數據分析與決策科學 新統計達人之訣竅 藍忠孚 國立陽明大學 教授兼醫學院副院長 98年醫療機構調查研討會 2009 January 16,
序幕 2
奈米與千兆位元組 1 • International System of Units (abbreviated “SI” from the French Le Systeme International d’Unites) • The SI was developed in 1960 from the old meter-kilogram-second (mks) system rather than the centimeter-gram-second (cgs) system. • Because the SI is not static, units are created & definitions are modified through international agreement among many nations as the technology of measurement progresses, & as the precision of measurements improves. 3
奈米與千兆位元組 2 • Data for digital information storage is usually measured in bits (20); 1 bit – 0 or 1, true or false, low or high. • A group of 8 bits in a computer is called a byte (23). A byte is the common unit of measurement for computer architectures. • In architectures that are 16 -bit or 32 -bit, which operate on 2 or 4 bytes per clock cycle. • The decimal (十進位 ) prefixes kilo, mega etc, are strictly powers of 10. The powers of 2 are the binary (二進位 ) prefixes kibi, mebi etc. 4
Standard Prefixes for the SI Units of Measure deca- hecto- kilo- mega- giga- tera- peta- exa- zetta- yotta- da h k M G T P E Z Y 102 103 106 109 1012 1015 1018 1021 1024 Name Multiples Symbol Factor 100 101 Name Subdivisions deci- centi- milli- micro- nano- pico- femto- atto- zepto- yocto- Symbol d c m μ n p f a z y 10 -1 10 -2 10 -3 10 -6 10 -9 10 -12 10 -15 10 -18 10 -21 Factor 100 10 -24
資料集 (Result Set) • 百萬位元組 (megabytes), 十億位元組 (gigabytes), 兆位元組 (terabytes), 千兆位元組 (petabytes). • U. S. Library of congress的藏書 , 只有二十兆位元 組. • Wal Mart 資料庫有五百七十兆位元組. • Google有約四千兆位元組資料需日夜分析. • 兆位元資料採礦 (tera mining) • 超級數據分析 (super crunching) • 超級數據達人 (super crunchers) 6
蘋果橘子經濟學 • 廠商如何知道我們的品味和偏好 ? • 政府如何瞭解民眾需要和期望 ? • 面對金融海嘯 , 人們生活行為會如何改變 ? • 喜歡看棒球和喜歡看籃球的人 , 飲食習慣 有何不同 ? • 重視節能減碳的人與用信用卡的人 , 有什 麼關係 ? 7
Conventional Approach • 依賴單一小型數據分析、統計方法、邏 輯推理、科學知識 ; • 已可局部掌握變遷趨勢，發掘問題癥結， 預測未來方向，協助我們作選擇. • 預測目前還是不精確的科學. • 直覺和經驗和有限 數據分析為主的 決策. 8
Super-crunching Approach • 電腦資料庫的容量大增 , 且速度飛快. • 超大資料網路和採礦技術在發威. • 改變企業和政府更準確預測民眾需要 , 精 密的社會 程 (social enginery), 正在席捲 世界. • 超級數據分析可 發掘出隱藏的關聯 , 開展 具更佳視野的資料導向決策. • 藉 超級 數據和邏輯分析來思考 , 可增加更 多、更好、更具創意之抉擇與決策. 9
提升分析力與決策力 : 掌握趨勢、抓住重點、強化競爭 10
新世紀的知識經濟 五層次 (Knowledge-based Economy) • 數據 (Data) • 資訊 (Information) • 知識 (Knowledge) • 內化 (Internalization) • 洞察力 (Insights) 11
當高中的老師 被誤為高中的同學時 … 12
認知、思考、實踐和預測之層 次 • 發生何事？ (Know-what) – description – to describe, to understand, … • 為何如此？ (Know-why) – inference association, causation/correlation – to explore, to examine, to diagnose, … • 如何解決？ (Know-how) – prescription/intervention effectiveness, efficiency, equity, impact, … – to test, to intervene, to correct, to cure, … • 未來如何？ (Know-what-next) – forecast – to project, to estimate, to predict, …
現代 科學之兩大支柱 • 實驗 Experiments (觀察 observations) (manipulation, control, randomization, & replication) • 邏輯與推理 Logic & Reasoning (數學 mathematics) 14
關聯性 (Association) • 因果關係 (causation; causal relationship) • 相關關係 (correlation; correlational relationship) • Multivariate analysis • True experimentation 15
因果關係之實例 : 健康 • Therapy: treatment improvement in patient’s condition • Etiology: causal factors disease incidence • Health services: services improvement in health of community 16
Scientific Methods & Its Applications in Policy Making • Knowing / Learning / Analyzing & Evaluating • Thinking (macro/micro, alternatives/options) • Speculation (trade-offs, pros & cons, the best/worst scenarios, game theory) • Innovation (creative/strategic thinking) • • Logic Reasoning Argumentation Conclusion
