5.1.1 sufficient component cause model
- 1. Causality Outline – Causality – Models of causality • Sufficient-component cause model, aka Rothman’s pies • Counterfactual framework • Hill’s criteria or “viewpoints” • Graphical models/DAGs (Presented by Dr. Jade Benjamin-Chung) • Causal perspective on effect modification (Presented by Dr. Jade Benjamin-Chung) – Summary 1
- 2. 2 Causality in epidemiology • Goals of epidemiology – Describing distribution of disease – Predicting disease – Identifying causes of disease
- 3. 3 Causality • What is a cause? • “An antecedent event, condition, or characteristic that was necessary for the occurrence of the disease at the moment it occurred, given that other characteristics are fixed.” (emphasis added)
- 4. 4 Causality • “An event, condition, or characteristic that preceded the disease onset and that, had the event, condition, or characteristic been different in a specified way, the disease either would not have occurred at all or would not have occurred until some later time.”
- 5. 5 Causality • “We may define a cause to be an object followed by another… where, if the first object had not been, the second never had existed” (Hume 1748)
- 6. 6 Causality Outline – Causality – Models of causality • Sufficient-component cause model, aka Rothman’s pies • Counterfactual framework • Hill’s criteria or “viewpoints” • Graphical models/DAGs (Will be presented by Dr. Jade Benjamin-Chung) • Causal perspective on effect modification (Will be presented by Dr. Jade Benjamin-Chung) – Summary
- 7. 7 Models of causality • Epidemiology uses several “models of causation that may be useful for gaining insight about epidemiologic concepts” – Conceptual model vs statistical model – Why are there several models and not just one? • Each causal model has different strengths and helps illuminate different epidemiologic concepts and accomplish different tasks relevant to epidemiologic research
- 8. 8 Models of causality – Sufficient-component cause model, aka Rothman’s pies – Counterfactual framework – Graphical models/DAGs – Hill's criteria or “viewpoints”
- 9. 9 Causality Outline – Causality – Models of causality • Sufficient-component cause model, aka Rothman’s pies • Counterfactual framework • Hill’s criteria or “viewpoints” • Graphical models/DAGs (Will be presented by Dr. Jade Benjamin-Chung) • Causal perspective on effect modification (Will be presented by Dr. Jade Benjamin-Chung) – Summary
- 10. Sufficient-component cause • Model oriented around mechanisms of disease causation A U B 10
- 11. Sufficient-component cause • Example: sufficient cause of impaired brain function PKU 1 1 phenylalanine
- 12. 1 2 Sufficient-component cause • Sufficient cause = “a complete causal mechanism, a minimal set of conditions and events that are sufficient for the outcome to occur.” – Can be (and almost always are) more than one for any outcome – If none occur, then the outcome will not occur – Can (and almost always does) include unknown causes
- 13. Sufficient-component cause • Example of two sufficient causes of CVD – A=smoking, B1=genetic profile 1, B2=genetic profile 2, C=high cholesterol U1 A U2 1 3 A B1 B2 C
- 14. Sufficient-component cause • Component cause = one member (one “slice”) of a set of causes that creates a sufficient cause; blocking it will result in the outcome not happening U1 A B1 U2 1 4 A B2 C
- 15. Sufficient-component cause • Necessary cause = occurs in all sufficient causes; same “slice” is in everypie – HPV is a necessary cause of cervical cancer – HIV is a necessary cause of AIDS U1 A B1 U2 1 5 A B2 C
- 16. Sufficient-component cause • Causal complement = for any component cause, the set of other component causes in the sufficient cause is the causal complement (the rest of the pie) – A, B2, U2 are the causal complement of C U1 A B1 U2 1 6 A B2 C
- 17. 1 7 Sufficient-component cause • Unknown component causes (Us) – The Us are also very important – almost every pie has some unknown component U – For almost every outcome there is an entire U pie
- 18. 18 Example • Bicycle commuting According to the US Census Bureau’s 2008 American Community Survey (ACS) 0.55% of American workers use a bicycle as the primary means of getting to work. This is up 14 percent since 2007, 36 percent from the first ACS in 2005, and 43 percent since the 2000 Census.
- 20. Example • My bicycle fall – A=road surface, B=traffic, C=speed, D=bicycle characteristics A B U C D 2 9
- 21. 2 1 S-c cause contributions • Depicts multifactorial causality in disease • Clearly illustrates “necessary” and “sufficient” causes • Illuminates meaning of strength of associations, attributable percentages, and effect modification (will return to this throughout the course) – Note: effect modification exists when the effect of one exposure is different depending on the value of another exposure
- 22. 2 2 S-c cause limitations • More useful in concept than in application to particular research questions – What scope of causes get into a pie? – So many pies for most outcomes • Not helpful for depicting sequential mechanisms or direct and indirect effects • Specifies details that go beyond the scope of epidemiologic data
- 23. 2 3 Example • Bicycle fall – Illustration of sufficient-component cause model illuminating strength of associations
- 24. Example U A B C D • Study of the effect of traffic on bicycle falls – A=road surface, B=traffic, C=speed, D=bicycle characteristics Setting 1: 10% of bicycling routes have poor quality road surface Setting 2: 90% of bicycling routes have poor quality road surface 2 4
- 25. 2 5 Example • Will the strength of the association between traffic and falls appear the same in settings 1 and 2? – Assuming frequency of other component causes of this sufficient cause and other distinct sufficient causes are the same between settings • Strength of association depends on how common the rest of the pie is in the population you study (i.e., strength of association depends on the prevalence of the causal complement)
- 26. 2 6 Sufficient-component cause • We will return to this model when we discuss – Strength of associations (on different scales) – Attributable percentage measures – Effect modification