Abstract
Background: Previous research has demonstrated the predictive validity of the Infant Risk of Overweight Checklist (IROC). This study further establishes the predictive accuracy of the IROC using data from the Avon Longitudinal Study of Parents and Children (ALSPAC) and examines the optimal threshold for determining high risk of childhood overweight.
Methods: Using the IROC algorithm, we calculated the risk of being overweight, based on International Obesity Task Force criteria, in the first year of life for 980 children in the ALSPAC cohort at 5 years. Discrimination was assessed by the area under the receiver operating curve (AUC c-statistic). Net reclassification index (NRI) was calculated for risk thresholds ranging from 2.5% to 30%, which determine cutoffs for identifying infants at risk of becoming overweight.
Results: At 5 years of age, 12.3% of boys and 19.6% of girls were categorized overweight. Discrimination (AUC c-statistic) ranged from 0.67 (95% confidence interval [CI], 0.62–0.72) when risk scores were calculated directly to 0.93 (95% CI, 0.88–0.98) when the algorithm was recalibrated and missing values of the risk factor algorithm were imputed. The NRI showed that there were positive gains in reclassification using risk thresholds from 5% to 20%, with the maximum NRI being at 10%.
Conclusions: This study confirms that the IROC has moderately good validity for assessing overweight risk in infants and offers an optimal threshold for determining high risk. The IROC algorithm has been imbedded into a computer program for Proactive Assessment of Obesity Risk during Infancy, which facilitates early overweight prevention through communication of risk to parents.
Introduction
The risk factors for being overweight and obese in childhood are identifiable antenatally and/or during early infancy,1 and some of these, such as rapid weight gain, are potentially modifiable. Interventions that enable parents to address modifiable risk factors could contribute to early prevention of childhood overweight and obesity, and therefore identifying infants at greatest risk is important. Current guidance2–4 around the identification and assessment of overweight and obesity risk in infants and young children is contingent on a passive approach with little explicit guidance for health professionals. Health policy3,4 from the UK National Institute for Health and Care Excellence and US Institute of Medicine states that primary prevention and evidence-based interventions are important, but there is little consensus on whether childhood overweight risk can be identified by health professionals and, if so, how it should be communicated.5,6
The Infant Risk of Overweight Checklist (IROC)7 was developed from a comprehensive systematic review1 of early life risk factors associated with being overweight in childhood and validated using data from the UK Millennium Cohort Study (MCS).8 The MCS is a prospective birth cohort in the UK that consists of data from longitudinal interviews with parents. The study analysis was restricted to 13,513 children who had completed anthropometric data at 3 years. The sample was divided into two cohorts: 80% were randomly selected to a derivation cohort for development of the risk algorithm, and the remaining 20% were used to validate the algorithm. Step-wise logistic regression was used to determine the best predictor model for overweight at 3 years.8 Seven predictors were found to be significantly associated with overweight at 3 years in a mutually adjusted predictor model: sex; birth weight; age-adjusted weight gain; maternal prepregnancy BMI; paternal BMI; maternal smoking in pregnancy; and breastfeeding status during the first year. The model yielded a moderately good ability to predict whether the infant would be overweight at 3 years of age (area under the receiver operator curve [R2 = 0.92; AUC] = 0.72). Additionally, the IROC algorithm was found to have a sensitivity (true-positive rate) of 70% (corresponding positive predictive value [PPV] = 38%) and a specificity (true-negative rate) of 68% (corresponding negative predictive value [NPV] = 87%).8
However, the IROC has only been validated in the MCS birth cohort and can therefore only be generalized to similar populations. Recruitment to the MCS was purposively sampled to ensure adequate representation of children from all four UK countries, deprived areas, and areas with high concentrations of black and Asian families.8 It is therefore important to validate the IROC in a second population-based birth cohort. The Avon Longitudinal Study of Parents and Children (ALSPAC) is a UK-based birth cohort that contains a higher proportion of families with white British ethnicity and lower levels of deprivation compared to the MCS cohort.9 The aim of this study is to confirm the predictive accuracy of the IROC algorithm using data from the ALSPAC. The previous validation8 used rapid weight gain over the first 12 months of life as a key predictor of child overweight. The present study evaluates whether earlier assessment of rapid weight gain, at 4 months, has acceptable predictive value given that, arguably, earlier intervention will be more effective and allow parents to make informed choices about weaning. A further aim is to establish the optimal threshold for determining high childhood overweight risk in infants and consider the operationalization of the IROC algorithm for prevention of childhood obesity in a clinical setting.
