Presentation is loading. Please wait.

Presentation is loading. Please wait.

DAY 4 Considerations for Interpretation Confounding 1.

Similar presentations


Presentation on theme: "DAY 4 Considerations for Interpretation Confounding 1."— Presentation transcript:

1 DAY 4 Considerations for Interpretation Confounding 1

2 Introduction Understanding what the data can tell us and cannot tell us 2

3 3 Approach to understanding the data If we see an association, what other possible explanations can there be besides causation? - Confounding

4 Confound “To confuse” 4 Webster’s Online Dictionary www.webster-dictionary.orgwww.webster-dictionary.org “To mistake for another”

5 5 Guilt by association

6 Not Observed James Cause X Observed Broken Window George Guilt by association

7 Scenario 1 – Individual level Hugo Lennie

8 Hugo Not Observed Empty Beer Bottles Lennie Cause Association Cause X Observed Scenario 1 – Individual level

9 Scenario 2 – Population level

10 Try Alcohol Not Try Alcohol 6 Yes R-Rated Movies No R-Rated Movies 90 b d a c Prevalence Prevalence Ratio Scenario 2 – Population level People who _____________ are ______times as likely to __________ as people who ____________.

11 Trying Alcohol Watching R-Rated Movies “The study establishes a clear progression that begins with watching R-rated movies and leads to teenagers trying alcohol.” “If we can keep our teenagers from watching R-rated movies, then we can go a long way towards preventing teenagers from trying alcohol.” Cause Scenario 2 – Population level

12 ? Trying Alcohol Watching R-Rated Movies Cause Association Cause X Scenario 2 – Population level

13 13 A situation in which an association between an exposure and an outcome is observed as a result of the influence of a third, unobserved, factor, called a confounder. The true effect of an exposure on an outcome is distorted because it is mixed up with another factor that can influence the outcome. The confounder is both associated with the observed exposure and is a possible cause of the outcome. Confounding

14 Unobserved Exposure Try alcohol Watch R-rated movies Limited Adult Supervision Watch out for lurking variables!

15 Unobserved Exposure Try alcohol Watch R-rated movies Willingness to Break the Rules

16 Unobserved Exposure Try alcohol Watch R-rated movies Poor Self-Control

17 Unobserved Exposure Try alcohol Watch R-rated movies Parental attitudes

18 Unobserved Exposure Try alcohol Watch R-rated movies Risk taker

19 Lack of Supervision or Poor Self- Control or Risk taking or... Try alcohol Watch R-rated movies X … and you avoided or eliminated the hypothesized cause, what would happen to the outcome? ? If the association was due to confounding …

20 Cause Association Cause X Confounder Playing Volleyball Skin Cancer

21 Cause Association Cause X Confounder Drinking Alcohol Lung Cancer

22 Cause Association Cause X Confounder Eating Pretzels Auto Accidents

23 Cause Association Cause X Confounder Watching TV Acne

24 Cause Association Cause X Confounder Drinking Coffee Cancer

25 Cause Association Cause X Confounder Driving Motorcycles Hepatitis C

26 Cause Association Cause X Confounder Playing Baseball Oral Cancer

27 Cause Association Cause X Confounder Sand Skin Cancer

28 Association is not necessarily causation. Always Remember

29 Supplementary Lesson about Confounding

30 The confounder is associated with the observed exposure (but not affected by the exposure). The confounder is a possible cause of the outcome. Two criteria for Confounding Possible Confounder Outcome Exposure

31 31 We suspect a factor to be a potential confounder based on intuition and evidence that it is associated with the exposure under study and a cause of the outcome under study But is this true in the data at hand? We can assess this by calculating the strength of association between: 1) exposure and the suspected confounder; and 2) between suspected confounder and outcome. Possible Confounder Outcome Exposure PR = ?

32 32 400 1000 40 % 4006001000 a b c d Exercise less Exercise more Total Increased BP No increased BP Prevalence 200 1000 20 % 2008001000 Prevalence Ratio 2.0 ÷ a a+b c c + d People who exercise less are more likely to have high blood pressure. People who exercise less are 2.0 times as likely to have increased blood pressure as those who exercise more. 2000

33 Cause Association Cause X Confounder? Exercise less Increased blood pressure

34 Smoking Increased blood pressure Exercise Less PR = ? PR = 2.0 Assessment of smoking as a potential confounder See if it meets two criteria for confounding 1. Associated with the exposure? 2. A possible cause of the outcome?

35 35 450 1000 45.0 % 4505501000 a b c d Less exercise More exercise Total Smokers Non- smokers Prevalence 175 1000 17.5 % 1758251000 Prevalence Ratio 2.6 People who exercise less are 2.6 times as likely to be smokers as those who exercise more. 2000 1. Is the potential confounder associated with the exposure? (Is smoking associated with exercise?)

