Introduction to Epidemiological Methods in Life Course Research PDF

Introduction to Epidemiological Methods in Life course Research Dr Tom McAdams  some key concepts in life course epidemiology Lecture study designs outline analytic strategies relevant in longitudinal/life course investigations measuri...

Introduction to Epidemiological Methods in Life course Research Dr Tom McAdams  some key concepts in life course epidemiology Lecture study designs outline analytic strategies relevant in longitudinal/life course investigations measuring mental health problems and disorders in large-scale studies Life course Epidemiology long-term effects on health of adverse exposures across the life-span: gestation, childhood, adolescence, young adulthood, and later adult life - and across generations Initially developed in relation to chronic physical health problems (Barker hypothesis) also appropriate for mental health: risk factors occur across the lifespan, from gestation onwards onsets and trajectories of disorders show systematic patterning by age interest in the importance of biological and social transitions in the life course on risk for disorder outcomes include impacts on a range of health, economic, and social difficulties Key concepts: models for understanding the impact of early risks critical period (biological programming) exposure in critical developmental period affects structure/function of organism sensitive period exposures have greatest effect in periods of rapid development – but more scope for modification accumulation of risk adverse exposures accumulate over life course chains/cascades of risk sequence of exposures early and later exposures unlikely to be independent – early risks may increase the likelihood of later exposures Life course causal models of risk Pathways to symptoms of depression and anxiety: women in the 1946 British birth cohort Study designs for life course epidemiology life course study: ‘a cohort study that has information from at least one early stage of development and in adult life’ key design: pregnancy/birth cohort studies representative samples born close in time, and followed prospectively over time Britain has a rich tradition of birth cohort studies: 1946 birth cohort: national study of ~5,000 individuals born in one week in 1946 1958 British birth cohort: national study of ~17,000 individuals (all births in one week) 1970 British birth cohort: national study of ~17,000 individuals (all births in one week) ALSPAC: pregnant women in one UK area with expected delivery dates April 1991- Dec 1992: ~15,000 pregnancies – offspring now studied to late 20s Millennium Cohort: national sample of ~19,000 children born Sept 2000-Jan 2002 TEDS: twins born Jan 1994-Dec 1996 to ~14,000 families  Advantages: prospective follow-up and representative sampling (common to many epidemiological designs) narrow age range → ability to study within-person Advantages & age-related change comparisons of different birth cohorts with the disadvantages same/similar assessments at comparable ages informative re changes in the prevalence of disorder of birth cohort Disadvantages: designs costly to establish and maintain allows mental health to be set in social and biomedical context but can be at the expense of detailed domain-specific measures well-suited for research into common mental health problems, less so for rarer conditions vulnerable to bias from selective participation and attrition Related alternative designs reconstructing birth cohorts from records, or ‘reviving’ earlier studies that were interrupted/not originally planned for long-term follow-up – eg: records-based studies: Lothian 1921 and 1936 birth cohorts New England Family Studies: following in adulthood samples from the Collaborative Perinatal Project that originally ran to age 7 years Related alternative designs accelerated longitudinal designs: recruit 2 or 3 cohorts of different ages, and follow each prospectively – eg Great Smoky Mountains Study Infancy childhood adolescence adulthood late adulthood Cohort 1 Cohort 2 Cohort 3 Related alternative designs natural experiments to assess impacts of specific exposures: English Romanian Adoptees Study (impact of early institutional rearing) assisted conception (IVF etc) – (impact of pregnancy exposures) policy-related variations – eg age at starting/finishing school in education systems with fixed school entry dates start cohorts at older ages (eg Biobank, English Longitudinal Study of Ageing) and use record linkage/retrospective reporting for data on early exposures specialized cohorts of individuals with specific disorders (autism, schizophrenia, etc) Analytic strategies As for most longitudinal studies – nothing specific to life course studies methods for dealing with repeated exposures and outcomes – eg random effects models, longitudinal latent class analyses Intercept Slope 1 1 1 1 1 0 1 2 3 4 t1 t2 t3 t4 t5 Longitudinal classes of depressive/anxious symptomatology: adolescence-mid-life (1946 British birth cohort) Analytic