Term
|
Definition
| 1. Confounding, 2. Misclassification/Info bias, 3. Selection bias |
|
|
Term
| If bias is conservative, effect will be biased towards... |
|
Definition
|
|
Term
|
Definition
| Probability that the test statistic is more extreme than the observed, given that the Ho is true |
|
|
Term
| Name 4 threats to causal inference |
|
Definition
| 1. Lack of precision, 2. Lack of internal validity, 3. Incorrect asses. of direction of causality, 4. Lack external validity |
|
|
Term
| What is the counterfactual definition of confounding? |
|
Definition
| The freq. of the disease in the unexposed is different from the freq. of disease in the counterfactual exposed pop. had they been unexposed |
|
|
Term
| What are the 3 necessary criteria to define confounding? |
|
Definition
| 1. Confounder is a cause of outcome, 2. Confounder is associated with exposure, 3. Confounder is no affected by exposure (not in causal pathway) |
|
|
Term
| What is a backdoor pathway? |
|
Definition
| A non-causal association between exposure and outcome. They begin with arrows pointing into the exposure |
|
|
Term
| What is Simpson's paradox? |
|
Definition
| Non collapsability does not assure non- confounding |
|
|
Term
| By how much is commonly accepted that an adjusted OR differ from a crude OR to say that confounding is present? |
|
Definition
|
|
Term
| Name 3 methods to control confounding in design stage |
|
Definition
| !. Randomization, 2. Restriction, 3. Matching |
|
|
Term
| Name 3 methods to control confounding in the analysis |
|
Definition
| 1. Standarization, 2.Stratification, 3. Multivariate models |
|
|
Term
| How do you standardize with the direct method? |
|
Definition
| Multiply observed incidence rates by age by a standard population weight |
|
|
Term
| Name limitations of standardized rates |
|
Definition
| 1. Need to know stratum spec. rates, 2. Absolute magnitude depends on choice of standard pop., 3. Adjusted rates can only be compared across pops if same standard pop is used |
|
|
Term
| How do you calculate standardized mortality rates (SMR)? |
|
Definition
| SMR= observed/expected x 100 |
|
|
Term
| What is non-differential error? |
|
Definition
| Measurement error doesn't depend on levels of other variables (error is the same for outcome, confounder, etc.) |
|
|
Term
| What is independent error? |
|
Definition
| Measurement error in a variable is not associated with errors in measuring other variables |
|
|
Term
| Non-differential, independent measurement errors bias estimate towards... |
|
Definition
|
|
Term
| What is regression towards the null? |
|
Definition
| When a variable is measured with random error and we select participants with observed extreme values, their true underlying values are on average closer to the pop.mean |
|
|
Term
| When can regression towards the mean be important? |
|
Definition
| in pre/post studies. An association may appear for the intervention just because there is regression towards the mean |
|
|
Term
| Name 2 biases that can result in differential measurement error |
|
Definition
| 1. Surveillance bias, 2. Diagnostic bias |
|
|
Term
|
Definition
| When selection into study is affected both by exposure and cause of outcome |
|
|
Term
| What is emigrative selection bias? |
|
Definition
| When losses to follow-up are differential |
|
|
Term
| What types of biases can result from a healthy worker effect? |
|
Definition
| Selection bias and counfounding |
|
|
Term
| What is confounding by indication? |
|
Definition
| A variable is a risk factor for a disease among nonexposed persons and is associated with the exposure of interest in the population from which the cases derive |
|
|
Term
|
Definition
| Cohort study in which exposure assignment is determined by study |
|
|
Term
|
Definition
| Technique used in RTC to avoid serious imbalances in # of participants in each group |
|
|
Term
|
Definition
|
|
Term
|
Definition
|
|
Term
|
Definition
| Randomized controlled study |
|
|
Term
|
Definition
|
|
Term
| What is a factorial trial for? |
|
Definition
| To test >1 intervention compared to a control in a single trial |
|
|
Term
| What is intention to treat analysis? |
|
Definition
| To analyze groups as they were assigned (analyze as randomized) |
|
|
Term
|
Definition
| all cohort members with the same opportunity for observing the event, and that have been followed for the same amount of time without the event |
|
|
Term
| How do you assemble a nested-case control study? |
|
Definition
| !. Identify cases, 2. Reorder according to outcome time scale, 3. For each case, identify risk set, 4. Randomly select controls from risk set |