Methods
Participants
The data for this study were a 10% sample (1432 families) of births, known as the Children in Focus (CiF) cohort, from ALSPAC, a prospective birth cohort that recruited 14,541 pregnant women in Avon, UK, from 1991 to 1992. Participants in the CiF cohort attended clinics at the University of Bristol at various time intervals between 4 months and 61 months of age (5 years). Follow-up data for child weight at 5 years are available for 980 families. Details on the selection, enrollment, and cohort profile are published elsewhere.9,10 Ethical approval was obtained from the ALSPAC Ethics and Law Committee and the Local Research Ethics Committees.
Infant Risk of Obesity Checklist
The IROC algorithm was based on seven predictors which can be easily assessed in the early months of life: sex; infant birth weight; infant rapid weight gain; maternal prepregnancy BMI; paternal BMI; maternal smoking in pregnancy; and breastfeeding in the first year (Table 1). The variables of sex (male/female), maternal smoking in pregnancy (no/yes), ever breastfed (no/yes), and infant rapid weight gain (no/yes) were dichotomized. Rapid weight gain was classified as weight gain >0.67 standard deviation (SD),11–13 from birth to the 4-month assessment by trained interviewers. Infant birth weight, obtained from the birth certificates or health records, was categorized into quintiles: <2.93 kg; 2.93–<3.24 kg; 3.24–<3.49 kg; 3.49–<3.81 kg; and ≥3.81 kg. Self-reported maternal prepregnancy BMI and paternal BMI were categorized into clinically recognized categories of <18.5 kg/m2 for underweight, 18.5–<25 kg/m2 normal weight, 25–<30 kg/m2 overweight, or ≥30 kg/m2 for obese.
Table 1.
Risk factors | Categories |
---|---|
Sex | Male |
Female | |
Infant birth weight quintiles (kg) | <2.93 |
2.93–<3.24 | |
3.24–<3.49 | |
3.49–<3.81 | |
≥3.81 | |
Infant weight gaina | ≤0.67 SD |
>0.67 SD | |
Maternal prepregnancy BMI (kg/m2) | <18.5 |
18.5–<25 | |
25–<30 | |
≥30 | |
Paternal BMI (kg/m2) | <18.5 |
18.5–<25 | |
25–<30 | |
≥30 | |
Mother smoked in pregnancy | No |
Yes | |
Ever breastfed | Yes |
No |
Change in weight-for-age z-score from birth to assessment time using the World Health Organization 2006 Child Growth Standards.
SD, standard deviation.
Outcome Measure
The primary clinical outcome measure for the validation study was overweight in childhood defined by the International Obesity Task Force (IOTF)14 sex- and age-specific cutoffs, which correspond to an adult BMI ≥25 kg/m2. In addition, we compared IOTF overweight criteria14 assessed at 5 years to UK national guidelines4 using the UK 1990 growth reference,15 which defines clinical overweight based on the 91st centiles, respectively. Both criteria were used to ensure that the IROC can also be interpreted in a broader context.
Statistical Analysis
Data from all children in the ALSPAC CiF cohort on the seven risk factors utilized in the IROC algorithm were extracted to calculate risk and predict actual outcome of being overweight at 5 years. A total of four models were developed and analyzed: (1) a clinical model, which uses the original algorithm and assigns null values to missing data on risk factors; (2) a recalibrated model, which uses multivariate logistic regression to generate and apply a recalibrated algorithm to better reflect the demographics of the ALSAPC CiF cohort; (3) an imputed model, using multiple imputation16 where 10 copies of the existing data set were generated to predict missing risk factor from multivariate models (assuming that the data were “missing-at-random”); and (4) a recalibrated imputed model, applying the recalibrated algorithm to the imputed data set where there are no missing risk factor data. Discrimination, a measure of predictive accuracy defined by the ability to distinguish a case from a noncase, for each of the four models was assessed by the AUC c-statistic. This calculates the probability that the IROC predictive risk score is higher for overweight children than for those who are healthy weight or underweight. The AUC can range from 0.5 (algorithm predicts based on complete random chance) to 1.0 (algorithm has perfect predictive accuracy). Models with an AUC of greater than 0.70 for predicting clinical outcomes