strategies careful specification of proposed pathways, possible confounding factors, intermediate variables & potential biases (Directed Acyclic Graphs can be useful here – DAGs: causal diagrams) Causal inference methods Structural equation modelling Methods for dealing with attrition and missing data to reduce bias in estimates: multiple imputation; inverse probability weighting; full information maximum likelihood Methods of assessing mental health problems/disorders Direct Questionnaires Interviews observations Health Records/ Administrative Wearables Apps Data  Questionnaires Strengths Limitations flexible, and suitable for large Can be specific to informant’s samples - can be mailed, completed perspective (may be rater biases etc) on line etc may not be detailed enough to suitable for parents, teachers and assess diagnostic categories older children informant agreement… can be used to assess symptoms of very specific disorders/traits if desired reliable economical Interviews Strengths Limitations typically designed to assess relatively expensive diagnostic categories as well (interviewers, training etc) as symptoms completion time much longer opportunity to probe than for questionnaires responses may be lower reliability than suitable for use with parents, questionnaires (esp for teachers and children younger children) – can be administered in informant agreement… respondents’ home, labs/clinical settings, on line Observations Strengths Limitations detailed records of specific may require labs/special behaviours equipment can include ‘experimental’ intensive (and often expensive) to conditions collect and code data can help identify environmental contingencies valuable for young children, and groups with limited language skills Health records and administrative data Strengths Limitations Clinical records and Typically will not capture diagnoses ensure clinical sub-clinical significance of research symptoms/risks Representativeness: it is Some people who meet possible to obtain “whole- eligibility criteria for population data” diagnoses may not seek help/diagnosis Records can be incomplete/missing Mental Health Questionnaires  Exploring the structure of emotional/behavioural and mental health problems → developing taxonomies Understanding associations between mental health problems and predictors/outcomes As screening instruments in two-stage epidemiological surveys Questionnaires: main uses Diagnosis/case identification → prevalence estimates Service planning Examining change across development and over time Characteristics of questionnaire rating scales Format: Usually 3-point or 5-point likert scale ratings of extent to which individual behaviours apply can produce both dimensional and categorical measures scores summed to give total ‘difficulties’ scores sub-sets of items (often derived from factor analyses): sub-scales, syndromes cut-points to identify disorder/’case-ness’ Reflect judgements of child’s functioning by particular informants interact with child over extended periods know child under different conditions key figures in child’s environment Questionnaires assume a shared understanding between researcher and informant of: attributes/behaviours to be rated which behaviours reflect scale items severity/scaling of items: how often is often? Possible rater effects/biases belief/norms of respondent mental state of respondent social desirability Major questionnaires used to study child mental health Child Behavior Checklist (CBCL) - Achenbach versions for parents, teachers (TRF) and children/young people assess broad range of competencies and problem behaviours relatively long: >100 items Conners’ Rating Scales versions for parents, teachers and children /young people assess a wide spectrum of behaviours, emotions, and problems – especially hyperactivity/ADHD relatively long: 40-~90 items Rutter Scales versions for parents (A) and teachers (B) initially designed as screening instruments ~30 items Strengths and Difficulties Questionnaire (SDQ) - Goodman versions for parents, teachers and young people modified version of Rutter scales, with addition of prosocial items – 25 tems Syndromes/subscales CBCL syndromes SDQ subscores Aggression Conduct Hyperactive Externalising Emotional Delinquent Hyperactivity Total difficulties Peers Anxiety Prosocial Internalising Depression Social withdrawal Somatic complaints Emotional Problems Subscale sdqinfo.org SDQ: Impact supplement Questionnaires for specific disorders/problem areas review series in Child & Adolescent Mental Health Informants All contribute differing information: complementary Parents useful in all domains may under-report hyperactivity, and emotional problems in adolescents Teachers useful for hyperactivity and attentional problems may under-report emotional problems Children/young people under-report hyperactivity, inattentiveness & oppositional behaviours provide complementary information on conduct problems important for internalising problems readability issues for younger