is to be anticipated because teenage birth apparently disrupts schooling, which results in low education and low income.

For area-level characteristics, since an inverse correlation of college enrollment and per capita income with proportion of single-mother households were observed, we considered that these indicators may capture the dimension of area deprivation. The proportion of three-generation households has been considered an indicator of social support. Abortion rate has primarily been considered an indicator of unintended pregnancy and attitudes for conforming to be cultural norm, since births outside of marriage represent only 1.7% of all births in Japan [31]. Furthermore, the juvenile crime rate has been considered an indicator of area security.

We found an association of indicators for deprivation (high proportion of single-mother households, low college enrollment) with teenage motherhood, which was accordance with our hypothesis. However, high per capita income, high proportions of three-generation households, and low juvenile crime rates, all of which might imply rich social supports, showed a positive association with teenage motherhood. To interpret these findings, we may need to consider that the outcome “teenage motherhood” could be broken down into two elements; teenage pregnancy and choosing to purse pregnancy to birth. In fact, most teenage pregnant women in Japan terminate their pregnancies by induced abortion, with 2.2 abortions for each live birth, in contrast with 0.4 abortions for each live birth in the US [31]. Teenage motherhood might thus reflect a positive decision to avoid induced abortion.

Taking these related factors into account, the effect by high college enrollment (representing least social deprivation) might be inversely associated with the first element (teenage pregnancy) [21, 23, 25] and high proportions of three-generation households, low juvenile crime rates, and a high per capita income at an area level might be associated with the latter element (decision to pursuing pregnancy), which may provide teenage pregnant women with their motivation to continue with the pregnancy and ensure child rearing in a safe environment with rich social support [27, 28].

Our results showed that higher proportions of single-mother households were associated with teenage motherhood, except in those areas with the very highest proportions of single-mother households. The mechanism for this effect is unknown. In our study population, 18.1% (80/440) of teenage mothers were not living with their partners (i.e., they were single parents) compared with only 3.5% (730/21159) of mothers of all ages. Since teenage motherhood is a strong predictor for single motherhood [7], we speculate that as the proportion of single-mother households rises, young parenting which might lead to single parenting becomes the social norm in these communities and is considered more acceptable [21, 36]. However, the highest single-mother household areas showed no increased association.

The findings of this study demonstrate the need for implementing strategies for the prevention of teenage motherhood, especially aimed at the prevention of teenage pregnancy, and the following adverse events. Primarily, rising levels of general education is fundamentally important as our findings showed an inverse association with high college enrollment at an area-level [2]. Entry rates into university-level education in Japan ranks 23rd among 34 Organization for Economic Co-operation and Development (OECD) countries, which may partly be due to the fact that public college tuition fees in Japan are the second highest among OECD countries and the public subsidies for higher education, mostly through student loan, but not scholarship [37]. Furthermore, achieving 100% high school enrollment and completion is desirable because 17.8% of teenage mothers did not graduate from high school at an individual level in our study even though the rate for enrollment and completion of high school was generally high (97% and 98%, respectively, in 2000) in Japan [38, 39]. Secondly, robust sex education should be provided both at school and by other local sectors that are accessed outside of the public educational system because more than 80% of teenage mothers were either 18 or 19 years of age and had not achieved completion of high school education. Basic knowledge concerning reproductive rights, including access to contraception, emergency pills, and induced abortion, should be carefully taught, and information on the availability of social welfare such as public assistance should also be provided when teenage pregnancies or other difficult pregnancies arise. Moreover, when these measures result in choosing to pursue a pregnancy, educational and supportive programs for young expectant mothers are essential for the prevention of further unintended pregnancies and for the development of healthy children [40].

Several weaknesses of our study need to be addressed before a conclusion can be achieved. First, the associations of area-level and individual-level variables with teenage motherhood might reflect reverse causation because the temporality is difficult to establish in this study. For example, the association of living in a three-generation household with teenage motherhood might be observed because the young mothers need the support from their parents by living in the same household. However, the characteristics of teenage motherhood both in individual- and area-levels could be captured through this study. Second, a Japanese prefecture, with a population ranging from 0.6 million to 13 million may be too large to determine the effect of area-level factors on individual behavior. A smaller area such as a city may be more appropriate for analysis of the area-level effect. However, the number of teenage mothers in each city may be too small for the analysis of teenage motherhood. In this study, we were able to determine the related factors at a prefecture level, probably because some policies relating to maternal and child health are set at prefecture level as well as at national and municipal levels. Third, information on individual characteristics before pregnancy was limited because the target population for the national survey was determined after delivery. Also, the variable “not living with their partners” may include those whose partner did not live with because of the partner’s job transfer.

