Causal relationships, bias, and research designs Professor Anthony DiGirolamo
Causal Relationships
Sufficient Cause
Causal Relationships Sufficient Cause Precedes diseases, factor always present with disease
Causal Relationships Sufficient Cause Precedes diseases, factor always present with disease Necessary Cause
Causal Relationships Sufficient Cause Precedes diseases, factor always present with disease Necessary Cause Factor must be present for disease, but factor can be present without developing disease
Causal Relationships Sufficient Cause Precedes diseases, factor always present with disease Necessary Cause Factor must be present for disease, but factor can be present without developing disease Risk Factor
Causal Relationships Sufficient Cause Precedes diseases, factor always present with disease Necessary Cause Factor must be present for disease, but factor can be present without developing disease Risk Factor An exposure, behavior, or attribute that, if present, clearly influences the probability of disease
Determining Cause and Effect
Mill’s Cannons
Determining Cause and Effect Mill’s Cannons Strength Association shows a large difference
Determining Cause and Effect Mill’s Cannons Strength Association shows a large difference Consistency Association is always present with the disease
Determining Cause and Effect Mill’s Cannons Strength Association shows a large difference Consistency Association is always present with the disease Specificity No disease if the factor isn’t present
Determining Cause and Effect Mill’s Cannons Strength Association shows a large difference Consistency Association is always present with the disease Specificity No disease if the factor isn’t present Biological Plausibility Fits the natural history of the disease
Common Pitfalls What is bias?
Common Pitfalls What is bias? An event that produces deviations that shift data in a particular direction (skew your data)
Common Pitfalls What is bias? An event that produces deviations that shift data in a particular direction (skew your data) Common Types
Common Pitfalls What is bias? An event that produces deviations that shift data in a particular direction (skew your data) Common Types Assembly bias – characteristics of a group are not evenly distributed
Common Pitfalls What is bias? An event that produces deviations that shift data in a particular direction (skew your data) Common Types Assembly bias – characteristics of a group are not evenly distributed Selection bias – participants allowed to select which part of the study they are in
Common Pitfalls What is bias? An event that produces deviations that shift data in a particular direction (skew your data) Common Types Assembly bias – characteristics of a group are not evenly distributed Selection bias – participants allowed to select which part of the study they are in Detection bias – failure to detect true cause of disease
Common Pitfalls What is bias? An event that produces deviations that shift data in a particular direction (skew your data) Common Types Assembly bias – characteristics of a group are not evenly distributed Selection bias – participants allowed to select which part of the study they are in Detection bias – failure to detect true cause of disease Measurement bias –variations due to instrumentation or user error
Common Pitfalls Additional concerns when doing research…?
Common Pitfalls Additional concerns when doing research…? Random error
Common Pitfalls Additional concerns when doing research…? Random error Confounding
Common Pitfalls Additional concerns when doing research…? Random error Confounding Synergism
Research Design All research is descriptive, and results are directly related to the data assembled
Research Design All research is descriptive, and results are directly related to the data assembled What is a hypothesis ?
Research Design All research is descriptive, and results are directly related to the data assembled What is a hypothesis ? An “educated guess” about the relationship that exists in an observed phenomenon
Research Design All research is descriptive, and results are directly related to the data assembled What is a hypothesis ? An “educated guess” about the relationship that exists in an observed phenomenon Not always right; often need to re-test to truly discern relationship
Research Design All research is descriptive, and results are directly related to the data assembled What is a hypothesis ? An “educated guess” about the relationship that exists in an observed phenomenon Not always right; often need to re-test to truly discern relationship It’s called RE-SEARCH for a reason ! : )
Research Design 4 Key Factors of Research Designs
Research Design 4 Key Factors of Research Designs Enable comparison of a variable for two or more groups at a specified time
Research Design 4 Key Factors of Research Designs Enable comparison of a variable for two or more groups at a specified time Comparison must be quantified in absolute or relative terms
Research Design 4 Key Factors of Research Designs Enable comparison of a variable for two or more groups at a specified time Comparison must be quantified in absolute or relative terms Allow determination of the temporal sequence (when and how the factor and disease occur)
Research Design 4 Key Factors of Research Designs Enable comparison of a variable for two or more groups at a specified time Comparison must be quantified in absolute or relative terms Allow determination of the temporal sequence (when and how the factor and disease occur) Minimize bias, confounding, and other outside elements that may skew from true results
Research Designs Designs for Generating Hypotheses
Research Designs Designs for Generating Hypotheses Cross-Sectional Surveys
Research Designs Designs for Generating Hypotheses Cross-Sectional Surveys Quick, cost-effective studies done of a population at a certain point in time (calls, interviews, appointments, etc)