are seen to be valid for use in practice, with consideration given to the balance of benefits of prevention against the harms of treatment.17 To generate 95% confidence intervals (CIs) around AUC c-statistic, standard errors were bootstrapped using a jackknife procedure.18 To test the effect of various risk thresholds for identifying “high-risk,” the net reclassification index (NRI) was calculated for risk thresholds ranging from 2.5% to 30% for the primary clinical model.19 The NRI in this study can be interpreted as the net change in percentage of children correctly reclassified from an initial baseline risk threshold of 2.5%.19 This is calculated as: NRI = (net increase of classification for cases/total number of cases) + (net decrease of classification of non-cases/total number of non-cases) × 100. All analyses were conducted using STATA software (13 MP; StataCorp LP, College Station, TX), imputation using the Multiple Imputation Chained Equations (MICE) mim module16; calibration using multivariate logistic regression (logit command) of IROC risk factors against the overweight outcome; discrimination analysis generating AUC and associated CIs by the somersd module18; and net reclassification analysis using diagt module.20
Results
Study Population
At 5 years of age, 12.3% of boys and 19.6% of girls were categorized overweight by IOTF standards (Table 2). The proportion of children who were overweight increased as birth weight increased, with the exception of children with birth weights less than 2.93 kg. Twenty-four percent of children who experienced rapid weight gain in the first 4 months were considered overweight at 5 years, whereas only 13.3% of children who had not experienced rapid weight gain were considered overweight. There was an overall directly proportional relationship between maternal and paternal BMI and childhood overweight apart from the lowest BMI category <18.5, possibly attributable to the low numbers of parents in that category. Mothers who smoked in pregnancy also had a high proportion of overweight children compared to those who did not smoke, whereas breastfeeding resulted in a lower proportion of children who were overweight compared with not breastfeeding. Missing values ranged from 0.70% (n = 7) for birth weight to 38.9% (n = 381) for paternal BMI. Additional analysis on the relationship between IROC risk factors and overweight status at 5 years is shown as a multivariate logistic model provided in Supplementary Table 1 (see online supplementary material at http://www.liebertpub.com).
Table 2.
Risk factors | N (%) | Mean BMI (SD) | % Overweight (IOTF)a | % Overweight (UK 1990)b |
---|---|---|---|---|
Sex | ||||
Male | 535 (54.6) | 16.0 (1.4) | 12.3 | 12.2 |
Female | 445 (45.4) | 16.1 (1.6) | 19.6 | 11.5 |
Birth weight (kg) | ||||
<2.93 | 136 (13.9) | 15.7 (1.5) | 11.8 | 8.1 |
2.93–<3.24 | 173 (17.7) | 15.7 (1.3) | 11.0 | 7.5 |
3.24–<3.49 | 201 (20.5) | 16.0 (1.5) | 13.4 | 9.5 |
3.49–<3.81 | 242 (24.7) | 16.2 (1.6) | 17.4 | 13.6 |
≥3.81 | 221 (22.5) | 16.4 (1.6) | 22.2 | 18.1 |
Missing | 7 (0.70) | 15.3 (0.9) | 0.0 | 0.0 |
Infant weight gainc (0 to 4 months) | ||||
≤0.67 SD | 624 (63.7) | 15.9 (1.4) | 13.3 | 9.9 |
>0.67 SD | 76 (7.8) | 16.8 (1.7) | 23.7 | 18.4 |
Missing | 280 (28.5) | 16.1 (1.5) | 18.5 | 14.3 |
Maternal prepregnancy BMI (kg/m2) | ||||
<18.5 | 37 (3.8) | 15.7 (1.4) | 13.5 | 8.1 |
18.5–<25 | 541 (55.2) | 15.8 (1.3) | 11.7 | 8.1 |
25–<30 | 117 (11.9) | 16.4 (1.5) | 24.8 | 17.9 |
≥30 | 47 (4.8) | 17.1 (2.1) | 44.7 | 38.3 |
Missing | 238 (24.3) | 16.1 (1.6) | 14.7 | 12.6 |
Paternal BMI (kg/m2) | ||||
<18.5 | 5 (0.50) | 15.9 (1.9) | 40.0 | 0.0 |
18.5–<25 | 312 (31.8) | 15.7 (1.4) | 9.3 | 6.1 |
25–<30 | 230 (23.5) | 16.3 (1.5) | 20.4 | 14.8 |
≥30 | 52 (5.3) | 16.7 (1.7) | 26.9 | 23.1 |
Missing | 381 (38.9) | 16.0 (1.5) | 16.0 | 13.4 |
Mother smoked in pregnancy | ||||
No | 816 (83.3) | 16.0 (1.5) | 14.7 | 11.3 |
Yes | 144 (14.7) | 16.2 (1.7) | 20.1 | 13.9 |
Missing | 20 (2.0) | 16.1 (1.4) | 20.0 | 20.0 |
Ever breastfed | ||||
Yes | 761 (77.7) | 16.0 (1.5) | 15.4 | 11.3 |
No | 180 (18.3) | 16.1 (1.6) | 18.3 | 13.9 |
Missing | 39 (4.0) | 15.6 (1.1) | 7.7 | 20.0 |
Determined by the International Obesity Task Force sex- and age-specific cutoffs.
