Abstract
Objectives To study sexually transmitted Chlamydia trachomatis infections (STCTs), teenage pregnancies, and payment defaults in individuals born preterm as proxies for engaging in risk-taking behavior. Study design Our population-based register-linkage study included all 191 705 children alive at 10 years (8492 preterm 4.4%) born without malformations in Finland between January 1987 and September 1990 as each mother's first child within the cohort. They were followed until young adulthood. We used Cox regression to assess the hazards of STCTs, teenage pregnancies, payment defaults, criminal offending, and substance abuse by gestational age. Gestational age was considered both as a continuous and categorical (extremely, very, moderately, late preterm, early term, post term, and full term as reference) exposure. Results A linear dose-response relationship existed between gestational age and STCT and teenage pregnancy; adjusted hazard for STCT decreased by 1.6% (95% CI, 0.7%-2.6%), and for teenage pregnancy by 3.3% (95% CI, 1.9%-4.8%) per each week decrease in gestational age. Those born extremely preterm (23-27 completed weeks) had a 51% (95% CI, 31%-83%) lower risk for criminal offending than their full-term born counterparts, and those born very preterm (range, 28-31 weeks) had a 28% (95% CI, 7%-53%) higher hazard for payment defaults than those born at full term. Gestational age was not associated with substance abuse. Conclusions The lower risk-taking that characterizes people born preterm seems to generalize to sexual and to some extent criminal behavior. Those born very preterm are, however, more likely to experience payment defaults.
Abbreviations∗∗
HR - hazard ratio
SEP - Socioeconomic position
STCT - Sexually transmitted Chlamydia trachomatis infection
Individuals born extremely (before 28 weeks of gestation) or very preterm (28-31 weeks of gestation), or with extremely low birth weight (<1000 g) or very low birth weight(<1500 g) more often have poorer neurocognitive function than those born at full term.1, 2, 3, 4, 5 Those born preterm also tend to have a behavioral profile characterized by cautiousness in social relationships, inattention, and internalizing problems such as depression and anxiety. They, in contrast, have lower levels of externalizing problems, such as maladaptive behavior toward one's environment.6, 7, 8, 9 The influence of suboptimal gestational age at birth on behavioral outcomes seems to continue to some extent to adolescence.3,6,10, 11, 12, 13
Preterm born individuals are as likely or less likely to commit criminal offences, and have similar or lower rates of risk-taking behaviors, namely smoking, use of illicit drugs and alcohol abuse than their term born peers.14, 15, 16, 17, 18, 19, 20, 21, 22, 23 However, less is known about risk-taking behavior in other domains of life such as in peer or partner romantic relationships and in managing personal finances among those born preterm. In this nationwide, individual-level register linkage study we assessed whether preterm born young adults have, compared with full-term born controls, lower rates of teenage pregnancies, sexually transmitted infections, payment defaults, criminal offences, and care episodes for substance abuse.24
Methods
The data originated from 11 nation-wide administrative registers: (1) the Finnish Medical Birth Register, (2) the Central Population Register (updated through April 2012), (3) Register of Congenital Malformations (January 2015), (4) The Finnish Care Register for Health Care (December 2015), (5) Census Register (December 2014), (6) the Register on Disability Allowances (December 2015), (7) National Infectious Disease Register (December 2012), (8) Register on Induced Abortions and Sterilizations (December 2017), (9) register on credit ratings (August 2018), (10) register on crimes (February 3, 2020), and (11) register on fines and punishments (December 10, 2019). The 5 last-mentioned registers are described in the Appendix (available at www.jpeds.com ). Other registers are described in detail elsewhere.25,26 Register data were merged by using encrypted personal identification codes, which enable full-coverage, accurate individual linkages between registers. Based on Finnish and European Union legislation, individual consents are not required in studies based on pseudonymized register data if the registered persons are not contacted. The local ethics committees and applicable register authorities approved the study protocol. The study was approved by the Local Ethical Review Board, Helsinki (Helsinki University Central Hospital Ethics Committee Dnr: 200/13/03/00/08, and Dnr: HUS/3580/2017). The study was performed in accordance with the declaration of Helsinki.
