1 / 58

Healthy Ireland A Conceptual Approach towards a Research and Data Plan

Presentation at Department of Health, 12 th February 20 14. Healthy Ireland A Conceptual Approach towards a Research and Data Plan. Key concepts in Healthy Ireland.

Télécharger la présentation

Healthy Ireland A Conceptual Approach towards a Research and Data Plan

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Presentation at Department of Health, 12th February 2014 Healthy IrelandA Conceptual Approach towards a Research and Data Plan

  2. Key concepts in Healthy Ireland Current adverse health trends in Ireland are similar to those causing concern in other developed countries. They include projected significant increases in levels of chronic disease, exposure to health risks, growing health inequalities, and difficulty in accessing care when it is needed. Healthy Ireland, page 6

  3. Key concepts in Healthy Irelandcont. … Healthy Ireland is designed to bring about real, measurable change and is based on an understanding of the determinants of health. Health and well-being are affected by all aspects of a person’s life; economic status, education, housing, the physical environment in which people live and work. Health and well-being are also affected by policy decisions taken by Government, the individual choices people make about how they live, and the participation of people in their communities. Healthy Ireland, page 6

  4. Key concepts in Healthy Irelandcont. … This means addressing risk factors and promoting protective factors at every stage of life from pre-natal, through early childhood, adolescence, adulthood and into old age, to support lifelong health and well-being. Healthy Ireland, page 6

  5. Highlighted Risk factors affecting Positive Health Outcomes The major risk factors identified in Healthy Ireland are: Overweight and Obesity Mental Health Smoking Alcohol and Drugs But note the conceptual confusion in the above: smoking, alcohol and drugs are pure behaviours and thus potential risk factors. Obesity is a health outcome, but could still be seen as a proxy for behaviour in a similar way. By contrast, mental health is first and foremost a health outcome! Healthy Ireland, page 10

  6. In Summary Healthy Ireland provides an integrated approach to looking at higher level Health and Well-being outcomes, HI attempts to identify the key Risk and Protective Factors that affect these outcomes, HI attempts to understand the differentials in health outcomes for different groups within Irish society (i.e. the Social Gradient) and HI aims at mapping out the movement of key indicators over time, both in terms of health and well-being outcomes and the key factors which affect these.

  7. A Conceptual Approach to developing a research and Data Plan When formulating a conceptual approach to a Research and Data Plan aimed at supporting evidence-based policy making, it is our contention that this can best be achieved on the basis of understanding the structural relationships between key health outcomes and their principal determinants. This can best be achieved by analysing major Irish datasets (GUI, TILDA, SLAN) using appropriate statistical techniques. Key latent concepts can be operationalised as composite indices, to facilitate ongoing monitoring of risk and protection factors and health outcomes. Data collection should be based on a need-to-know approach and guided by the a priori identification of key concepts.

  8. Overview over rest of Presentation The remainder of this presentation is organised in four sections: The first section will demonstrate the usefulness of a Structural Equation Model in understanding Child Well-being using the Growing UP in Ireland data. This is followed by a demonstration of the determinants of Health and Well-being of older people using TILDA. The third Section discusses the underlying principles informing Composite Index Construction. The final Section outlines how the approach(es) may be applied in the context of Healthy Ireland, and particularly the development of a Research and Data Plan.

  9. Understanding Child Well-beingA study Based on the 9-Year cohort of GUI The central question guiding our research is:What influences the well-being of children and their families? Drawing on Ryff and Keyes (1995), we define well-being as a multi-dimensional construct situated between the individual and the social whole, comprising: Emotional well-being (absence of depression, internalising behaviours) Subjective well-being (e.g. life satisfaction) Relational well-being (including family and intimate relationships) Positive self-concept (self-esteem, self-efficacy) Positive work and/or study role Absence of symptoms or externalising behaviours

  10. Our Approach We situate the well-being of children within the context of the “family system”. We develop an integrated theoretical model of well-being and the family system, based on previous research. We seek robust latent multi-item measures of key concepts in this model. We distinguish between the measurement model (items and scales used to measure key concepts), the structural model (relationships between the key concepts), and the risk and protective factors that constitute the context of child development. We use Structural Equation Modelling techniques to estimate parameters in our model.