Determined by UK 1990 sex- and age-specific growth charts.
Change in weight-for-age z-score from birth to assessment time using the World Health Organization 2006 Child Growth Standards.
SD, standard deviation; IOTF, International Obesity Task Force.
External Validation
The primary clinical model (Model 1), which directly applied the original IROC algorithm to the ALSPAC cohort of children, resulted in a moderate discrimination using both IOTF (AUC c-statistic, 0.67; 95% CI, 0.62–0.72) and UK 1990 (AUC c-statistic, 0.65; 95% CI, 0.60–0.71) overweight criteria (Table 3). In Model 2, the algorithm was simply recalibrated reflecting the ALSPAC population characteristics. Compared to Model 1, recalibration (Model 2) increased the discrimination by an average of 3% and 2% based on IOTF and UK 1990 overweight criteria, respectively. Compared to Model 1, multiple imputation (Model 3) increased the discrimination by an average of 12% and 8% based on IOTF and UK 1990 overweight criteria, respectively. Moreover, if both recalibration and multiple imputation were used in combination (Model 4), an average of 26% and 25% increase in discrimination could be achieved, based on IOTF and UK 1990 overweight criteria, respectively. Further analysis on the IROC algorithm performance compared to single risk factors can be found in Supplementary Figure 1 (see online supplementary material at http://www.liebertpub.com).
Table 3.
Models | IOTF Overweight Criteria AUC c-statistic (95% CI)+ | UK 1990 Overweight Criteria AUC c-statistic (95% CI)a |
---|---|---|
Primary analysis | ||
1. Clinical model | 0.67 (0.62–0.72) | 0.65 (0.60–0.71) |
Sensitivity analysis | ||
2. Recalibrated model | 0.70 (0.65–0.74) | 0.67 (0.62–0.72) |
3. Imputed model | 0.79 (0.72–0.86) | 0.73 (0.65–0.81) |
4. Recalibrated imputed model | 0.93 (0.88–0.98) | 0.90 (0.84–0.96) |
95% confidence intervals generated by bootstrapping standard errors using jackknife procedure.
IROC, Infant Risk of Overweight Checklist; IOTF, International Obesity Task Force; AUC, area under the receiver operating curve; CI, confidence interval.
Risk Threshold Analysis
The risk threshold for determining “high-risk” increased as the number of cases of children who were correctly identified decreased (sensitivity: true positive rate), whereas the number of noncases correctly identified (specificity: true negative rate) increased (Table 4).
Table 4.
Risk threshold | No. of cases correctly identified (% of total cases) | No. of noncases correctly identified (% of total noncases) | Net reclassification index (NRI)a |
---|---|---|---|
IOTF defined overweight | |||
2.5% | 151 (0.99) | 39 (0.05) | — |
5% | 138 (0.90) | 197 (0.24) | 11 |
10% | 81 (0.53) | 587 (0.71) | 21 |
15% | 37 (0.24) | 761 (0.92) | 13 |
20% | 19 (0.12) | 806 (0.98) | 6 |
25% | 6 (0.04) | 821 (0.99) | 0 |
30% | 4 (0.03) | 827 (1.00) | −11 |
UK 1990 defined overweight | |||
2.5% | 114 (0.98) | 39 (0.05) | — |
5% | 104 (0.90) | 200 (0.23) | 10 |
10% | 60 (0.52) | 603 (0.70) | 19 |
15% | 30 (0.26) | 791 (0.92) | 15 |
20% | 16 (0.14) | 840 (0.97) | 8 |
25% | 4 (0.03) | 856 (0.99) | 0 |
30% | 2 (0.02) | 862 (1.00) | −1 |
NRI is the net change in percentage of individuals correctly classified from a baseline risk threshold of 2.5%.
