, Causation of Disease Occurrence, Multilevel Approach
?
?.? : : :,,,.
(conditionalism) One-to-one relationship : The cause of any effect must consist of a constellation of components that act in concert (John Stuart Mill, 1862).
Rothman s Pie Rothman, Greenland. Causation and causal inference. In: Rothman & Greenland (eds). Modern Epidemiology..
Necessary vs. Sufficient cause? A specific disease
Sufficient causes of a disease Smoking Lung cancer (cause)? Smoking Lung cancer? Smoking Lung cancer sufficient cause?
(population) Risk Risk 10%, 10% risk? complement component 0% ( 90%) 100% ( 10%)., complement component.
A, B, E prevalence..
A complement component Population 1? E (strength of effects)?
A complement component Population 2 E (strength of effects)??
E complement component E complement cause A, B prevalence
Rothman Pie causal complement common,. (strength of a factor s effect), biologic significance., prevalence ( : incidence).
Quiz risk factor population attributable risk fraction (etiologic fraction) 100%? 100% 100%? 100%,? Phenylketonuria
Rothman s Pie? Cause causal component. Causal component,,, = causal component... PAR causal component,,, (causalism, causality)
Thomas McKeown s Thesis McKeown. The Role of Medicine: Dream, Mirage, or Nemesis? (,.. ) Szreter, AJPH (2002) Colgrove, AJPH (2002)
Thomas McKeown Physician, demographic historian Who had a strong interest in the relation of class, income, and living environment with health Department of Social Medicine, Birmingham Medical School, UK Publishing: 1950s-1980s 1955-1972: 4 articles in Population Studies Medicine in Modern Society, 1965 The Modern Rise of Population, 1976 The Role of Medicine: Dream, Mirage, or Nemesis?, 1976 The Origin of Human Disease, 1985
vs.
The McKeown Thesis..,,.,,,,,
? 1950 1970, 1950,,,. Effectiveness and Efficiency: Random Reflections on Health Services by Archibald Cochrane Medical Nemesis: The Expropriation of Health by Ivan Illich
Historical Controversy by Some Researchers In UK,,, (NHS ) New Right. In Canada, US, and UK government-supported medical service, (Lalonde ).
McKeown???
Controversies on McKeown s Findings The Population History of England 1541-1871 (1981) by E.A. Wrigley & Roger Schofield. The importance of social intervention in Britain s mortality decline 1850-1914: a reinterpretation of the role of public health (1988) by Simon Szreter (rising standards of living). 19.( ),.
, McKeown Consensus?., flawed.,, antithesis??
, McKeown s Thesis?,,.? / vs. /? (fundamental cause).
Type III Error: Right Answer to Wrong Question Schwartz & Carpenter. The right answer for the wrong question: consequences of type III error for public health research. AJPH (1999) Marmot. Improvement of social environment to improve health. Lancet (1998)
(inter-individual differences) (differences between populations or time periods)? (causes of individual cases within a population) (causes of incidence rates)?
Rothman s Pie & Sufficient Causes (an example) Rothman s Pie.,?
Population differences in the prevalence and patterning of risk factors for stroke, based in part on Rothman s model Diabetes A stroke, B. Chronic stress variation, variation. A chronic stress.
, -?,?
Type III Error : Homeless Homeless Homeless individual risk factor (,,, ) homeless? homeless?,? Individual susceptibility susceptibility intervention? Population incidence of homelessness vs. individual adaptation skill
Type III Error : Obesity? propensity for appetite dysregulation increased food advertising, Interaction btw the causes of interindividual differences within a population and the causes of the rate increase
Type III Error : Black-White Difference in Infant Mortality? infant mortality ( : low birthweight), ubiquitous ( : discrimination) Intervention infant mortality.,.
, Type III error?
