Predicting multiple long-term conditions with role limitation at age 46 using early-life data from the 1970 British...

There is minimal research investigating whether multiple long-term conditions (MLTCs) can be predicted using early-life data. We examined whether the risk of later-life MLTCs that impact everyday functioning can be predicted from factors recorded at birth and age 5.

Data from 5007 participants aged 46 in the 1970 British Cohort Study were analysed. The outcome was two or more self-reported MLTCs with/without role limitation (ie, impacting everyday life functioning) due to a physical or mental health condition. Backwards stepwise logistic regression was used as a variable selection approach to identify early-life predictors that improved model discrimination for four models: (1) age 5 predictors only and MLTC with role limitation, (2) age 5 predictors only and MLTC without role limitation, (3) age 5 and birth predictors and MLTC with role limitation and (4) age 5 and birth predictors and MLTC without role limitation. Model discrimination was assessed using the area under the curve (AUC).

11.9% of the sample reported MLTCs with role limitation. The set of variables selected differed across models, reflecting differences in which predictors most improved predictive discrimination for each outcome. The highest AUC (0.66) was for model 3 including age 5 and birth predictors of MLTCs with role limitation at age 46. Retained age 5 predictors included the number of immunisations, sex, Rutter behaviour score, maternal malaise index (an indication of psychological distress), asthma diagnosis, if the child was seen by a doctor/nurse in school, whether the parent viewed their child as equals, and if the parent believed the child should have unquestioning obedience towards them. Retained birth predictors included: maternal smoking, gestational age, parity and if the mother ever had a teenage pregnancy.

Using early-life birth cohort data from birth and age 5 did not predict MLTCs with role limitation at age 46 well. Exploration of predictors during other time points in childhood could better inform preventative interventions.