IROC, Infant Risk of Overweight Checklist; ALSPAC, Avon Longitudinal Study of Parents and Children; IOTF, International Obesity Task Force.
The NRI showed that there were positive gains in reclassification anywhere from 5% to 20%, with the maximum NRI being at 10%. This can be interpreted as the optimal point of balance between sensitivity and specificity. At this 10% threshold using IOTF defined overweight, 53% (n = 81) of all overweight children were correctly identified whereas 71% (n = 587) of all healthy weight children were correctly ruled out, resulting in a maximum NRI of 21. Using UK 1990 defined overweight at a 19% threshold, 52% (n = 60) of all overweight children were correctly identified whereas 70% (n = 603) of all healthy weight children were correctly ruled out, resulting in a maximum NRI of 19.
Discussion
The results confirm that the IROC algorithm is a valid measure of risk, predicting overweight in childhood up to 5 years of age. The IROC algorithm incorporates seven factors that are routinely recorded or can easily be assessed in clinical practice. Although some risk factors in the ALSPAC were missing or undocumented, the algorithm remains clinically valid by assuming the average population risk for those particular missing risk factors. Even with missing values and no recalibration of the original algorithm to reflect target population, the IROC can achieve moderate prediction accuracy. With a simple recalibration, this accuracy improves. With complete and accurate information for all the predictor variables, extremely high levels of predictive accuracy could be achieved. This study has also determined that the optimal risk threshold for determining “high-risk” based on a balance between sensitivity and specificity is 10% using ALSPAC, a large UK-based longitudinal population study. Although several prediction algorithms21–24 exist, the significant advantage of the IROC is that it has proved valid in a heterogeneous UK population-based cohort, using parameters identified in a systematic review of the available literature.1
Clinical Implications
Studies have shown that UK healthcare professionals are often unsure whether and how to intervene with infants who gain weight rapidly.5,6 Similar studies in the United States have found that clinicians only diagnose overweight or obesity in 1.1–31% of all overweight children, leading to suboptimal levels of advice given and failure to refer to appropriate interventions.25,26 Although it is possible that “at-risk” infants could be identified solely by rapid weight gain/growth27–29 during the first few months of life, this will not account for interactions between, for example, birth weight and weight gain30 or the protective effects of breastfeeding.1 This could increase the number of false-positive cases resulting in unnecessary intervention. The IROC algorithm has the potential to increase identification of infants at risk, whereas the high specificity and corresponding NPV of the algorithm indicates a relatively low chance of engaging parents unnecessarily. IROC is designed to assess overweight in infants as young as 4 months, thus offering a valuable opportunity for effective early intervention. Counseling on weight status of preschool children results in parents making positive lifestyle changes.31,32 There is also some evidence from randomized, controlled trials of interventions delivered during infancy, typically including parental support, nutrition modification, healthy eating, breastfeeding, and weaning advice that it is possible to reduce obesity risk during infancy.33–38
ProAsk Tool
Confirmation of the validity of the IROC and an optimal cut-off threshold for overweight risk enables potential implementation in a clinical setting. The IROC algorithm (Model 1) has been embedded within an interactive digital program named ProAsk (Proactive Assessment of Obesity Risk during Infancy) for use on a hand-held tablet device. ProAsk is designed for clinical use with parents and prompts them to enter information on the seven risk factors used by the IROC algorithm. If information on any of the risk factors is unavailable, the program defaults to the average population risk of the missing risk factor. A percentage risk is calculated using the original algorithm16 and, based upon the 10% risk threshold determined by the current study, gives text-based feedback of risk of overweight compared to other infants. Rather than using the terminology “high-risk,” as is appropriate in this technical report, “above-average” risk is employed. ProAsk prompts the health professional to share this information with the parent with supporting text to aid appropriate, nonstigmatizing language. The ProAsk program then directs the healthcare professionals to a therapeutic wheel based on a recent systematic review,38 which promotes evidence-based behavior change strategies in four areas: active play; milk and solid foods; sleeping and soothing; and infant feeding cues. Parents are encouraged to identify one area and coproduce solution-focused, yet evidence-based, strategies for behavior change with health professionals. ProAsk draws on the extended Health Belief model of behavior change to increase parents' perceived susceptibility to the risk of child obesity and promote self-efficacy for behavior change.39 Importantly, the use of a digital tool supports the healthcare professional to communicate infant obesity risk in a personalized, nonstigmatizing manner.