Research Designs Designs for Generating Hypotheses Cross-Sectional Surveys Quick, cost-effective studies done of a population at a certain point in time (calls, interviews, appointments, etc) Useful in determining prevalent risk factors in a population
Research Designs Designs for Generating Hypotheses Cross-Sectional Surveys Quick, cost-effective studies done of a population at a certain point in time (calls, interviews, appointments, etc) Useful in determining prevalent risk factors in a population Surveys sometimes have low response rates
Research Designs Designs for Generating Hypotheses Cross-Sectional Ecological Studies
Research Designs Designs for Generating Hypotheses Cross-Sectional Ecological Studies Relate the frequency of one characteristic (smoking) and an outcome (lung cancer) occurring in a geographical area
Research Designs Designs for Generating Hypotheses Cross-Sectional Ecological Studies Relate the frequency of one characteristic (smoking) and an outcome (lung cancer) occurring in a geographical area Downside is too many other factors and temporal issues may be overlooked or incorrectly identified
Research Designs Designs for Generating Hypotheses Cross-Sectional Ecological Studies Relate the frequency of one characteristic (smoking) and an outcome (lung cancer) occurring in a geographical area Downside is too many other factors and temporal issues may be overlooked or incorrectly identified Longitudinal Ecological Studies
Research Designs Designs for Generating Hypotheses Cross-Sectional Ecological Studies Relate the frequency of one characteristic (smoking) and an outcome (lung cancer) occurring in a geographical area Downside is too many other factors and temporal issues may be overlooked or incorrectly identified Longitudinal Ecological Studies Long-term surveillance or frequent cross-sectional surveys to measure trends in disease rates over many years
Research Designs Designs for Generating or Testing Hypotheses
Research Designs Designs for Generating or Testing Hypotheses Cohort Studies
Research Designs Designs for Generating or Testing Hypotheses Cohort Studies Compares two groups of clearly defined individuals, randomly selected
Research Designs Designs for Generating or Testing Hypotheses Cohort Studies Compares two groups of clearly defined individuals, randomly selected Observations made to examine if one group develops a condition with a risk factor, the other without
Research Designs Designs for Generating or Testing Hypotheses Prospective vs. Retrospective Cohort Studies
Research Designs Designs for Generating or Testing Hypotheses Prospective vs. Retrospective Cohort Studies Prospective cohorts are done at present time, baseline data recorded and then observed over time
Research Designs Designs for Generating or Testing Hypotheses Prospective vs. Retrospective Cohort Studies Prospective cohorts are done at present time, baseline data recorded and then observed over time Retrospective cohorts look back in time to a risk group and follow members in a current day scenario to see what conditions are now prevalent
Research Designs Designs for Generating or Testing Hypotheses Prospective vs. Retrospective Cohort Studies Prospective cohorts are done at present time, baseline data recorded and then observed over time Advantages are control of standardized procedures and diagnosis, estimates of risk generated are true, many different disease outcomes may be studied at once Disadvantages are high costs, loss of follow-up, long wait for final results Retrospective cohorts look back in time to a risk group and follow members in a current day scenario to see what conditions are now prevalent
Research Designs Designs for Generating or Testing Hypotheses Prospective vs. Retrospective Cohort Studies Prospective cohorts are done at present time, baseline data recorded and then observed over time Advantages are control of standardized procedures and diagnosis, estimates of risk generated are true, many different disease outcomes may be studied at once Disadvantages are high costs, loss of follow-up, long wait for final results Retrospective cohorts look back in time to a risk group and follow members in a current day scenario to see what conditions are now prevalent Good to estimate absolute risk, but lacks ability to control data collection and standardization
Research Designs Designs for Testing Hypotheses
Research Designs Designs for Testing Hypotheses Randomized Controlled Clinical Trials (RCCT)
Research Designs Designs for Testing Hypotheses Randomized Controlled Clinical Trials (RCCT) Patients placed into two groups, one receiving a treatment the other a placebo (usually a therapeutic treatment) Use of “blinding” in trials ?
Research Designs Designs for Testing Hypotheses Randomized Controlled Clinical Trials (RCCT) Patients placed into two groups, one receiving a treatment the other a placebo (usually a therapeutic treatment) Use of “blinding” in trials ? Patients are unaware of which group they are in (single), those dosing the patients also do not know (double) Key is that both groups must be treated equally
Research Designs Designs for Testing Hypotheses Randomized Controlled Clinical Trials (RCCT) Patients placed into two groups, one receiving a treatment the other a placebo (usually a therapeutic treatment) Use of “blinding” in trials ? Patients are unaware of which group they are in (single), those dosing the patients also do not know (double) Key is that both groups must be treated equally Randomized Controlled Field Trials (RCFT)
Research Designs Designs for Testing Hypotheses Randomized Controlled Clinical Trials (RCCT) Patients placed into two groups, one receiving a treatment the other a placebo (usually a therapeutic treatment) Use of “blinding” in trials ? Patients are unaware of which group they are in (single), those dosing the patients also do not know (double) Key is that both groups must be treated equally Randomized Controlled Field Trials (RCFT) Focuses on using a preventative measure, rather than a therapeutical one (RCCT)
Well Done!!