Evidence-based Policymaking 18
Policy Challenges Encountered by the Decision Makers • • • Dealing with scarcity Funding systems equitably & sustainably Allocating resources effectively Delivering care efficiently Implementing change
Demographics & Burden of Disease/Disability • • • Living longer & aging society Increasing diversity Lengthening years of education Widening household income gap Shifting burden of disease/disability – lifestyle behaviors, chronic diseases, mental illness, & emerging & reemerging infectious diseases
Health Finance & Health Insurance • Escalated healthcare costs • Market dynamics • The poor, the near poor, the vulnerabled/disadvantaged & the unprotected • Increasing tiering of health insurance • Continuous debates on the role & responsibility among the family, the employer, & the society
Health Paradigm Shifts • Adherence to the biomedical model -> expansion to a multifactorial view of health; • Attention to acute episodic illness -> extension to chronic illness management; • Focus on disease -> broader focus to diseased person & the disease; • Focus on individuals -> more focus on communities & defined specific populations; • Cure as uncompromised goal -> adjustment to better care & better quality of life.
The Direction in Resolving Issues in Health System • • Improving access to care for people; Containing health care costs; Maintaining quality of services; Empowering patient’s rights. Convergence of “health development” with “socioeconomic development”
Steps of Evidence-based Policymaking • • Asking an answerable question; Tracking down the best evidence; Critical appraisal; Integrating the appraisal w/ expertise & people preference; & • Auditing performance in step 1 -4.
Factors to Consider When Selecting a Health Topic for Audit • Important impact on health – Likely increase in morbidity/mortality if care is poor. • Affects a large number of people – Improving quality of care in common conditions usually has more impact than in rare conditions. • Convincing evidence is available about appropriate care – Otherwise, efforts to change current performance are difficult to justify. • Good reasons for believing that current performance could be improved – Concentrates effort on optimum elements of care
統計指標 26
Questions • How healthy are we? • How long can we live? • How good are our life should be? 27
The Major Categories of Health Statistics • Health status: – prevalence & burden of disease (BOD) & epidemiological concerns • Health services: – Service delivery, treatments, quality of care • Health care financing: – Costs, access, insurance • Outcomes of care & efficacy: – Health utility, quality of life
The Significance of Health Statistics 1 • To document (also to monitor trends in) health status/needs of the citizen & of the special population groups; • To analyze strength & weakness of health care financing & delivery; • To identify health problems or health policy issues (e. g. , health disparities); • To provide evidence-based information for health system reform;
The Significance of Health Statistics 2 • To evaluate the effects/impacts of health policies, programs, & activities; • To support biomedical & health services research; • To optimize the health resources allocation & health priority setting; • To do comparative systems analysis & to learn from each others.
