挖掘大數據以指導政策并提高健康公平性.pdf

《挖掘大數據以指導政策并提高健康公平性.pdf》由會員分享，可在線閱讀，更多相關《挖掘大數據以指導政策并提高健康公平性.pdf（33頁珍藏版）》請在三個皮匠報告上搜索。

1、Mining large data to guide policy and improve health equityMARCELA HORVITZ-LENNON MD MPHSENIOR PHYSICIAN SCIENTIST,RAND CORPORATIONASSOCIATE PROFESSOR OF PSYCHIATRY(PART-TIME),HARVARD MEDICAL SCHOOLSENIOR SCIENTIST,HEALTH EQUITY RESEARCH LAB,CAMBRIDGE HEALTH ALLIANCE5th National Big Data Health Scie

2、nce Conference,Columbia SC,February 3,2024OutlineDefining and understanding health equity Words and concepts The role of public policy Inequities in mental health and mental health careBig data an opportunity to identify and address inequities that carries risks Bias in AI and why it matters for equ

3、ity What it takes to do impactful health equity research2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY2Defining and Understanding Health Equity2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY3Words and concepts2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY4Health Equity i

4、nfluential definitions2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY5Braveman.Health Disparities and Health Equity:Concepts and Measurement.Annu Rev Public Health 2006.Absence of unfair,avoidable or remediable health differences among groups of people based on social stratifiersAchieved wh

5、en everyone can attain their full potential for health and well-beingSocial stratifiers:Socioeconomic position(social class)Race and Ethnicity Place of residence Sex,Gender,or Sexual orientation Age DisabilityWorld Health Organization Health equity(who.int)“People often use health inequalities in wh

6、at may be an effort to avoid the judgmental or moral connotations that may be associated with health inequities”Braveman&Gruskin.Defining health equity.J Epidemiol Community Health 2003Health inequities faced by people disadvantaged based on social stratifiers are compounded by stereotyping,prejudic

7、e,and discriminationStereotyping=oversimplified generalizationsPrejudice=biased thinking Discriminatory practices often embedded in institutional and political processesDiscrimination=bias-driven actions against a group of people2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY6The role of pu

8、blic policy2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY7Understanding Inequities WHO CSDHs Conceptual Framework*2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY8CSDH(2008).Closing the gap in a generation:health equity through action on the social determinants of health.Final Repo

9、rt of the Commission on Social Determinants of Health.Geneva,World Health OrganizationHealth(and disease)regarded as a socially produced phenomenon Locates health as a social justice topic(human rights framework)Health inequities flow from patterns of social stratification Social stratification-the

10、systematically unequal distribution of resources,power,and prestige among social groups Addressing determinants-political process that engages the agency of disadvantaged communities and the responsibility of the state Emphasis on creative self-empowerment of previously oppressed groups*Draws from t

11、he work of Finn Diderichsen and his Model of Mechanisms of Health Inequality Diderichsen et al.The Social Basis of Disparities in Health.In:Evans et al.,eds.Challenging inequities in health.New York,Oxford University Press 2001CSDHs Conceptual Framework:Structural and Intermediate Determinants2024MI

12、NING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY9Social Determinants of Health Inequities-stratifying contextual factors that create social hierarchies*Groups stratified based on resources,power,and prestige*Proxy indicators:income,education,occupation,gender,race/ethnicity,disability,othersSocial

13、 Determinants of Health Inequities influence health through differential exposure to Social Determinants of Health*Material circumstances(e.g.,housing,physical environment)*Psychosocial circumstances(e.g.,stress,social support)*Behavioral and biological/genetic factors*Health system*Social determina

14、nts of health:conditions in which people are born,grow,live,work,and ageWords and concepts matter if policies seek to address the root causes of health inequities2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY10Castrucci&Auerbach.Meeting individual social needs falls short of addressing soc

15、ial determinants of health.Health Affairs Forefront.2019Inequities in mental health and mental health care2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY11Schizophrenia and social disadvantage 2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY12Onset early in life-first episode of psy

16、chosis(FEP)typically in late teens,early20sLow prevalence(1%)but high disease burden Disability that can be ameliorated Premature mortality that can be preventedRisk Factors Genetic etiological factors(familial clustering)Sociodemographic:age,sex,marital status,employment and educational status,race

17、/ethnicity Non-genetic etiological factors:Stressful life events,urbanicity,cannabis and other substance abuseWorse prognosis in the US relative to other industrialized countries Relapses leading to hospitalization or incarceration are common Overrepresented among homeless persons More than 3 in 4 l

18、ack gainful employment&depend on income supportsIllness is costly to US society In 2013,total costs=$155.7 billion(healthcare costs:24%;unemployment:38%)Using policy to improve outcomes2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY13Use health policy levers to expand access to care and pro

19、mote high-quality and equitable care Enforce and strengthen parity lawsReduce barriers for gaining Medicaid coverageIncrease Medicaid provider rates&budgetsImprove oversight-penalize abusive practices&network inadequaciesInvest in educating and training the workforceIncentivize high-quality care Als