There is currently very little research guidance about the most appropriate way for health professionals to assess and communicate infant risk of overweight to parents in an effective, nonjudgemental, and sensitive way. Interactive digital technology has the potential to deliver personalized information about health risks in a way that increases understanding of risk and facilitates appropriate behavior change without increasing patient anxiety.40 Evidence from qualitative research suggests that general practitioners appreciate the options offered by tailored digital technology for health promotion in primary care,41 and dietitians have identified a need for digital resources to support their communication with children and parents around the sensitive topic of child obesity.42 Although the results of a large systematic review suggest that health information technology can promote and support patient-centred care,43 the potential of digital technology to enhance communication in child-focused clinical settings has been largely neglected.42
The MCS and ALSPAC databases both experienced higher attrition rates from study participants from lower socioeconomic backgrounds leading to missing data. There is no current research evidence that demonstrates the best way to collect data on obesity risk factors in clinical practice. A multifaceted approach is likely to be needed, such as face-to-face data collection during a clinical consultation, by a questionnaire sent to parents, supplemented through access to medical records.44 For instance, risk factors such as parental BMI could be obtained from consenting participants through their general practice computer records.44 A study funded by the UK Medical Research Council Public Health Intervention Development Scheme (MRC-PHIND PH01/14-15) is exploring the feasibility of health visitors (public health nurses), using ProAsk to identify infants at increased risk of obesity, and to communicate that risk in conjunction with brief targeted health advice. The ProAsk study is being conducted in areas of high deprivation to determine the feasibility of collecting these data clinically as well as the acceptability of the approach.
Limitations
There were several limitations of the study. The most substantial contributor toward missing data was paternal BMI attributed to lack of availability, or single-parent status. Although the sensitivity analysis performed in this study shows that the effects of missing values may be marginal, the missing values have the potential to generate greater variability in the observed overweight outcomes. This process allows assessment of all children despite missing data. Imputation techniques allow for the efficient analysis of all children in the cohort despite having missing data. This method of accounting for missing data is a robust, widely accepted45 technique to improve the efficiency of the analysis and validity of the results. Additionally, the overweight outcomes assessed were limited to BMI and a relatively short follow-up duration (5 years of age). Although there is a strong association between childhood overweight and adult obesity,46 it is still unknown whether identifying infants at risk of overweight would extend to adulthood limited by the 5-year follow-up duration.
Conclusions
This study has confirmed that the IROC algorithm has moderately good validity for assessing overweight risk in infants as early as 4 months. It has also determined the optimal threshold for determining “high-risk” infants, enabling services to prioritize those at greatest risk. Interactive digital technology (ProAsk) has the potential to facilitate the identification of “high-risk” infants in clinical settings, support the sensitive communication of this information to parents, and empower parents to make evidence-based, solution-focused behavioral changes.
Supplementary Material
Acknowledgments
The authors are extremely grateful to all the families who took part in this study, the midwives for their help in recruiting them, and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists, and nurses. The UK Medical Research Council, the Wellcome Trust, and the University of Bristol provide core support for ALSPAC. This publication is the work of the authors (S.A.R., S.F.W., D.N., J.A.S., and C.G.) and will serve as guarantors for the contents of this article and does not reflect the views of the ALSPAC executive. This research was specifically funded by the UK National Institute for Health Research Collaboration for Leadership in Applied Health Research and Care–Nottinghamshire, Derbyshire and Lincolnshire (NIHR CLAHRC-NDL).
Author Disclosure Statement
No competing financial interests exist.