|
|
Term
| What is incidence density sampling? |
|
Definition
| Compare cases to non-cases (at event time) among cohort members |
|
|
Term
| What is cumulative incidence sampling? |
|
Definition
| Select all cases. Those without outcome at the end of study period are controls |
|
|
Term
| How do you assemble a case-cohort? |
|
Definition
| 1. Select a subcohort at start of risk period, 2. Identify cases in subcohort, 3. Include remaining cases outside subcohort |
|
|
Term
| What is an absorbent state? |
|
Definition
| Once exposed, always exposed (no need for more assessments) |
|
|
Term
| In a case-cohort, can cases outside the cohort be controls? |
|
Definition
|
|
Term
| In a nested case-control, can a case be a control? |
|
Definition
| YES. A case may serve as a control for a previous case |
|
|
Term
| In a nested case-control, can an individual be a control at several times? |
|
Definition
|
|
Term
| In a nested case-control, outcome is/are... |
|
Definition
|
|
Term
| In a case-cohort, the outcome is/are... |
|
Definition
|
|
Term
| Strengths of the case control studies |
|
Definition
| Efficient, shorter time period, good for rare diseases, may assess multiple outcomes |
|
|
Term
| In case control studies, when does the OR approximate the RR? |
|
Definition
| When incidence sampling is used |
|
|
Term
| How do you avoid selection bias in case control studies? |
|
Definition
| By selecting cases independent of exposure |
|
|
Term
|
Definition
| Type of selection bias in hospital based C-C studies. Combination of exposure and disease increases risk of admission to hospital |
|
|
Term
| What type of sampling better represents person-time exposure in C-C studies? |
|
Definition
| Incidence density sampling. Controls are selected at the time of each incident case |
|
|
Term
| How do you assemble a case control study? |
|
Definition
| 1. Identify target and source pop, 2. Recruit cases and controls, 3. Ascertain exposure status |
|
|
Term
| Is target and source population always the same for cases and controls? |
|
Definition
| Target pop. should be the same, source can differ |
|
|
Term
| What are elegibility criteria? |
|
Definition
| Characteristics that define the target and source pop. Exclusion+ inclusion criteria |
|
|
Term
|
Definition
| Awareness of a possible association between exposure and outcome, where exposed individuals are followed more closely and detection of cases is thus increased |
|
|
Term
| What are concerns when using prevalent cases? |
|
Definition
| 1. Survivor cases, 2. Change of exposure due to disease, 3. Reverse causation |
|
|
Term
| Name sources for control selection |
|
Definition
| 1. Same as cases, 2. Geographic related (neighbors), 3. Random digit calling, 4. friends, family, 5. Same hospital/clinic |
|
|
Term
| What is individual matching? |
|
Definition
| Pairing or grouping controls to cases by a known risk factor |
|
|
Term
| How do you calculate a match OR? |
|
Definition
| OR=#pairs expo case and unex control/ #pairs case unexpo and control expo |
|
|
Term
| Can you produce a selection bias with matching? |
|
Definition
| YES. If selecting on a confounder in case control studies |
|
|
Term
| On what depends the bias of the crude OR compared to a matched OR? |
|
Definition
| On how much the matched factor is associated with the exposure |
|
|
Term
| What is frequency matching? |
|
Definition
| Selection of controls based on freq. of case characteristic so that they have the same distribution of potential confounder |
|
|
Term
| What are limitations of matching? |
|
Definition
| Effect of match var on outcome cannot be assessed, costly, can overmatch, can match on exposure or surrogate |
|
|
Term
| What is the invariance of the OR mean? |
|
Definition
| OR of exposure comparing diseased & non-diseased = OR of disease comparing expo & un expo |
|
|
Term
| Can the RR be calculated from the OR? |
|
Definition
|
|
Term
| The OR of disease is the ______ of the OR of non-disease |
|
Definition
|
|
Term
| 2 definitions of effect modification |
|
Definition
| 1. The causal effect of exposure is different by the categories of other variable(s) 2. Joint causal effect of 2 vars is different from what would be expected by adding the independent effects |
|
|
Term
| Important characteristics of effect modification |
|
Definition
| 1. Scale dependent (additive & multiplicative), 2. Is symmetrical, 3. Causal effect that should be estimated |
|
|
Term
| Type of scale (multiplicative or additive) more relevant for PH |
|
Definition
|
|
Term