20、o,promote reductions in Duration of Untreated Psychosis(DUP)Association between DUP and outcomesIn the US:longer DUP for racial/ethnic minorities Using policy to improve outcomes2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY14Incentivize equitable care Improve race/ethnicity data collectio

21、n Require reporting of data stratified by race/ethnicity and appropriately risk-adjusted Incentivize culturally competent care,greater workforce diversity,and disparities reductions Ensure that providers caring for disadvantaged individuals are not penalized and paid adequately Require that private

22、payers reinvest into the communities they serveAddress modifiable risk factors through social policyDisentangling inequities by place of residence and race/ethnicity among Medicaid beneficiaries with schizophreniaMINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY202315Trends in quality of mental h

23、ealth care and quality inequities among Medicaid beneficiaries with schizophreniaMINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY202416Differences by race/ethnicity in access to critical physical health care during the COVID-19 surge in NYS among Medicaid beneficiaries with schizophreniaMINING L

24、ARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY202417Big data offers an opportunity to identify and address inequities but it carries risks2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY18Big data can shine a light on health inequities and its drivers2024MINING LARGE DATA TO GUIDE POLICY AND I

25、MPROVE EQUITY19Large observational databases assembled from multiple sources containing varying amounts,types,and quality of health,healthcare,social needs,and social care informationArea-level indices assembled with data from multiple sources(American Community Survey,US Census,etc.)describing SDH

26、at various geographic levels)Trinidad et al.Use Of Area-Based Socioeconomic Deprivation Indices:A Scoping Review And Qualitative Analysis.Health Affairs 2022.Bias in AI and why it matters for Equity2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY202024MINING LARGE DATA TO GUIDE POLICY AND IM

27、PROVE EQUITY21Statistical/computational bias is only the tip of the iceberg2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY22Bias in AI Statistical and Computational 2024o Stem from systematic error arising from use of unrepresentative samples Can occur in the absence of prejudice or discrim

28、inatory intent o Present in the datasets and algorithmic processes used in the development of AI applicationso Multiple potential drivers:Heterogeneous or wrong data Representation of complex data in simpler mathematical representations Algorithmic biases Treatment of outliers OthersMINING LARGE DAT

29、A TO GUIDE POLICY AND IMPROVE EQUITY23Bias in AI HUMANReflect systematic errors in human thought based on a limited number of heuristic principles Heuristics-adaptive mental shortcuts that permit complexity reduction in tasks of judgement and choice but can lead to cognitive biasesHeuristics and hum

30、an biases are implicit-increasing awareness does not ensure mitigationOmnipresent in the institutional,group,and individual decision-making processes across the AI lifecycle and in the use of AI applications 2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY24Bias in AI SYSTEMIC2024MINING LARG

31、E DATA TO GUIDE POLICY AND IMPROVE EQUITY25Result from procedures and practices of specific institutions leading to certain social groups being advantaged or favored and others being disadvantaged or devaluedPresent in the datasets used in AI;institutional norms,practices,and processes across the AI

32、 lifecycle;and broader culture and societyA way forward Bias may not be fully eliminated but value of AI may be safely realized only if all AI stakeholders,including policymakers and other end users,are aware of its pervasiveness,sources,and possible mitigation strategies2024MINING LARGE DATA TO GUI

33、DE POLICY AND IMPROVE EQUITY26What it takes to do impactful health equity research2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY27Key ingredients Develop multidisciplinary hubs of researchers that leverage existing knowledge and other resources and use multiple research methodologies to:-i

34、dentify community priorities for addressing inequalities-deepen the understanding of mechanisms of inequities-investigate potential additional impacts-design and evaluate strategies to eradicate them-motivate and train the next generation of researchers2024MINING LARGE DATA TO GUIDE POLICY AND IMPRO

35、VE EQUITY28What it takes to do impactful health equity researchMultidisciplinary hubs of researchersDisciplines-stress biology and other basic sciences;clinical medicine,psychiatry and other medical specialties;population health;social epidemiology;health services research;health economics;public po

36、licy;anthropology2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY29What it takes to do impactful health equity researchMultidisciplinary hubs of researchers o Disciplines Methodologies-disparities research,qualitative methods,community-based participatory research,statistical and artificial

37、intelligence methods,simulations,complex systems,translational research,dissemination&implementation research,econometrics,policy analysis&evaluation2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY30What it takes to do impactful health equity researchMultidisciplinary hubs of researchers o D

38、isciplines o MethodologiesStudy designs-randomized trials,quasi-experimental studies,social epidemiological(cohort)studies,qualitative and mixed-method studies 2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY31What it takes to do impactful health equity researchMultidisciplinary hubs of rese

39、archers o Disciplines o Methodologieso Study DesignsFunding Heightened interest in health equity research among funderso NIH study sections and institutes including the National Institute on Minority Health and Health Disparitieso Centers for Medicare&Medicaid Services(Office of Minority Health)o Pa

40、tient-Centered Outcomes Research Institute grantso Foundations(e.g.,RWJF-Evidence for Action-E4A:Innovative Research to Advance Racial Equity)o DHHS Office of the Assistant Secretary for Planning and Evaluation(ASPE)2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY32THANK YOU2024MINING LARGE DATA TO GUIDE POLICY AND IMPROVE EQUITY33