References
- 1.Weng SF, Redsell SA, Swift JA, et al. . Systematic review and meta-analyses of risk factors for childhood overweight identifiable during infancy. Arch Dis Child 2012;97:1019–1026 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Department of Health. Healthy Child Programme: Pregnancy and the first five years of life. Department of Health: London, 2009 [Google Scholar]
- 3.Institute of Medicine. Early childhood obesity prevention policies. Institute of Medicine: Washington, DC, 2011 [Google Scholar]
- 4.National Institute of Health and Care Excellence. Managing overweight and obesity among children and young people: Lifestyle weight management services. National Institute of Health and Care Excellence: London, 2013 [Google Scholar]
- 5.Redsell S, Atkinson P, Nathan D, et al. . Preventing childhood obesity during infancy in UK primary care: A mixed-methods study of HCPs' knowledge, beliefs and practice. BMC Fam Pract 2011;12:54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Redsell S, Swift JA, Nathan D, et al. . UK health visitors' role in identifying and intervening with infants at risk of developing obesity. Matern Child Nutr 2013;9:396–408 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Weng SF, Redsell SA, Nathan D, et al. . Estimating overweight risk in childhood from predictors during infancy. Pediatrics 2013;132:e414–e421 [DOI] [PubMed] [Google Scholar]
- 8.Millenium Cohort Study, First, Second, Third, and Fourth Surveys. London, UK, Centre for Longitudinal Studies. Institute of Education, University of London, 2010 [Google Scholar]
- 9.Fraser A, Macdonald-Wallis C, Tilling K, et al. . Cohort profile: The Avon Longitudinal Study of Parents and Children: ALSPAC mothers cohort. Int J Epidemiol 2013;42:97–110 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Boyd A, Golding J, Macleod J, et al. . Cohort profile: The ‘children of the 90s’—The index offspring of the Avon Longitudinal Study of Parents and Children. Int J Epidemiol 2013;42:111–127 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Beyerlein A, Ness AR, Streuling I, et al. . Early rapid growth: No association with later cognitive functions in children born not small for gestational age. Am J Clin Nutr 2010;92:585–593 [DOI] [PubMed] [Google Scholar]
- 12.Mihrshahi S, Battistutta D, Magarey A, et al. . Determinants of rapid weight gain during infancy: Baseline results from the NOURISH randomised controlled trial. BMC Pediatr 2011;11:99. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Ong K, Loos R. Rapid infancy weight gain and subsequent obesity: Systematic reviews and hopeful suggestions. Acta Paediatr 2006;95:904–908 [DOI] [PubMed] [Google Scholar]
- 14.Cole TJ, Bellizzi MC, Flegal KM, et al. . Establishing a standard definition for child overweight and obesity worldwide: International survey. BMJ 2000;320:1240–1243 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Cole TJ, Freeman JV, Preece MA. Body mass index reference curves for the UK, 1990. Arch Dis Child 1995;73:25–29 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Royston P WI. Multiple Imputation by Chained Equations (MICE): Implementation in Stata. J Stat Softw 2011;45:1–20 [Google Scholar]
- 17.Traeger A, Henschke N, Hübscher M, et al. . Development and validation of a screening tool to predict the risk of chronic low back pain in patients presenting with acute low back pain: A study protocol. BMJ Open 2015;5:e007916. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Newson R. Confidence intervals for rank statistics: Somers' D and extensions. Stata J 2006;6:309–334 [Google Scholar]
- 19.Leening MJ, Vedder MM, Witteman JC, et al. . Leening MJG VM, Witterman JCM, Pencina MJ, Steyerberg EW. Net reclassification improvement: Computation, interpretation, and controversies. Ann Intern Med 2014;160:122–131 [DOI] [PubMed] [Google Scholar]
- 20.Seed P. DIAGT: Stata module to report summary statistics for diagnostic test compared to true disease status. My Ideas. 2010. Available at http://ideasrepecorg/c/boc/bocode/s423401html Last accessed December1, 2015
- 21.Druet C, Stettler N, Sharp S, et al. . Prediction of childhood obesity by infancy weight gain: An individual-level meta-analysis. Paediatr Perinat Epidemiol 2012;26:19–26 [DOI] [PubMed] [Google Scholar]
- 22.Morandi A, Meyre D, Lobbens S, et al. . Estimation of newborn risk for child or adolescent obesity: Lessons from longitudinal birth cohorts. PLoS One 2012;7:e49919. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Santorelli G, Petherick ES, Wright J, et al. . Developing prediction equations and a mobile phone application to identify infants at risk of obesity. PLoS One 2013;8:e71183. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Dahly D, Rudolf M. Identifying obesity risk in the early years. Cross Government Obesity Unity: Aylesbury, UK, 2010 [Google Scholar]