Cohort Members
A total of 235 624 index children with valid personal identity code (99.8% of all live-born children) born in Finland between January 1, 1987, and September 30, 1990, were identified from the Medical Birth Register. Of them 7073 (3.0%) were excluded owing to missing Central Population Register data, missing or inaccurate information on gestational age, or/and major congenital anomaly (for details see Figure 1 and the Appendix). Our main analysis included only each mother's first child born during the recruiting period (191 705 [83.9%]) to avoid within-family correlations and within family influences. To avoid follow-up in such ages that certain outcome events are not recorded to the official registers or are extremely unlikely to occur as a consequence of risk-taking behavior, we started the follow-up at 10, 15, or 18 years of age (depending on the outcome, as described elsewhere in this article).
Exposure, Outcomes, and Covariates
Gestational age was assessed according to ultrasound estimates and/or estimates based on maternal last menstrual period and was considered either as a continuous exposure or categorized into 7 subgroups: extremely preterm, 23-27 completed weeks; very preterm, 23-31 weeks; moderately preterm, 32-33 weeks; late preterm, 34-36 weeks; early term, 37-38 weeks; full term, 39-41 weeks (reference); and post term, 42 weeks.
The most useful proxies of risk-taking behavior that can be straightforwardly and reliably derived from the national registers were (1) sexually transmitted Chlamydia trachomatis(STCT) infections (yes vs no), (2) teenage pregnancies among females (live birth of a child or induced abortion before 20 years of age, yes vs no), (3) payment defaults (defaulting on paying back debts; yes vs no), (4) substance abuse diagnosed in a hospital inpatient or outpatient clinic (yes vs no), and (5) criminal offending (yes vs no). Detailed information on these outcomes, including precise definition of each outcome, is available in the Appendix, description of the outcomes, and Table I (available at www.jpeds.com ).
The covariates in the models are described in detail elsewhere and in the Appendix, description of the covariates, and Table II (available at www.jpeds.com ).25,26 In the 4 analyses including both sexes, the sex of the study subject was included as a stratum in the Cox model to meet the assumptions for Cox proportional hazards, and accounting for sex differences in pubertal timing, adolescent brain, and behavior.27,28 Concerning each of the 4 outcomes the gestational age associations were similar across strata (male/female) for all of them; P for interaction of less than .57 (STCT), less than .58 (payment defaults), less than .76 (substance abuse), and less than .75 (criminal offending). The sex of the study subject (male vs female) originated from the Medical Birth Register, to which it is reported within first postnatal week.
The other categorical covariates covered data on parental highest ever attained level of socioeconomic position (SEP), parental ages separately, maternal marital status, smoking, and gestational disorders, severe medical conditions of the adolescent study subject, and parental behavior including substance abuse, criminal offending, and payment defaults. Missing data (Table III) were coded to a separate category when they were missing for 3 or more study subjects; otherwise, we applied single imputation with most likely values; data on parental payment defaults missing—no parental payment defaults; data on parental criminal offending missing—no parental criminal offending; data on parental substance abuse missing—no parental substance abuse. In analyses no data was missing for continuous variables (gestational age or birth-weight SD score).
Statistical Analyses
In the analyses on STCTs, teenage pregnancies, and substance abuse the follow-up started at 10 years of age. The follow-up started at 15 years of age when assessing criminal offending, and at 18 years of age when payment defaults were assessed. These ages were chosen because juveniles younger than 15 years of age cannot be held legally responsible for criminal behavior in Finland and therefore records are unavailable, and only those 18 years of age or older can sign financial agreements.29 Total follow-up time varied according to outcome, owing to varying end dates of outcome data, and the birth year of the study subject. Thus, the ages of the study subjects at the end of follow-up were as follows: STCTs, mean 24.6 years; teenage pregnancies, mean, 20.0 years; payment defaults, mean 29.9 years; substance abuse mean 27.3 years; and criminal offending, mean 31.2 years. Those who emigrated or died during the follow-up, or reached the end of follow-up without an event, whichever occurred first, were considered as censored observations.
