Close
Supervisors: Dr Leah Li, Professor Ruth Gilbert
A machine learning approach to study life course patterns of maternal adverse experiences and the occurrence of maltreatment in offspring using linked UK administrative data
Background:
Repeated exposure to adversity in childhood, such as maltreatment (abuse and neglect) and household dysfunction (e.g. domestic conflict, witness abuse of family members, parental mental illness) have been associated with increased risk of premature death and a range of health outcomes later in life (1). In particular, maternal adverse experiences (i.e. interpersonal violence, mental health problems, and substance misuse) and child maltreatment often co-occur and share many common risk factors, e.g. young motherhood and lower socio-economic position (SEP). For example, children who witnessed intimate partner violence at home were ~4 times more likely to experience maltreatment than children who did not (2). Child maltreatment and household dysfunction also tend to co-occur (3). However, little is known about the longitudinal association between maternal adverse experiences and child maltreatment in the offspring.
Aims and Methods:
This PhD project aims to explore the life course patterns of maternal adverse experiences and the occurrence of maltreatment in the offspring, and importantly, the correlations/links among them.
The project will use the data of over a million children and their mothers registered in the UK CPRD mother-baby link database, which includes patients followed from birth onwards, and is linked to the Hospital Episode Statistics (HES), Index of Multiple Deprivation (IMD), and ONS for mortality data. The primary outcomes are repeated measures of maternal adverse experiences and child maltreatment, and these will be identified from multiple sources, such as read codes, prescriptions, clusters of ICD codes, social service referrals, or self-report instruments in the database (CPRD/HES) (4).
Specifically, child maltreatment will be based on any recording in children or mothers of neglect, physical, emotional/psychological and sexual abuse by a parent/caregiver, any code relating to social service involvements/referrals, or codes indicating high risk of maltreatment. Maternal adverse experiences will include mental health problems (psychiatric disorders, depression, or above cut-offs of validated self-report instruments), risky behaviours (codes indicating alcohol and substance misuse), and interpersonal violence (codes indicating domestic violence). Machine learning techniques will be used to capture the cases and also high risk groups (likely cases) of mother and children who had repeated adverse experiences and maltreatments. Other important factors will also be considered, such as birth year, age of the child, prenatal exposure to drugs and alcohol, teen/single motherhood, lower SEP (using IMD as a proxy), and frequency of GP visits and hospital admissions. Patterns of trajectories for maternal interpersonal violence, substance misuse and mental health problems, and maltreatment in offspring will be studied simultaneously using joint modelling or latent class analysis to account for their correlations.
References:
1. Norman RE, Byambaa M, et al. The long-term health consequences of child physical abuse, emotional abuse, and neglect: a systematic review and meta-analysis. PLoS medicine. 2012; 9(11):e1001349.
2. Hamby S, Finkelhor D, et al. The overlap of witnessing partner violence with child maltreatment and other victimizations in a nationally representative survey of youth. Child abuse & neglect. 2010;34:734-41.
3. Denholm R, Power C, Thomas C, Li L. Child maltreatment and household dysfunction in a British Birth Cohort. Child Abuse Review. 2013;22(5):14.
4. Syed S, Ashwick R, Schlosser M, Gonzalez-Izquierdo A, Li L, Gilbert R. Predictive value of indicators for identifying child maltreatment and intimate partner violence in coded electronic health records: a systematic review and meta-analysis. Archives of Disease in Childhood 2020; In press.
