We aim to develop and validate a set of ACE indicators, and to study the impact of ACEs on future health outcomes and healthcare use in a birth cohort of mothers and children (born 1990-2018). The cohort will comprise the CPRD mother-baby link (MBL), with patients followed from birth to adulthood, and linked to the Hospital Episode Statistics (HES), Index of Multiple Deprivation (IMD), and ONS for mortality data. The project is divided into three parts:
Part I: Refinement and validation
(1) An exploratory study, to refine and classify ACE indicators in CPRD and HES using supervised machine learning techniques such as random survival forests.
(2) A cross-validation study, to quantify the overlap and agreement between ACE indicators across HES and CPRD. We will calculate prevalence's, positive predictive values and cross concordances using methods such as the Cohens Kappa statistic to estimate agreement between sources.
(3) A derivation and validation study, using prediction models investigating the association between clinically relevant ACE indicators (e.g. predictors amenable to intervention during a limited time period) and the risk of future adversity (e.g. re-occurrence of child maltreatment). Predictors will be modelled using supervised machine learning techniques, combined with manual multivariable hazard regression models.
Part II: Descriptive study
(4) A descriptive study, to estimate prevalence and, annual incidence rates and attributable economic costs of ACEs in CPRD and HES. Estimates will be stratified by demographics, IMD and overall ACE type.
Part III: Hypothesis testing study
(5) An hypothesis-testing birth-cohort study, investigating whether ACEs increase the risk of poor health outcomes among mothers and children later in life, compared to unexposed mother-child pairs. ACEs will be modelled using multivariable cox-proportional hazard models, and outcomes will include psychiatric disorders, chronic conditions, potentially preventable health events (e.g. emergency injury admissions), and all/cause-specific mortality.
Primary outcomes as recorded in maternal/child CPRD/HES-APC/ONS records:
Child maltreatment (CM), all-cause and cause-specific mortality;
Adverse childhood experiences/environments (ACEs; e.g. child maltreatment, social service involvement, maternal mental health problems, and other adverse family environments).
Secondary outcomes as recorded in maternal/child CPRD/HES-APC records:
Re-occurrence of ACEs; frequency of ACEs
All-cause hospitalisations; cause-specific hospitalisations;
ACEs and secondary outcomes can be considered as exposures interchangeably, with descriptions provided in section N.
To ascertain applicable code lists, we will prioritise existing phenotypes and validated algorithms. Additional Read and ICD codes will be identified using free-text searches of code dictionaries and by searching online code repertoires. If not already classified, ICD-10 codes will be categorised into broad diagnosis groups using the Clinical Classifications Software (CCS developed by the Agency for Healthcare Research and Quality (AHRQ). The cross-map package by NHS digital will be used to guide combinations of Read and ICD-10 codes for classification of similar diagnoses across sources. The final list of ACEs and health conditions will reflect a refined set of indicators clinically relevant or of public health importance based on prevalence, that have associations with risk factors and outcomes consistent with external evidence and input from a patient and expert stakeholder group co-ordinated by NIHR Children and Families Policy Research Unit (CPRU)
Ruth Gilbert - Chief Investigator - University College London ( UCL )
Shabeer Syed - Corresponding Applicant - University College London ( UCL )
Ania Zylbersztejn - Collaborator - University College London ( UCL )
Arturo Gonzalez-Izquierdo - Collaborator - University College London ( UCL )
Claire Powell - Collaborator - University College London ( UCL )
Estela Capelas Barbosa - Collaborator - University College London ( UCL )
Harry Hemingway - Collaborator - University College London ( UCL )
Heng Fan - Collaborator - University College London ( UCL )
Janice Allister - Collaborator - Not from an Organisation
Katie Harron - Collaborator - University College London ( UCL )
Leah Li - Collaborator - University College London ( UCL )
Linda Wijlaars - Collaborator - University College London ( UCL )
Pia Hardelid - Collaborator - University College London ( UCL )
Rebecca Lacey - Collaborator - University College London ( UCL )
Spiros Denaxas - Collaborator - University College London ( UCL )