Using Cox regression, we calculated the HRs and 95% CIs in the 6 nonreference gestational age subgroups in weeks, and also modelled the log HR either as a linear or a spline function of the continuously measured gestational age in days.30,31 We applied restricted cubic splines with 4 knots set at 28, 33, 37, and 41 weeks of gestation. Linear trend was assessed by comparing the model involving the continuous gestational age variable with a model with a constant log HR and nonlinearity by comparing the spline model to the model with a linear trend. For illustrations, we calculated 95% CIs for the estimated linear and spline functions.
We also investigated whether the gestational age category-specific HRs for any of the proxies of risk-taking behavior were dependent on age. We estimated age-dependent HRs by using the interaction terms calculated separately for the gestational age categories and a categorized age variable with cut-off points at 13, 15, 18, 22, and 25 years of age (when the cut-off point was within the follow-up period).
Because some evidence exist on associations between birth weight and behavioral problems in adolescence, we examined potential interactions between gestational age and fetal growth (for which birth-weight SD score served as a proxy) by introducing the corresponding interaction term in the models (Appendix).32, 33, 34
SPSS 26 (IBM Corp) was the main statistical software. R software (R Foundation for Statistical Computing) version 3.6.0 with packages “survival” and “rms” was employed to estimate the linear and spline regression models.31,35,36
Results
The study population in main analyses with follow-up from 10 years of age consisted of 191 705 individuals (Table III). In comparison with full-term born study subjects, those born preterm had a lower birth-weight SD score and were more likely to have severe medical conditions in adolescence. The mothers of those born preterm were more likely to have a gestational disorder, smoke during pregnancy, and be unmarried. Both the mothers and the fathers of those born preterm were older. They were also more likely to have blue collar SEP, payment defaults, and substance abuse as compared with the parents of full-term born study subjects (P < .001; χ2 or t test). The rates of parental criminal offending were similar between the groups (P < .12; χ2 test).
The age period and gestational age category-specific HRs for each of the proxies of risk-taking behavior indicated that the effect of gestational age was consistent across the 6 predefined age periods (cut-off points at 13, 15, 18, 22, and 25 years when applicable). Therefore, the age period-specific estimates are not shown. Table IV (available at www.jpeds.com ) shows that the risk-taking-related outcomes were associated with male sex (except STCT), maternal smoking, younger age of parents, and parental SEP of blue collar class.
STCTs
Table V shows the proportion of individuals with STCT by gestational age category. In the fully adjusted model, gestational age as a categorical variable was not statistically significantly associated with STSC (Figure 2, A, Table VI; available at www.jpeds.com). However, a linear dose-response relationship was statistically significant, corresponding with a 1.6% (95% CI, 0.7%-2.6%) decrease in the hazard per each week decrease in gestational age (Figure 2, A, Table VII; available at www.jpeds.com).
Teenage Pregnancies
Table V shows the proportion of women who experienced teenage pregnancy (birth or induced abortion before 20 years of age) by gestational age category. In the fully adjusted model, women born late preterm (0.83; 95% CI, 0.73-0.94) and early term (0.91; 95% CI, 0.73-0.94) had lower hazard of teenage pregnancy than women born full term (Figure 2, B, Tables VII and VIII; available at www.jpeds.com ). A linear dose-response relationship was statistically significant, corresponding with a 3.3% decrease (95% CI, 1.9%-4.8%) in teenage pregnancies per each week decrease in gestational age.