  11. An Ecological Model of child Well-being(Bronfenbrenner) Measurement Model for PCG Well-being Measurement Model for SCG Well-being Risk and Protective Factors Risk and Protective Factors SCG Well-being PCG Well-being Child Well-being Measurement Model for Child Well-being

  12. broad outline of a Second Order Latent Variable Model Dyadic Relationship Dyadic Relationship Depression Parenting Depression Parenting Risk and Protective Factors Risk and Protective Factors SCG Well-being PCG Well-being Child Well-being Scholastic Achievement Self-Concept Child Difficulties

  13. Measuring Child Well-being Child Well-being Self-Concept Scholastic Achievement Child Difficulties Emotional - PCG Conduct - PCG Hyperactivity - PCG Peer Relations - PCG Happiness - PCG Appearance - PCG Popularity - PCG Intellectual - PCG Reading - PCG Maths - PCG Teacher Evaluation Strengths & Difficulties Questionnaire Piers – Harris II Drumcondra

  14. Measuring Parental Well-being PIANTA - Child Parent Relationship Scale Dyadic Adjustment Scale Consensus Cohesion Satisfaction Positive Aspect – P1 Positive Aspect – P2 Positive Aspect – P3 Dyadic Relationship Depression Parenting PCG Well-being

  15. Risk and Protective Factors Financial Difficulties Non-Irish Ethnicity SCG Well-being Local Problem Scale Low Social Class Local Services Scale Equivalised Household Income Decile Haase-Pratschke Deprivation Score ESRI Basic Deprivation PCG Well-being Health Status (Child) Low Education (PCG) Life Events (Child) Health Status (PCG) Child Well-being Gender (Child) Age (PCG)

  16. A structural equation Model of Child and Family Well-being Child well-being SCG well-being PCG well-being d d d d d d d d d d d d DCON_P DCOH_P DCON_S DCOH_S P2_PCG P3_PCG P1_SCG P2_SCG P3_SCG DSAT_P DSAT_S P1_PCG ParentingPCG ParentingSCG DyadicSCG DyadicPCG DepressionSCG DepressionPCG d d d d d Financialdifficulties Localproblems Non-Irishethnicity Localservices Low socialclass Haase-Pratschke HHincome Health ofChild ESRIDeprivation Life Eventsof child Low educ.of PCG Genderof child Healthof PCG ChildDifficulties Self-Concept ScholasticAchievement Age ofPCG Note 1: covariances between disturbance terms for Child Well-being and Parenting (PCG and SCG) not included in figure. Note 2: all covariances between independent variables omitted from figure D_MATH D_READ EMO_P CON_P PH_POP PH_HAP HYP_P PH_APP T_EVAL REL_P PH_INT d d d d d d d

  17. Influence of Risk and Protective Factorson Family Well-being Standardised coefficients

  18. Significant Influences on Child Well-being Based on the 9-Year cohort of GUI Financial Difficulties - . 08 Non-Irish Ethnicity SCG Well-being - . 10 Local Problem Scale - . 10 Low Social Class R²=.04 - . 06 Local Services Scale Equivalised Household Income Decile Haase-Pratschke Deprivation Score - . 15 ESRI Basic Deprivation - . 11 PCG Well-being . 09 R²=.17 - . 10 Health Status (Child) - . 11 - . 06 - . 28 . 41 Low Education (PCG) . 04 - . 10 . 08 Life Events (Child) - . 07 - . 04 Health Status (PCG) - . 04 Child Well-being Gender (Child) . 07 . 12 Age (PCG) R²=.31 Goodness of Fit: N: 4,881 CFI: .95 RMSEA: .02 All effects significant at p < .05

  19. Key Findings The analysis confirms the importance of the mother’s well-being as a mediating factor on the child. A one unit improvement in the mother’s well-being is associated with a 0.4 unit direct improvement in child well-being. In stark contrast, the direct effect of the father’s well-being on the child (.04) is almost negligible once we control for other factors. A striking result is the strongly mediated effect of many contextual influences, in harmony with the ecological model of child well-being. With the exception of the mother’s health and the Haase-Pratschke Deprivation Index, which have a significant direct effect on child well-being, all other socio-economic factors, including financial variables and local area problems, have a distaleffect on child well-being that is mediated by the mother’s well-being.

  20. Discussion The conceptualisation of well-being as a higher-order latent concept reveals itself to be a powerful and well-supported hypothesis. The assumption that the well-being of children cannot be understood without simultaneously analysing the well-being of their parents is reinforced. All of the key influences identified in this analysis are in line with our previous research on child and family well-being using independent data – including the finding that a unit change in maternal well-being is associated with almost half a unit change in child well-being. Parents act as a buffer between economic risk factors and child well-being. Socio-economic risks doinfluence parental well-being, and thus have a mediated effect on children. The model presented here reflects the situation of two-parent families only. As we elected to study the dyadic relationships between caregivers and between caregivers and children, single parents were excluded. The next step would therefore be to focus on the primary caregiver and child, thus including single parent families.