| Type of scale (additive or multiplicative) that better describes observed patterns of disease |
|
Definition
|
|
Term
| What type of error is increased when testing for interaction? |
|
Definition
| Both type I (increased in multiple comparisons) and type II more diff, to identify true interactions because of great random variability) |
|
|
Term
| What are Hill's criteria for causation? |
|
Definition
| 1. Strength of association, 2. Temporality, 3. Plausibility, 4. Consistency, 5. Specificity, 6. Bio gradient, 7. Coherence, 8. Experiment, 9. Analogy |
|
|
Term
| What is a sufficient cause? |
|
Definition
| Minimal set of conditions and events that are sufficient for the outcome to occur. A complete causal mechanism |
|
|
Term
| What is a necessary cause? |
|
Definition
| A component cause that appears in every sufficient cause |
|
|
Term
|
Definition
| Resistance of a group to an attack by a disease to which a large proportion of members are immune |
|
|
Term
| What is a DAG (Directed Acyclic Graph)? |
|
Definition
| Set of arrows drawn along a time line, characterizing causal and temporal relationships between variables |
|
|
Term
| What is the Study population? |
|
Definition
| Group under observation in study with the same exposure-disease assoc. (although not necessarily representative) |
|
|
Term
| What is the Target population? |
|
Definition
| Etiologic goal of Epi research, not always numerable |
|
|
Term
| What is the Source population? |
|
Definition
| Group expected to have the same expo-disease association and can be numerated |
|
|
Term
| If individuals enter together, non are added in follow up and exit for endpoint only, what type of cohort is it? |
|
Definition
|
|
Term
| What is an open cohort based on entry and exit? |
|
Definition
| A cohort where ind. enter at different times and exit for various reasons |
|
|
Term
| What are censored individuals? |
|
Definition
| !. Those that didn't experience endpoint of interest during follow up, 2. Those that cannot be observed because lost, dead, schedule end |
|
|
Term
|
Definition
| Cumulative time spent by each individual at risk in the population |
|
|
Term
| What is a lagged exposure? |
|
Definition
| The period of exposure time not relevant for outcome |
|
|
Term
| Why is etiological relevant exposure measured? |
|
Definition
| Counting wasted exposure tends to dilute associations |
|
|
Term
| 3 reasons for apportioning person-time |
|
Definition
| 1. Risk varies with age, 2. Risk varies over time (secular trend), 3. Risk for disease may vary over expo levels |
|
|
Term
| What is the cumulative incidence (R)? |
|
Definition
| The proportion of a closed cohort at risk that presents the outcome within a period of time (total events/total people at risk) |
|
|
Term
| 3 characteristics of cumulative incidence (R) |
|
Definition
| 1. Unitless, 2. Represented as probability or %, 3. Numerator included in denominator |
|
|
Term
| What is cumulative survival (S)? |
|
Definition
| Compliment of cumulative incidence (S= 1-R) |
|
|
Term
| When do you use a life table vs a Kaplan Meier to estimate S? |
|
Definition
| Life table if exact time to event is unknown (assumes W is at t1/2), K-M if event time is known |
|
|
Term
| What is an incidence rate? |
|
Definition
| Number of new cases that occur in a specific period of time in a pop. at risk (#new cases/person-time) |
|
|
Term
| What is the relation between incidence and prevalence? |
|
Definition
|
|
Term
|
Definition
| The probability of an event divided by the prob. of non-event (Odds= p/1-p) |
|
|
Term
| How can risk be estimated form an odds? |
|
Definition
|
|
Term
| What is the counterfactual model? |
|
Definition
| The risk experience an exposed individual would have had, had he not been exposed, all else being equal |
|
|
Term
| What is the immortal person-time? |
|
Definition
| Time before the study the individual had to survive in order to be included |
|
|
Term
| 2 consequences of disregarding immortal person-time |
|
Definition
| 1. Overestimation of total time at which the individual was at risk, 2. Underestimation of incidence rate for endpoint |
|
|
Term
|
Definition
| Proportion of positive tests given the disease is truly present (a/a+c) |
|
|
Term
|
Definition
| Proportion of negative tests given they are truly disease free (d/d+b) |
|
|
Term
|
Definition
| Given a positive test, proportion of individuals who actually have the disease |
|
|
Term
|
Definition
| Given a negative test, proportion that are not diseased |
|
|
Term
| What is the goal of cross-sectional studies? |
|
Definition
| To compare the prevalence of an outcome in exposed and unexposed |