Payment Defaults
Table V shows the proportion of individuals with payment default by gestational age category. In the fully adjusted model, those born very preterm had 28% higher HR (95% CI, 7%-53%) for payment default than those born full term. This association remained also when parental payment defaults were included as an additional covariate (Figure 2, C, Table IX; available at www.jpeds.com ). Gestational age as a continuous variable was associated with payment defaults in a nonlinear way (for the linear model, P = .54; for the spline model P = .05) (Table VII).
Substance Abuse
Table V shows the proportion of individuals with substance abuse by gestational age category. In fully adjusted models, gestational age as a categorical variable was not statistically significantly associated with substance abuse (Figure 2, D, Table X; available at www.jpeds.com ). The P value for the linear trend between gestational age and substance abuse was .08 (Table VII).
Criminal Offending
Table V shows the proportion of individuals with criminal offending by gestational age. In the fully adjusted models, those born extremely preterm had a 51% lower hazard (95% CI, 31%-83%) for criminal offending than those born full term (Figure 2, E, Table XI; available at www.jpeds.com ). The P value for the linear trend between gestational age and criminal offending was less than .01 (Table VII).
Sensitivity Analyses
The inclusion of all of each mother's children born during the recruiting period (not only the first born during that period) and who were alive at 10 years of age (n = 227 323) had no effect on the interpretations (data not shown).
Discussion
In this population-based register-linkage cohort study of 191 705 individuals, we found that the lower the gestational age is at birth, the lower is the risk for teenage pregnancies and STCTs in adolescence and young adulthood, respectively. Preterm birth was, however, unrelated to the rates of substance abuse. Furthermore, those born preterm had similar relative frequencies of criminal offending as their full-term born peers, except those born extremely preterm (23-27 full gestational weeks) who had a lower risk for criminal offending. Payment defaults were more common among those born very preterm (28-31 weeks of gestation). These associations were not confounded by family SEP, pregnancy-related factors such as maternal pregnancy disorders and smoking in pregnancy, or by severe medical conditions, including neurosensory impairments and other medical disabilities.
The main strength of our study is that it uses a whole population cohort, being large enough to analyze a range of gestational age categories. The study combines, by means of unique personal identity codes, information from 11 population-wide, individual-level administrative registers with only marginal loss to follow-up. The study improves the knowledge on the range of long-term impacts of prematurity; several of our study outcomes are novel and have not been assessed before in this context.
There are also limitations. We had no data on whether the estimation of the gestational age of the child was based on fetal ultrasonography or on last maternal menstrual period. The last maternal menstrual period method tend to overestimate the gestational age, and because fetal ultrasound examination was only being introduced in clinical practice in Finland in 1987 through 1990, the distribution of the gestational ages may have moved toward a slight overestimation of gestational age and decreased rates of preterm birth, with a negligible effect on our estimation.37 Further, information on deaths and emigrations were only accessible until April 2012. The aggregate data provided by Statistics Finland, however, indicate a mortality rate of only 65 per 100 000 and emigration rate of 806 per 100 000 among Finns aged 25-34 years in 2019.38, 39, 40 The availability of the information on the payment defaults and criminal offending were affected by the abolition terms of the registers, as described in the Appendix. Register data on hospital-treated miscarriages were not included. The bias caused by these issues is moderate, leading to slight underestimations of the rates of outcome events, slight inaccuracy or inexactness in defining the study cohort, and to delay of capturing a study subject's first record (payment defaults, criminal offending, and pregnancies). Finally, prenatal corticosteroid treatment, although beneficial, can also potentially harm the brain development of the fetus and may be associated with mental and behavioral disorders of the child.41, 42, 43, 44 This common strategy treatment was not established as a routine in Finland in the late 1980s.45
Our results indicating fewer STCTs and teenage pregnancies among preterm adolescents and young adults are in line with those of Swamy et al and with a recent meta-analysis summarizing the contemporary knowledge on the effect of preterm birth on social relationships in adulthood, including sexual partnership and parenthood.46,47 The results of these 2 studies indicate that, in particular, those born extremely or very preterm have been less likely experienced romantic relationships, sexual intercourse, or parenthood. Although these results do not reflect exactly the same phenomenon as ours, the lower rates of teenage pregnancies and STCTs in our study could be explained by less engagement in sexual behavior. However, these results could also be explained by other factors, such as more frequent contraceptive use.48 To our best knowledge, no prior studies on preterm birth and STCTs, or personal payment default records exists. Basten et al showed that, at 42 years of age, adults born preterm more often reported personal financial difficulties compared with their term-born peers.49 They, however, interpret the study findings more as a result of a weaker cognitive capacity, resulting in poorer education and wealth in adulthood, and less as an expression of a certain behavioral trait.