  21. Significant Influences on Child Well-beingBased on a Meta-study of 6 Well-being Studies Parent Child Relationship Partner Relationship Problem Solving Depression Life Satisfaction R²=.58 R²=.22 R²=.43 R²=.67 R²=.50 . 76 . 47 . 71 - . 59 . 82 Age Educ. Local Problem Scale At Work - . 12 Medical C. Socio-economic Well-being Local Services Scale . 05 . 36 Financial D. Support Networks Parental Well-being . 15 - . 26 Negative Affect (Parent) R²=.79 . 06 Neighbourliness . 43 . 46 Positive Affect (Parent) - . 04 . 09 Age (Child) Gender (Parent) - . 08 Child Well-being Gender (Child) . 11 Age (Parent) R²=.36 Goodness of Fit: N: 1,600 CFI: .95 RMSEA: .02 Conduct P. Emotional P Hyperactivity. Peer Prob. All effects significant at p < .05

  22. Reflecting on the design of the GUI Dataset The GUI 9 year-old cohort data has a number of strengths… a large sample, panel design, multiple outcome measures, independent assessments, and a clustered sampling design consistent with an “ecological” approach. …but, given the overwhelming influence of parental well-being on child well-being, the data has also some weaknesses: It does not provide sufficient information on relationships (reciprocity, support, intimacy, conflict) within the neighbourhood, family or friendship group. It lacks a range of important measures, such as conflict between intimate partners, subjective well-being, physical symptoms, positive/negative affect, adult self-concept. In developing a Research and Data Plan for Healthy Ireland it is important that the key structural components that determine overall health and well-being are well understood before signing off on what data should be collected for the future.

  23. Learning from TILDA There are four questions which guide our analysis of the TILDA data: What is the relationship between Overall Health and Overall Well-being? What is the influence of mediating factors such as an Active Lifestyle and the level of Health Care utilised? How are these concepts influenced by a wide range of contextual factors? How do we develop an understanding of Ageing, other than being simply synonymous with age? Note the similarity of questions raised in developing a Research and Data Plan for Healthy Ireland!

  24. Key Concepts to be considered Well-being Again, we define well-being as a multi-dimensional construct situated between the individual and the social whole, comprising: Emotional well-being (absence of depression) Loneliness Subjective well-being (e.g. life satisfaction) Positive self-concept (self-esteem, self-efficacy)

  25. Key Concepts to be considered Overall Health Rather than looking at the effect of contextual factors on each specific dimension of Health, we construct a measure of Overall Health Status as a higher-level latent concept comprising five components: Blood Measures Movement Measures Neuropsychological Measures Eyesight, and Sensory Functioning

  26. The Dataset Our analysis is based on the first wave of TILDA data, which has many strengths… Large sample, panel design, multiple measures, independent assessments, clustered sampling design, “ecological” approach. Unlike the GUI dataset, TILDA also provides information on self-concept , subjective well-being, physical symptoms, reciprocity and support within the neighbourhood, family or friendship group. …but the TILDA data still has some weaknesses: It notably lacks a measure of personality traits (positive/negative affect), which has been shown to be of importance and conceptually and empirically different from depression. Nevertheless, almost all of the constituents to well-being thought to be of importance can be implemented within the TILDA analysis.

  27. An Ecological Model of the Well-being of Older People Measurement Model for Active Lifestyle Measurement Model for Social Class Measurement Model for Overall Health Status Social Class Risk and Protective Factors Risk and Protective Factors Active Lifestyle Overall Health Personal Well-being Measurement Model for Personal Well-being

  28. An Important Note on Missing Data The dataset used to estimate the TILDA Well-being Model comprises of all subjects where there exists a complete set of data for all constructs included. It is divided into two samples, a male sample with 1,828 cases, and a female sample with 1,833 cases. Missing data resulted in the following loss of records: Failure to return the essential self-completion questionnaire leads to a loss of roughly 15% of the sample, or about 1,300 cases. Failure to participate in the health assessment leads to a further reduction of almost 2,000 cases. Failure to provide data on incomes and assets imposes the loss of a further 1,300 cases approximately. The need to exclude young partners from the sample leads to a further loss of roughly 20 men and 150 women. Although the complete TILDA sample contains 8,504 cases, the cumulative impact of missing data implies a severe amount of data lost.