Measuring & Valuing Life, Health Care, & Health System 1 • • Life course approach Spectrum of care Individual health vs. Population health Physical, mental, & social health Quantity of life vs. Quality of life Biomedical & non-biomedical determinants Social gradients & health Aging & functioning, disability, &/or frailty 31
Measuring & Valuing Life, Health Care, & Health System 2 • • Negative & positive health indicators Single & multiple health risks & problems Single, composite, & summary measures Causation/correlation of health & other variables • Including health w/ wealth, w/ happiness, w/ well-being, … • Evidence-based health policy & practices 32
社會、經濟、政治、國際環境 （Macro Environment） 健康體制與政策 (Heatlh System & Policy) 國民健康狀況 (Health Status) 健康照護資源 (Health Care Resources) ˇ人口學指標 ˇ國民健康指標(負向為主) ˇ非醫療健康因子指標 ˇ經濟指標 ˇ醫療照護財源指標 ˇ醫療資源 ˇ長期照護資源 ˇ藥物市場指標 ˇ社會保障指標 健康照護品質 (Healthcare Quality) ˇ醫療品質指標 ˇ照護品質指標 健康照護支出 (Health Expenditure) 健康照護利用 (Health Care Utilization) ˇ醫療利用指標 ˇ長期照護利用指標 ˇ藥物利用指標 ˇ醫療費用指標 ˇ長期照護費用指標 ˇ藥物費用指標 健康效果 (Health Outcomes) ˇ國民健康指標(正向為主) ˇ健康生活品質指標 政策評估 (Health Policy Evaluation) 資源配置 (Health Resource Allocation)
Aging & Disability • Aging – Elder – young old, older old, oldest old • Disability – Disabled • Health & Functioning – Risk & uncertainty – Disease – Chronic health condition – Co-morbidity, multi-morbidity, complications – Disability – Impairment – Dysfunction – Handicap
Health & Disability Status of the Elderly & Disabled Population (1) • Positive indicators: – Quantity of life -- Life expectancy – Quality of life • Negative indicators: – Morbidity; Chronic health conditions; Co-morbidity – Sensory impairments & oral health – Memory impairment – Depressive symptoms – Disability – Mortality; Multiple causes of death • Respondent-assessed; Professional judged
Health & Disability Status of the Elderly & Disabled Population (2) • Healthy • With disability – mild, moderate, severe • With disease or chronic condition – mild, moderate, severe • With disease & disability – mild, moderate, severe • Home-bound, home-ridden • Chair-bound, chair-ridden • Bed-bound, bed-ridden
ICF vs. ICD • International Classification of Functioning, Disability & Health (ICF) • Comprehensive classification of eight health & healthrelated domains: mobility, self care, pain, cognition, interpersonal activities, vision, sleep & energy, & affect (mood) • International standards for use in surveys & clinical encounters
Frailty: An Overlapping Concept • Aging – Increased vulnerability to disease & accidents over time • Disability – Functional limitations resulting from impairments • Comorbidity – Disease processes resulting from biology & exposures • Frailty – Multidimensional, unstable, heterogeneous
Frailty: Consensus Definition (Canadian initiative on frailty & Aging, 2003) • A syndrome encountered in older persons that has diverse predisposing, precipitating, enabling & reinforcing factors. The key feature is a state of vulnerability to adverse health outcomes. There is a characteristic clustering of features that can lead to its recognition. The balance between assets & deficits will determine the consequences for an individual. Adaptability, physical environment & social environment are important determinants of the impact of frailty. (ecological model in epidemiology)
Frailty Phenotype • • • Weight loss Slow walking speed Low levels of physical activity Subjective exhaustion Weakness (low grip strength) • 3 -5 is “frail” • 1 -2 is “intermediate” • 0 is “no frail” Fried LP et al. J Ger Med Sci. 2001