Ecologic Fallacy Atomistic Fallacy
Fallacy of the Ecological Fallacy Individual-level models are more perfectly specified than ecological-level model. Ecological models are substitutes for individuallevel models. Only characteristics of individuals cause disease. Schwartz, AJPH (1994)
Water Quality and CVD Mortality in UK r=-0.67
Individuals and Populations Prevention and Prevention Paradox Rose. Strategy of prevention: lessons from cardiovascular disease. BMJ (1981) Rose. Sick individuals and sick populations. Int J Epidemiol (1985) Marmot. Improvement of social environment to improve health. Lancet (1998)
Relationship btw Population Average & Population Variation
Distribution of SBP in Middle-aged Men in Two Populations
The Contrasting Distribution of Serum Cholesterol in South Japan and Eastern Finland
The Shifting Distributions of SBP and BMI of Five Population Groups Aged 20-59 Derived from 52 Surveys in 32 Countries: Intersalt Study CV (coefficient of variation)?
Rates of homicide in Chicago and England and Wales by age and sex of perpetrator Homicide testosterone antagonists 20-24?
Alcohol,.,?,????
,? High-risk strategy Population strategy
Definition High-Risk Strategy, Intervention appropriate to individual Subject motivation/physician motivation Cost-effective use of resources Benefit:risk ratio favourable Difficulties and costs of screening Palliative and temporary-not radical Limited potential for individual and population
Screening and Predictability
Screening. : clinical preventive medicine
High-Risk Strategy
Relative Risk and Population Risk
Definition Population Strategy, Radical Large potential for population Behaviourally appropriate Small benefit to individual ( prevention paradox ) Poor motivation of subject/poor motivation of physician Benefit:Risk ratio worrisome
Definition Prevention Paradox intervention,, intervention Examples Immunization:, 600. Seat belt: 399. Life-style change: Framingham heart study, 1 heart attack, 49 40.,.
Multilevel Approach Diez-Roux. Multilevel analysis in public health research. Annu Rev Public Health (2000)
50 40 Mortality according to Class of Street and House, Manchester (1844) Class of the Streets 1st 2nd 3rd Mortality Rate per 1,000 30 20 10 0 1st 2nd 3rd 1st 2nd 3rd 1st 2nd 3rd Class of the Houses No Houses Engels (1845), p136
multilevel?
Math. Achievement Income level
Math. Achievement Catholic School Public School Income level
Multilevel Model: Example
Multilevel model Hierarchical model Random effect model Covariance components model
Compositional Effect vs. Contextual Effect
Multilevel Analysis HLM, Scientific Software International Lincolnwood, IL MLn, Institute of Education, London, UK VARCL, iec ProGAMMA, The Netherlands MIXREG, MIXOR & MIXNO, University of Illinois, Chicago, IL SAS procedure MIXED & SAS GLIMMIX macro, SAS Institute, Cary, NC BMDP5-V, Statistical Solutions Ltd., Ireland, UK
multilevel? Community, District, Province, National, International multilevel study.,.,.
Multilevel Lifecourse Context Multi-time point Accumulation Mechanisms Multi-level Extra-individual Biological, behavioral, psychological translation
Future for Epidemiology & Causation of Disease Krieger, Soc Sci Med (1994) Susser, AJPH (1996) Pearce, AJPH (1996) Susser, JECH (1998) McMichael, AJE (1999) McKinlay & Marceau, AJPH (1999)
Risk Factor Web of Causation Risk Factor Risk Factor Risk Factor Risk Factor Risk Factor Adult Disease Risk Factor Risk Factor Risk Factor Risk Factor Risk Factor Risk Factor McMahon (1960)
Epidemiology and the web of causation: has anyone seen the spider? (Nancy Krieger, 1994) Does risk factor epidemiology put epidemiology at risk? (Mervyn Susser, 1998) Prisoners of the proximate (A. J. McMichael, 1999)
Epidemiology, Disease Causation?
Eco-Epidemiology, Chinese Boxes Example: Global AIDS Epidemic Molecular level: means and timing of transmission Intermediate level: specific social behavior Population level: pattern of sexual relationships and of breast feeding Global level: interconnection btw societies Example: Peptic Ulcer Gastrophysiologist: wall of the stomach Neurophysiologist: autonomic nervous system Microbiologist: H. Pyroli Human geneticist: blood groups Epidemiologist: All + smoking, etc
Prisoner of the Proximate