  29. An SEM Model of the Well-being of Older Persons Social Class Age Active Lifestyle Lives Alone Drinks Regularly Relationship Quality Neuropsychol. Health Alcohol Problem Close Friends/Relatives Smokes Unemployed Overall Health Medical Screening Transport Problems Alcoholic Parent ADL/IADL Impairments Abused Childhood Care Received Helping Neighbours Personal Well-being Note: Paths between independent variables omitted from figure Religiosity

  30. Model Complexity The final Model is estimated separately for the male and female samples and comprises: 36 variables per sample A total of 1,332 observed variances and covariances Requiring the estimation of 363 parameters (107 regression coefficients for each sample, 16 constrained to be equal across samples 41 variances for each sample, 21 constrained to be equal across samples 52 covariances for each sample Overall more than 10 cases per parameter estimated. Despite the significant amount of data lost, this represents a very powerful and well-supported model.

  31. Measuring Social Class Third-level Education Assets Income Occupation .60 .29 .40 .51 Social Class

  32. Measuring Overall Health Status MMSE Montreal Test Memory Test Exec. Function .63 .74 .73 .53 Neuropsycho-logical Health Sensory Functioning Cholesterol Movement L/R Eyesight .20 -.62 .57 -.35 .33 .41 Overall Health .32 Social Class

  33. Measuring Active Lifestyle Active Social Participation Involved in Club/Group Works, Studies or volunteers Physical Exercise .60 .43 .27 .25 -.14 -.46 Active Lifestyle Age

  34. Measuring Personal Well-being Personal Well-being -.60 -.62 .51 .85 Life Satisfaction Depression Loneliness Self-Concept

  35. Significant Influences on Personal Well-being (Males) Goodness of Fit: N: 1,851 CFI: .95 RMSEA: .028 Social Class Age Lives Alone -.26 .28 Drinks Regularly .48 Relationship Quality Alcohol Problem -.75 Active Lifestyle Close Friends/Relatives Smokes .23 Medical Screening Helping Neighbours .42 .77 .32 Unemployed ADL/IADL Impairments -.39 Overall Health Transport Problems Abused Childhood .24 Care Received Personal Well-being Religiosity All effects significant at p < .05

  36. Key Findings The analysis confirms the pivotal importance of age for the health and well-being of older persons. A one unit (STD) increase in a person’s age is associated with a 0.75 unit deterioration in the person’s overall health. This, however, does not automatically translate into a similar deterioration of personal well-being. Whilst a person’s health is a strong influence on his/her well-being, the relationship (.24) is not as strong as one might expect and indeed, the moderate negative influence of age via the deterioration in health is partially offset by the positive direct effect of age on well-being (.23) . The other major factor which decouples a person’s well-being from his/her overall health alone, is the strong positive effect (.42) associated with a high quality relationship. It should, however, be noted that one could equally postulate a model in which the quality of relationship is taken as a measurement of a person’s well-being.

  37. Key Findings As is well-supported by other research, social class has a strong effect (.32) on the health of an older person, but surprisingly, no statistically significant direct effect on personal well-being. Instead, the effect of social class on well-being is entirely mediated; firstly via the already established path from health to well-being and secondly, by the complex way in which it affects a person’s active lifestyle. There exists a strong (.28) direct effect of social class on living an active life style. Poorer health has a strong negative effect (-.72) and an active life style has a positive effect (.14) on a person’s well-being.

  38. Other important Findings Medical Screening is positively associated with High Social Class (.18), but negatively associated with Age (-.26) and being Unemployed (-.09). The amount of Care Received from relatives reflects the degree of Impairments (.17), Close Relatives/Friends (.07) and Car Access Problems (.08). Lesser Care Received is associated with better Overall Health (-.08). Living Alone has a negative effect on Health and Personal Well-being, as well as an indirect negative effect on well-being via Alcohol Problems. A good Relationship Quality and having Close Friends and Relatives have positive effects on Active Lifestyle, Overall Health and Personal Well-being. Experiencing Transport Problems negatively affects both Overall Health and Personal Well-being. Religiosity does improve a person’s Well-being, both directly and indirectly via improved drinking and smoking behaviour. In contrast, being unemployed has a negative effect on well-being.

