How well can poor child health and development be predicted by data collected in early childhood?

Background Identifying children at risk of poor developmental outcomes remains a challenge, but is important for better targeting children who may benefit from additional support. We explored whether data routinely collected in early life predict which children will have language disability, overweight/obesity or behavioural problems in later childhood. Methods We used data on 10262 children from the UK Millennium Cohort Study (MCS) collected at 9 months, 3, and 11 years old. Outcomes assessed at age 11 years were language disability, overweight/obesity and socioemotional behavioural problems. We compared the discriminatory capacity of three models: (1) using data currently routinely collected around the time of birth; (2) Model 1 with additional data routinely collected at 3 years; (3) a statistically selected model developed using a larger set of early year's risk factors for later child health outcomes, available in the MCSbut not all routinely collected. Results At age 11, 6.7% of children had language disability, 26.9% overweight/obesity and 8.2% socioemotional behavioural problems. Model discrimination for language disability was moderate in all three models (area under the curve receiver-operator characteristic 0.71, 0.74 and 0.76, respectively). For overweight/obesity, it was poor in model 1 (0.66) and moderate for model 2 (0.73) and model 3 (0.73). Socioemotional behavioural problems were also identified with moderate discrimination in all models (0.71; 0.77; 0.79, respectively). Conclusion Language disability, socioemotional behavioural problems and overweight/obesity in UK children aged 11 years are common and can be predicted with moderate discrimination using data routinely collected in the first 3years of life.