Conclusion

In conclusion, our findings suggest that high proportions of single-mother household, three-generation households, and high per capita income were positively associated with teenage motherhood, while high levels of education and low juvenile crime rates were inversely associated. With the aid of panel data, it is essential to confirm whether changes in area-level factors over time may reduce teenage motherhood.

Author Contributions

  1. Conceptualization: SB.
  2. Formal analysis: SB.
  3. Funding acquisition: SB.
  4. Investigation: SB.
  5. Methodology: SB.
  6. Project administration: SB.
  7. Supervision: HI TF.
  8. Validation: TF.
  9. Visualization: SB HI TF.
  10. Writing – original draft: SB.
  11. Writing – review & editing: HI TF.

References

  1. 1. Gaudie J, Mitrou F, Lawrence D, Stanley FJ, Silburn SR, Zubrick SR. Antecedents of teenage pregnancy from a 14-year follow-up study using data linkage. BMC Public Health. 2010;10:63. pmid:20149221
  2. 2. UNICEF. A league table of teenage births in rich nations. Florence: 2001.
  3. 3. Hardy R, Lawlor DA, Black S, Mishra GD, Kuh D. Age at birth of first child and coronary heart disease risk factors at age 53 years in men and women: British birth cohort study. J Epidemiol Community Health. 2009;63(2):99–105. pmid:18782806
  4. 4. Patel PH, Sen B. Teen motherhood and long-term health consequences. Matern Child Health J. 2012;16(5):1063–71. pmid:21656056
  5. 5. Webb RT, Marshall CE, Abel KM. Teenage motherhood and risk of premature death: long-term follow-up in the ONS Longitudinal Study. Psychol Med. 2011;41(9):1867–77. pmid:21275086
  6. 6. Kane JB, Morgan S, Harris KM, Guilkey DK. The educational consequences of teen childbearing. demography. 2013;50(6):2129–50. pmid:24078155
  7. 7. Nanchahal K, Wellings K, Barrett G, Copas AJ, Mercer CH, Macmanus S, et al. Changes in the circumstances of young mothers in Britain: 1990 to 2000. J Epidemiol Community Health. 2005;59(10):828–33. pmid:16166353
  8. 8. Kirby D, Coyle K, Gould JB. Manifestations of poverty and birthrates among young teenagers in California zip code areas. Fam Plann Perspect. 2001;33(2):63–9. pmid:11330852
  9. 9. Hobcraft J, Kiernan K. Childhood poverty, early motherhood and adult social exclusion. Br J Sociol. 2001;52(3):495–517. pmid:11578006
  10. 10. Thornberry TP, Matsuda M, Greenman SJ, Augustyn MB, Henry KL, Smith CA, et al. Adolescent risk factors for child maltreatment. Child Abuse Negl. 2013;38(4):706–22. pmid:24075569
  11. 11. Olausson PO, Cnattingius S, Haglund B. Teenage pregnancies and risk of late fetal death and infant mortality. Br J Obstet Gynaecol. 1999;106(2):116–21. pmid:10426676
  12. 12. Chen XK, Wen SW, Fleming N, Demissie K, Rhoads GG, Walker M. Teenage pregnancy and adverse birth outcomes: a large population based retrospective cohort study. Int J Epidemiol. 2007;36(2):368–73. pmid:17213208
  13. 13. Roseboom TJ, van der Meulen JH, Osmond C, Barker DJ, Ravelli AC, Schroeder-Tanka JM, et al. Coronary heart disease after prenatal exposure to the Dutch famine, 1944–45. Heart. 2000;84(6):595–8. pmid:11083734
  14. 14. Barker DJ, Osmond C. Infant mortality, childhood nutrition, and ischaemic heart disease in England and Wales. Lancet. 1986;1(8489):1077–81. pmid:2871345
  15. 15. Chang Z, Lichtenstein P, D’Onofrio BM, Almqvist C, Kuja-Halkola R, Sjölander A, et al. Maternal age at childbirth and risk for ADHD in offspring: a population-based cohort study. Int J Epidemiol. 2014;43(6):1815–24. pmid:25355726
  16. 16. Lipman EL, Georgiades K, Boyle MH. Young adult outcomes of children born to teen mothers: effects of being born during their teen or later years. J Am Acad Child Adolesc Psychiatry. 2011;50(3):232–41 e4. pmid:21334563
  17. 17. Center for Disease Control and Prevention. Prepregnancy contraceptive use among teens with unintended pregnancies resulting in live births—Pregnancy Risk Assessment Monitoring System (PRAMS), 2004–2008. MMWR Morb Mortal Wkly Rep. 2012;61(2):25–9. pmid:22258415