  39. Implications For the development of the Research and Data plan Our analysis of TILDA provides strong support for a conceptual approach which builds on the use of latent variables in the context of a structural equation modelling environment. The approach is able to distinguish between the health and well-being of a person, as well as quantifying the influence that key concepts such as social class and lifestyle have on these outcomes. Of the four underlying tenets of Healthy Ireland, Mental Health, Smoking and Alcohol and Drugs are already successfully identified with their respective effects on Health and well-being. It may be possible to specify the model in such way as to align it further with the assumptions underlying Healthy Ireland.

  40. From Latent variable to Composite Indicator Construction The conceptualisation of key aspects of health and well-being and their determinants as latent concepts is in line with the rising interest internationally in the construction of Composite Indicators. It facilitates an understanding of the key dimensions of these concepts a definition as to how these can be measured the ability to model their structural relationships the identification and measurement of key influences on the latent concepts The monitoring of changes in both outcomes and key determinants over time.

  41. Composite Indicators are tools for policy-making Rapid development of economic indicators after WWII, but slow catch-up in “social” arena. Indicators are a proxy for unobservable phenomena (broad picture) Practical, intervention-oriented tools for decision-making Without theories, statistical data merely “describe the symptoms” The following slides are excerpts from a recent 2-day workshop delivered at the European Monitoring Centre for Drugs and Drug Addiction (EMCDDA). Composite index construction in context

  42. Principles in the construction of composite indicators What is badly defined is likely to be badly measured … A sound theoretical framework is the starting point in constructing composite indicators. The framework should clearly define the phenomenon to be measured and its sub-components and select individual indicators and weights that reflect their relative importance and the dimensions of the overall composite. Ideally, this process would be based on what is desirable to measureand not which indicators are available.

  43. Principles in the constructionof composite indicators Overview of Composite Index Construction:

  44. Principles in the constructionof composite indicators • The Unit of Analysis: • EU • Country • Region, or • Sector • Delivery Units (Hospitals, Specialities, Praxes etc.) • Households • Individuals • Episodes (time intervals)

  45. Principles in the constructionof composite indicators • Selection of Variables: • Think about what is desirable to measureand not which indicators are available • Identify the dimensions of the composite indicator • Use domains for organising the data, but be guided by an understanding of the underlying dimensionality • Select indicators on a “need to know” basis, not “everything you can get your hands on” • Evaluate quality of data according to set criteria • Evaluate what’s missing • State the limitations as the composite indicator is only as good as its components

  46. Principles in the constructionof composite indicators Evaluation Criteria for Selection of Variables: • Relevance: are the data what the user expects? • Accuracy: are the figures reliable? • Comparability: are the data in all necessary respects comparable across countries (units of analysis)? • Completeness: are domains for which statistics are available reflecting the needs expressed by users? • Coherence: are the data coherent with other data? • Timeliness and punctuality: does the user receive the data in time and according to pre-established dates? • Accessibility and clarity: is the figure accessible and understandable?

  47. Principles in the constructionof composite indicators • Preliminary Data-Analysis in Variable Selection: • Uni-dimensional Construct; i.e. construct covering a single dimension • Undertake Reliability Analysis Cronbach’s Alpha, which measures internal consistency of a set of items

  48. Principles in the constructionof composite indicators • Preliminary Data-Analysis in Variable Selection: • Multi-dimensional Construct; i.e. construct covering multiple dimensions • Undertake Factor Analysis to identify sufficiently supported dimensions (subject to sufficient n) • Ask whether the dimensions have substantive meaning • Avoid constructs which are “indicator rich but information poor” • Undertake Reliability Analysis (Cronbach’s Alpha) to assess internal consistency of each dimension.

  49. from objectives to design Dimensionality Composite indicators seek to measure multi-dimensional concepts which cannot be captured by single indicators Other terms: domain, facet, sub-component, factor, pillar Relates to the pattern of correlations between variables Multi-level, hierarchical phenomenon Has consequences for weighting and aggregation Analysis may be theory-driven or data-driven No empirical answers to questions about dimensionality

  50. from objectives to design Timeliness and Temporal Dynamics Timeliness refers to the period that elapses between the availability of an indicator and the phenomenon described Temporal dynamics are important for benchmarking, programme assessment and causal analysis Uses of composite indicators should be anticipated to ensure that design allows for required temporal dynamics: “For time-dependent studies, in order to assess country performance across years, the average across countries … and the standard deviation across countries … are calculated for a reference year, usually the initial time point...” (Nardo et al., 2005, pp. 60-1).

More Related