- 25.Riley MR, Bass NM, Rosenthal P, et al. . Underdiagnosis of pediatric obesity and underscreening for fatty liver disease and metabolic syndrome by pediatricians and pediatric subspecialists. J Pediatr 2005;147:839–842 [DOI] [PubMed] [Google Scholar]
- 26.Woolford SJ, Gebremariam A, Clark SJ, et al. . Incremental hospital charges associated with obesity as a secondary diagnosis in children. Obesity 2007;15:1895–1901 [DOI] [PubMed] [Google Scholar]
- 27.Monteiro PO, Victora CG. Rapid growth in infancy and childhood and obesity in later life–a systematic review. Obes Rev 2005;6:143–154 [DOI] [PubMed] [Google Scholar]
- 28.Ong KK, Loos RJ. Rapid infancy weight gain and subsequent obesity: systematic reviews and hopeful suggestions. Acta Paediatr 2006;95:904–908 [DOI] [PubMed] [Google Scholar]
- 29.Stettler N, Zemel BS, Kumanyika S, Stallings VA. Infant weight gain and childhood overweight status in a multicenter, cohort study. Pediatrics 2002;109:194–199 [DOI] [PubMed] [Google Scholar]
- 30.Yeung MY. Postnatal growth, neurodevelopment and altered adiposity after preterm birth—From a clinical nutrition perspective. Acta Pædiatrica 2006;95:909–917 [DOI] [PubMed] [Google Scholar]
- 31.Hernandez RG, Cheng TL, Serwint JR. Parents' healthy weight perceptions and preferences regarding obesity counseling in preschoolers: Pediatricians matter. Clin Pediatr 2010;49:790–798 [DOI] [PubMed] [Google Scholar]
- 32.Rhee KE, De Lago CW, Arscott-Mills T, et al. . Factors associated with parental readiness to make changes for overweight children. Pediatrics 2005;116:e94–e101 [DOI] [PubMed] [Google Scholar]
- 33.Campbell KJ, Lioret S, McNaughton SA, et al. . A parent-focused intervention to reduce infant obesity risk behaviors: A randomized trial. Pediatrics 2013;131:652–660 [DOI] [PubMed] [Google Scholar]
- 34.Daniels LA, Mallan KM, Battistutta D, et al. . Evaluation of an intervention to promote protective infant feeding practices to prevent childhood obesity: Outcomes of the NOURISH RCT at 14 months of age and 6 months post the first of two intervention modules. Int J Obes (Lond) 2012;36:1292–1298 [DOI] [PubMed] [Google Scholar]
- 35.Daniels LA, Mallan KM, Nicholson JM, et al. . Outcomes of an early feeding practices intervention to prevent childhood obesity. Pediatrics 2013;132:e109–e118 [DOI] [PubMed] [Google Scholar]
- 36.Wen LM, Baur LA, Simpson JM, et al. . Effectiveness of an early intervention on infant feeding practices and “tummy time”: A randomized controlled trial. Arch Pediatr Adolesc Med 2011;165:701–707 [DOI] [PubMed] [Google Scholar]
- 37.Wen LM, Baur LA, Simpson JM, et al. . Effectiveness of home based early intervention on children's BMI at age 2: Randomised controlled trial. BMJ 2012;344:e3732. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Redsell S, Edmonds B, Nathan D, et al. . Systematic review of behavioural and non-behavioural interventions, delivered during infancy and the antenatal period, that reduce the risk of overweight and obesity in childhood. Matern Child Nutr 2016;12:24–38 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Strecher V, Rosenstock I. The Health Belief Model. Jossey-Bass: San Francisco, CA, 1997 [Google Scholar]
- 40.Glazebrook C, Garrud P, Avery A, et al. . Impact of a multimedia intervention “Skinsafe” on patients' knowledge and protective behaviors. Prev Med 2006;42:449–454 [DOI] [PubMed] [Google Scholar]
- 41.Plaete J, Crombez G, DeSmet A, et al. . What do general practitioners think about an online self-regulation programme for health promotion? Focus group interviews. BMC Fam Pract 2015;16:3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Raaff C, Glazebrook C, Wharrad H. A systematic review of interactive multimedia interventions to promote children's communication with health professionals: Implications for communicating with overweight children. BMC Med Inform Decis Mak 2014;18:8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Finkelstein J, Knight A, Marinopoulos S, et al. . Enabling patient-centered care through health information technology. Evid Rep Technol Assess (Full Rep) 2012;(206):1–1531 [PMC free article] [PubMed] [Google Scholar]
- 44.Bhaskaran K FHJ, Douglas I, Leon D.A, Smeeth L. Representativeness and optimal use of body mass index (BMI) in the UK Clinical Practice Research Datalink (CPRD). BMJ Open 2013;3:e003389. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Sterne JA, White IR, Carlin JB, et al. . Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. BMJ 2009;338:b2393. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Ekelund U, Ong K, Linne Y, et al. . Upward weight percentile crossing in infancy and early childhood independently predicts fat mass in young adults: The Stockholm Weight Development Study (SWEDES). Am J Clin Nutr 2006;83:324–330 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.