Dynamic Frailty Measures • Functional Independence – Barthel Index – Functional Homeostasis • Prior geriatric Assessment – – Rules based definition GCIF (Geriatrician’s Clinical Impression of Frailty) CSHA Frailty Index FI-CGA • General Screening – Edmonton frail Scale
Markers of Frailty • Inability to perform one or more of basic ADL in the three days prior to admission • A stroke in the past three months • Depression • Dementia • A history of falls • One or unplanned admissions in the past three months • Difficulty in walking • Malnutrition • Prolonged bed rest • Incontinence
High Risk Groups • • • Very old (>75 years) Recently widowed Living alone / socially isolated Poor social/economic conditions Recently discharged from hospital Recently bereaved
Postponing Disability: Disability vs. Longevity • On population aging, researchers sought evidences to identify points of decline & potential intervention on disability. – Age structure & disability – Age at the onset of disability – Disability transition – Prospects for primary, secondary, & tertiary prevention of mortality, morbidity, disability, & frailty
Social & Economic Determinants of Health • • • Poverty Stress Early life Social exclusion Work Unemployment Social support Addiction Food Transport 45
Summary Measures of Population Health • • Healthy Life Year (Hea. LY) Disability-Adjusted Life Year (DALY) Quality-Adjusted Life Year (QALY) Health-Adjusted Life Expectancy (HALE) • Disease-free, Disability-free • life-year, person-day 46
Health-Adjusted Life Expectancy (HALE) • An extension of the notion of life expectancy to morbidity & disability, made possible the calculation of disease-free and disability-free life expectancy. • Providing the ability to monitor the consequences of the decrease in mortality among the oldest-old. • It implies successful aging or healthy and active aging.
非醫療之健康決定因子 (Non-medical Determinants of Health) • 飲食攝取、食品安全 • 香菸、酒精使用 • 體重、體脂肪、身體質量指數 (body mass index)、 腰臀比 • 職業相關勞動、運動習慣 • 環境：空氣品質
經濟指標 (Economic References) • 總體經濟指標 • 貨幣轉換指標 (PPP、匯率)
Determinants of Health • Gene • Environment • Life styles & behaviors – – Diets & nutrition Exercises Stress (Happiness) Social engagement (Vitality, Active) • Immunization • Health care 50
結語 51
Evidence-based Policymaking • 統計學 • 科學計量分析 (econometrics, psychometrics, sociometrics, health metrics, etc) • 數位資訊 • 資料採礦 • 超級數據分析 • 科技整合 • 證據為基礎之決策與實務 (evidence-based policymaking & practice) • 證據健康政策與計畫 (evidence-based health policy & program) 52
提升分析力與決策力之要領 • • 思維 (Thinking) 邏輯 (Logic)與推理 (Reasoning) 歸納 (Induction)和演繹 (Deduction) 質疑 (Skepticism) 辯證 (Argumentation) 證實 (Verification)和證偽 (Falsification) 創新 (Innovation) 決斷 (Determination) 53
References • • • 李維特和杜伯納 : 蘋果橘子經濟學 Levitt, SD & Dubner, SJ, Freakonomics: A Rogue economist explores the hidden side of everything, 2005. 派卡德 : 隱形的說客 Packard, V, the hidden persuaders, 艾瑞斯 : 什麼都能算，什麼都不奇怪 -- 超級數據分析的秘密 Ayres, I, Supercrunchers: Why thinking-by-numbers is the new way to be smart, 2007. 艾斯地和洛新 : 政策的測試 Esty, DC & Rushing, R, Data-driven policymaking, Center for American Progress, 2005. 潘和希雷 : 偏好搜尋引擎的問題 Pharm, A & Healey, J, Telling you what you like: “preference engines” track consumers’ choices online & suggest other things to try. But do they broaden tastes or narrow them? , L. A. Times, 2005. 蘇洛維基 : 群眾預測的準確性 Surowiecki, The wisdom of crowds, 2004. 54
Happy Chinese New Year to You & Your Family !! 55
Скачать презентацию 數據分析與決策科學 新統計達人之訣竅 藍忠孚 國立陽明大學 教授兼醫學院副院長 98年醫療機構調查研討會 2009 January
46f13a01d22e4e82cd6e7d7941d0d925.ppt