Our results indicate that very preterm born individuals have a higher risk for payment defaults. This finding may be due to more periods of unemployment, more receipt of social benefits, lower academic qualifications, poorer self-control, and weaknesses in certain cognitive tasks, such as judgements of risks.50 In comparison with those born very preterm, those born extremely preterm may more unlikely be in situation to take up credit or commit an offence, and consequently be at a lower risk of payment default(s) and record(s) of criminal offending.
Risk-taking behavior has been assessed primarily by questionnaires measuring proxies of risk taking or sensation seeking, including smoking, alcohol and illicit drug use, and violation of the law. The study results indicate lesser or similar level of risk-taking behavior among those born preterm.14,16,17,20, 21, 22, 23 Few studies based on register data only on preterm birth and risk-taking behaviors exist; a Norwegian study assessing drug felony, violence, criminal damage, and overall criminality across separate levels of gestational age found no differences between gestational age categories.15 Two nationwide Swedish register studies indicate that individuals born preterm appear to be at slightly higher risk for addictive disorder(s).51,52 This association was, however, attenuated when comorbidity with other psychiatric disorders was taken into account.51Two other Swedish register studies reported no association between preterm birth (without being small for gestational age) and substance-related disorders or drug dependence, and one of these studies showed increased rates for drug dependence for those born preterm and small for gestational age.19,53 A further Swedish study indicated decreased risk for problematic substance use and criminality among individuals born extremely preterm.18 A Danish study of 8000 individuals, supplemented with register data, suggested an increase in the risk of developing alcoholism in preterm born males, but not among females.54 We found no interaction between the effects of sex and preterm birth on substance abuse. Generally, studies on the association between gestational age and substance abuse based on questionnaires or interviews indicate lower rates of such behavior among those born preterm than among those born at term. Register-based studies employing inpatient and outpatient diagnosis data from health care registers do not capture substance abuse that does not need medical attention and therefore is likely to comprise more severe cases. This may in part explain the differences in the direction of estimates between the studies of different design.
The relationship between preterm birth and in general lower or similar likelihood of engaging in risk-taking behavior is not straightforward and is not wholly explained by this study. Reasons for differences between gestational age categories in the rates of such behavior that may be considered as risky remain unclear. It has been hypothesized that variation in behavior could be related to personality traits typical to preterm born individuals, qualitative differences in social interaction, a slower transition to young adulthood, parental behavior or even, especially among those born most preterm, to social isolation and lack of social and intellectual resilience, and fewer friends, as well as to increased bullying victimization, which have been found to be related to less financial planning previously.6,8,55, 56, 57, 58, 59, 60, 61, 62, 63, 64 In addition, other factors/mechanisms, such as not living independently, could contribute to this association.65,66 These factors could not be assessed within this register-based study.
Our findings on lower rates of chlamydia infections and teenage pregnancies with decreasing gestational age and lower rates of criminal behavior among those born extremely preterm suggest that the lower risk taking extends to sexual and to some extent criminal behaviors in adolescents and young adults born preterm (or early term). Higher rates of payment defaults were contrary to this hypothesis, but limited to those born very preterm. These novel findings provide knowledge on the range of the long-term impacts of prematurity, highlighting once again that preterm birth is not a pediatric condition only. They also expand the understanding of the preterm-born individual in the context of a life course approach and may also contribute to self-awareness among those born preterm.