36 Smoking Increased Blood Pressure Exercise Less PR = 2.6 PR = ? PR = 2.0

37 37 475 625 76.0 % 475150625 a b c d Smokers Non- smokers Total Increased BP No increased BP Prevalence 125 1250 10.0 % 12512501375 Prevalence Ratio 7.6 Smokers are 7.6 times as likely to have increased blood pressure compared to non-smokers. 2000 2. Is the potential confounder a possible cause of the outcome? (Is smoking a possible cause of increased blood pressure?)

38 Smoking Increased Blood Pressure Exercise Less PR = 2.6 PR = 7.6 PR = 2.0 Is there Possible Confounding?

39 39 Marijuana Use and Getting Tattoos

40 People who ____________________________________________ are ______ times as likely to _______________________________ compared to people who __________________________________ 40 50 200 25 % 50150200 a b c d Marijuana User Not a marijuana User Total Tattoos No Tattoos Prevalence 30 200 15 % 30170200 Prevalence Ratio 1.7 ÷ a a+b c c + d People who smoke marijuana are more likely to get tattoos than people who do not smoke marijuana.

41 People who use marijuana are 1.7 times as likely to get tattooed compared to people who do not use marijuana. 41 50 200 25 % 50150200 a b c d Marijuana User Not a marijuana User Total Tattoos No Tattoos Prevalence 30 200 15 % 30170200 Prevalence Ratio 1.7 People who smoke marijuana are more likely to get tattoos than people who do not smoke marijuana.

42 Cause Association Cause X Confounder? Use Marijuana Get Tattoos

43 Access to Money Get Tattoos Use Marijuana PR = ? PR = 1.7 Assessment of “access to money” as a potential confounder See if it meets two criteria for confounding 1. Associated with the exposure? 2. A possible cause of the outcome?

44 44 150 200 75.0 % 15050200 a b c d Total Prevalence 50 200 25.0 % 50150200 Prevalence Ratio 3.0 People who use marijuana are 3.0 times as likely to have access to money than non-users. 400 Marijuana User Not a marijuana User Access to Money No Access to Money 1. Is the potential confounder associated with the exposure? (Is access to money associated with marijuana use?)

45 Access to Money Get Tattoos Use Marijuana PR = 3.0 PR = ? PR = 1.7

46 46 60 140 42.8% 60140200 a b c d Total Prevalence 20 180 11.1 % 20180200 Prevalence Ratio 3.9 People with access to money are 3.9 times as likely to get tattoos than people without access to money. 400 Access to Money No Access to Money Tattoos No Tattoos 2. Is the potential confounder a possible cause of the outcome? (Is access to money a possible cause of getting tattoos?

47 Access to Money Get Tattoos Use Marijuana PR = 3.0 PR = 3.9 PR = 1.7 Is there Possible Confounding?

48 Possible Confounder Outcome Exposure PR = 0.97 PR = 1.10 PR = 1.8 Associated with the exposure? A possible cause of the outcome? Another Example - Is there Possible Confounding?

49 Day 4 First Morning Breakout Assessing Possibility of Confounding 49

50 50 Use the 54 cards in the deck to explore the potential for confounding in this data set Deck Worksheet – pages 3-5

51 Less Frequent Breakfast High Physical Activity High Sleep PR = ? Associated with the exposure? A possible cause of the outcome? Assess Possibility of Confounding by Frequency of Breakfast

52 Assessing confounding in your study Assembly to brainstorm ideas... 52

53 Day 4 Second Morning Breakout Possible confounders you should examine? 53

54 Possible Confounder(s)? Your Outcome Your Exposure Discuss possible confounding in your study Label your possible confounding triangle on large post-it

55 Breakout Deliverable for Assembly/Gallery walk 55 Large post-it with your team’s confounding triangle (During gallery walk after the break, use small post-its to provide constructive comments and suggestions on other teams’ triangles).

56 56 Study Proposal: Section 7a 7. Considerations for Interpretation 7a. Confounding Describe how a confounding variable could influence your hypothesized association Develop one or two additional study questions to help assess possible confounding If you have time, start this...

57 57 Breakout Session Tips Create one or two questions for your survey that will assess possible confounding. Your question(s) must be school appropriate Limit your answer options to two answers, such as Yes/No, Male/Female, or other 2- choice answers. Anything else?

58 58 Reminder It should be written so that its intended audience understands it. It should mean the same thing to everyone who reads it. The answer options must categorize and cover the entire range of possible behavior The time period to consider when answering a particular question must be appropriate (long or short enough) Criteria for a good survey question


Download ppt "DAY 4 Considerations for Interpretation Confounding 1."

Similar presentations


Ads by Google