  18. 18. Paul C, Fitzjohn J, Herbison P, Dickson N. The determinants of sexual intercourse before age 16. J Adolesc Health. 2000;27(2):136–47. pmid:10899475
  19. 19. Noll JG, Shenk CE. Teen birth rates in sexually abused and neglected females. Pediatrics. 2013;131(4):e1181–e7. pmid:23530173
  20. 20. Burack R. Teenage sexual behaviour: attitudes towards and declared sexual activity. Br J Fam Plann. 1999;24(4):145–8. pmid:10023100
  21. 21. Smith DM, Elander J. Effects of area and family deprivation on risk factors for teenage pregnancy among 13-15-year-old girls. Psychology, health & medicine. 2006;11(4):399–410.
  22. 22. Hogan DP, Sun R, Cornwell GT. Sexual and fertility behaviors of American females aged 15–19 years: 1985, 1990, and 1995. Am J Public Health. 2000;90(9):1421–5. pmid:10983200
  23. 23. McLeod A. Changing patterns of teenage pregnancy: population based study of small areas. BMJ. 2001;323(7306): 199–203. pmid:11473910
  24. 24. Uren Z, Sheers D, Dattani N. Teenage conceptions by small area deprivation in England and Wales, 2001–2002. Health Statistics Quarterly. Office for National Statistics. 2007(33):34–9.
  25. 25. Johns SE. Perceived environmental risk as a predictor of teenage motherhood in a British population. Health Place. 2011;17(1):122–31. pmid:20951627
  26. 26. Baba S, Goto A, Reich MR. Recent pregnancy trends among early adolescent girls in Japan. J Obstet Gynaecol Res. 2014;40(1):125–32. pmid:24033703
  27. 27. Silverstein M, Cong Z, Li S. Intergenerational transfers and living arrangements of older people in rural China: Consequences for psychological well-being. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences. 2006;61(5):S256–S66.
  28. 28. Saito E, Ueki S, Yasuda N, Yamazaki S, Yasumura S. Risk factors of functional disability among community-dwelling elderly people by household in Japan: a prospective cohort study. BMC Geriatr. 2014;14:93. pmid:25154498
  29. 29. Ministry of health, labour and welfare. Vital statistics (special issue) of 2000. Tokyo: Ministry of health, labour and welfare, 2001.(in Japanese)
  30. 30. Ministry of education, culture, sports, science and technology. School Basic Survey for fiscal year 2000. Ministry of education, culture, sports, science and technology, 2001. (in Japanese)
  31. 31. Baba S, Tsujita S, Morimoto K. The analysis of trends in induced abortion in Japan- An increasing consequence among adolescents. Environ Health Prev Med. 2005;10(1):9–15. pmid:21432158
  32. 32. Ministry of health, labour and welfare. Report on public health administration and service for fiscal year 2000. Tokyo: Ministry of health, labour and welfare, 2001 (in Japanese)
  33. 33. Cabinet Office, Government of Japan. Annual report on prefectural accounts for fiscal year 2000. Tokyo: Cabinet Office, Government of Japan, 2001 (in Japanese)
  34. 34. Statistics Bureau Ministry of Internal Affairs and Communications. Portal site of official statistics of Japan (e-Stat) Available: http://www.e-stat.go.jp/. (in Japanese)
  35. 35. National Police Agency. Overview of juvenile deliquency and protection in 2000. Tokyo: National Police Agency, 2001.(in Japanese)
  36. 36. Leventhal T, Brooks-Gunn J. The neighborhoods they live in: The effects of neighborhood residence on child and adolescent outcomes. Psychol Bull. 2000;126(2):309–37. pmid:10748645
  37. 37. OECD. Education indicators in focus. 2012. Available: http://www.oecd.org/education/skills-beyond-school/49729932.pdf.
  38. 38. Ministry of Education Culture, Sports, Science and Technology. FY 2014 School Basic Survey. Available: http://www.mext.go.jp/english/topics/__icsFiles/afieldfile/2015/09/03/1361507_1.pdf
  39. 39. Ministry of Education Culture, Sports, Science and Technology. Report on school refusaland dropout among high school students. Available: http://www.mext.go.jp/b_menu/shingi/chukyo/chukyo3/047/siryo/__icsFiles/afieldfile/2012/03/21/1318690_02.pdf. (in Japanese)
  40. 40. Doyle O, Harmon CP, Heckman JJ, Tremblay RE. Investing in early human development: Timing and economic efficiency. Econ Hum Biol. 2009;7(1):1–6 pmid:19213617