|
|
Term
| What is the goal of cohort studies? |
|
Definition
| To compare the incidence of a disease in exposed an unexposed |
|
|
Term
| What is the goal of case-control studies? |
|
Definition
| To compare the odds of exposure in cases and controls (implies odds of disease in exposed and unexposed) |
|
|
Term
| What is the true causal effect? |
|
Definition
| Difference in the freq. of disease in a pop. with everyone exposed minus the freq. of disease in the same pop., during the same time, but unexposed |
|
|
Term
| What is the exchangeability/randomization assumption? |
|
Definition
| The mean in the unexposed group is exchangeable with the mean in the counterfactual exposed group |
|
|
Term
| How does stratification provide with exchangeability? |
|
Definition
| After stratification, exchangeability will hold between exposed and unexposed within the strata of the confounder |
|
|
Term
|
Definition
| A variable that is the causal pathway between exposure and outcome |
|
|
Term
| What is the difference between a full and a partial mediator? |
|
Definition
| Full: exposure has an indirect effect on outcome. Partial: exposure has a direct effect on outcome |
|
|
Term
| According to the Prentice criteria, what is a surrogate? |
|
Definition
| 1. Must be correlated to outcome, 2. Must be a full mediator, 3. Must fully capture the effect of tx on an outcome |
|
|
Term
| What is a collider-stratification bias? |
|
Definition
| The opening of a backdoor pathway due to the adjustment of a mediator |
|
|
Term
| How many reference categories are there in a heterogeneity of effects table? |
|
Definition
| 2 reference categories (within group comparisons) |
|
|
Term
| What are Cook's types of prediction models? |
|
Definition
| 1.Diagnostic (accurate identification of current disease), 2. Prognostic (estimate risk of a future unknown event) |
|
|
Term
|
Definition
| Ability to correctly estimate a disease state or risk of future event |
|
|
Term
|
Definition
| Ability to separate persons with and without the disease |
|
|
Term
|
Definition
| Sensitivity (Y axis) and 1-specificity (X axis) |
|
|
Term
| What is the C statistic or AUC? |
|
Definition
| The probability that a randomly selected individual with disease has a test result indicating greater suspicion of disease |
|
|
Term
| What is the hierarchy for model validation? |
|
Definition
| 1. No validation, 2. Internal validation, 3. Prospective validation, 4. External validation |
|
|
Term
| Missing data can introduce what type of bias? |
|
Definition
|
|
Term
| What does missing completely at random assume? |
|
Definition
| the probability of missing data values does not depend on anything but chance |
|
|
Term
| What does missing at random assume? |
|
Definition
| the probability of missing data values may depend on observed values in the study, but not on missing values |
|
|
Term
| What does not missing at random assume? |
|
Definition
| the probability of missing data depends on the value of some missing data |
|
|
Term
| What are the causal response types? |
|
Definition
1. No effect. Doomed
2. Exp Causal
3. Exp Protective
4. No effect. immune |
|
|
Term
| What are the potential outcomes? |
|
Definition
D=1 outcome had they been exposed
D=0 outcome had they been un exposed
At best one is observed although both have a true value per person |
|
|
Term
| What is the ecological fallacy? |
|
Definition
| error of interpretation of statistical data; inferences about the nature of specific individuals are based solely upon aggregate statistics collected for the group to which those individuals belong. |
|
|
Term
| What is competing risk bias? |
|
Definition
| Type of seection bias where outcomes are mutually exclusive |
|
|
Term
| What is ascertaintment bias? |
|
Definition
| Type of selection bias where patients gathered do not represent the cases originated in the population |
|
|
Term
|
Definition
| Also called selective survival bias, type of selection bias where the exposure is determinant of or related to prognosis factors |
|
|
Term
| What is the healthy worker effect? |
|
Definition
| Lower mortality rate in employed individuals compared to the general population |
|
|
Term
| What is the Hawthorne effect? |
|
Definition
| Increase in the outcome under study in participants who are aware of being observed |
|
|
Term
| What is a protopathic bias? |
|
Definition
| When exposure is influenced by early (subclinical) stages of disease |
|
|
Term
| What is a contamination bias? |
|
Definition
| When intervention-type activities find their way to the control group |
|
|