children… Child self-reports: readability: SDQ  Agreement between informant/methods Agreement between informants and methods is typically low-moderate: Parent and teacher Parent and child Why is agreement reports: reports: limited? maximum 20% average correlations situation-specificity overlap of cases around.25 of behaviours identified as in higher in referred differing knowledge clinical range than in community of relevant average correlations samples phenomena on total scores:.3 differing thresholds agreement higher method factors and for externalising error variance than internalising problems De Los Reyes et al. (2015). The validity of the multi-informant approach to assessing child and adolescent mental health. Psychological Bulletin, 141, 858–900. Interviews Assessing Diagnostic Categories Two broad types of diagnostic interviews: Structured interview: fixed question wording ensures same probes given to all informant determines symptom presence and intensity brief training for interviewers Semi-structured interview initial questions specified, then interviewer probes detailed glossary to define symptoms interviewer must know glossary definitions interviewer varies wording as appropriate to respondent detailed questioning to provide descriptions of behaviours, frequency, intensity, duration more extended interviewer training Some of the major diagnostic interviews for assessing children and adolescents: CAPA: Semi-structured interview Comparing interviews (Angold et al, 2012) comparison of DISC, CAPA and DAWBA Time to administer DAWBA shortest (33 minutes vs 54-60) Prevalence rates (proportion with 1+ diagnoses) DISC leads to highest rates (high rates of specific phobias) excluding specific phobias… DAWBA 17%, DISC 27%, CAPA 31% levels of agreement ‘moderately encouraging’ when might each interview be most appropriate? DAWBA: services and trials DISC & CAPA: studies where important that non-cases have no problems (eg imaging, molecular genetics) Cross-cultural issues Culture shapes the meaning and social experience of mental health and illness influences exposures to factors that may be risks for mental ill health may moderate the relationship between risks and mental health outcomes cultural variations in describing, disclosing, and categorizing suffering are sources of potential bias in epidemiological research many questionnaires and interviews have been widely translated, and used in a range of countries/cultures Mean CBCL Total Problems scores in 42 societies International comparisons: comparing interviews and questionnaires Implications… Interview-based studies do point to the possibility of differences in rates of disorder between countries But… Brief questionnaire measures May also capture cross-cultural reporting effects May not be a valid method for comparing cross-national prevalence rates may point to need for population-specific norms Direct observation Naturalistic observations, structured, experimental procedures  Procedure for eliciting infant’s behaviour coded attachment-related for: behaviours in infants aged 9-18 months: 21 minute exploratory play procedure: reactions to departure of caregiver mother and infant reactions to stranger together: infant explores reunion with caregiver stranger enters, talks to parent, approaches infant; mother leaves mother returns, greets and comforts child – then leaves again infant alone stranger enters mother returns, picks up infant, stranger leaves Observational assessment of pre-school disruptive behaviour DB-DOS (Wakschlag et al, 2008) Berkeley Puppet interview Measelle, J., & Ablow, J. C. (2018) To assess young children’s (age 4-8 years): perceptions of self and others During the interview, two identical hand puppets (tan-coloured puppies named “Iggy” and “Ziggy”) make opposing statements about themselves and then ask the child to describe themselves Autism Diagnostic Observation Schedule (ADOS) Lord, Rutter et al, 1989 Structured and semi-structured tasks, providing opportunities for the child to show social and communication behaviours relevant to the diagnosis of autism 4 modules, dependent on language level of child For example, Module 3 includes: Construction Task Make-Believe Play Joint Interactive Play Demonstration Task Description of a Picture Telling a Story From a Book Conversation and Reporting Emotions Social Difficulties and Annoyance Etc.. Apps and wearables “Ecological Momentary Assessment” or “Experience sampling” Data collection methods for gathering systematic self-reports of behaviors, emotions, or experiences as they occur in the individual's natural environment. Regular intervals Specific times Can also be collected passively (e.g., heart rate) via wearables.  https://www.cataloguementalhealth.ac.uk/ Catalogue of Mental interactive catalogue of the mental health measures included in major UK longitudinal studies (currently 55 studies) Health developed by Louise Arseneault and Measures colleagues at SGDP most studies available for secondary analysis by other researchers