Epidemiology Final (copy) dadfadddd

Pros of RCTs asdfasdasdf

ability to randomize = reduces likelihood that grps will differ significantly, can control for known & unknown confounders; provide the greatest control over:

\- amount of exposure

%%- timing & freq. of exposure%%

%%- period of observation%%

%%editing a flashcard%%

\
\
\
asdfkljasdf

Selection Bias

the procedures used to select subjects & factors that influence participation in a study that distorts the causal relationship b/w exposure & outcome; can happen due to

\- procedures used to include/exclude subjects

\- differences b/w participants & nonparticipants

\- differences in follow-up/drop-out rates that differ according to a causal exposure

Minimizing Selection Bias

\- keep non-response as low as possible

\- if possible, collect info on non-responders

Information Bias

the incorrect determination of exposure and/or outcome; types of bias include:

\- diagnostic bias: exposed may be followed more intensively than unexposed (suspicion) or the probability of being classified as having the disease is related to exposure status (classification)

\- recall bias: diseased participants may recall exposure differently than non-diseased participants (relevant in case-control & cross-sectional studies)

\- bias in abstracting records: when abstracting medical records to find cases, exposure status may influence the depth of the inquiry in the record (ex. looking at influence of anxiety on heart attacks -> look more closely at cases w/ anxiety -> see more of a relationship b/s anxiety & heart attacks)

\- interview bias: interviewers not blinded to the exposure/disease status might ask questions differently according to exposure/disease status

\- analytic bias: when statisticians/epidemiologists analyze data with a preference for a specific outcome -> could consciously/unconsciously influence analysis process toward preferred outcome

\- Hawthorne effect: participants in a study change their behaviors because they know they're in a study

\- Social Desirability: tendency of people to respond to survey questions in a manner that will be viewed favorably by others

Minimizing Information Bias

address in design stage of study (can also be done for selection bias)

Confounding

the causal relationship b/w exposure and outcome is distored b/c of association of the exposure w/ other factors that influence the outcome (is there an extraneous factor blurring the effect)?

Statistical Significance

when p

Benefits of randomization

\- only way to control for known/unknown factors

\- randomization uses study design to control for potential confounding

\- investigators, participants, & analysts can be blinded, allows for some degree of subjectivity when assessing many outcomes (ex. patient answers more accurately b/c they don't know if they're in tx or control grp)

Why RCTs are best

\- population is controlled

\- confounding controlled through randomization

\- treatment is controlled (chosen by investigator)

\- natural history of disease is controlled (use of a control grp)

\- blinding

\- specific data is collected (control data quality)

Randomized Control Trials

RCTs (uses RR as measure of association); an experiment designed to assess relative efficacy of a test intervention in comparison to one/more alt interventions, in comparable grps of humans

\- participants randomized

\- can see what would have happened to exposed if they weren't exposed

\- control grp is as similar to intervention grp as possible, with regard to every feature besides exposure status- best in terms of hierarchy of study designs

Threats to randomization

unintended crossovers (ex. patient in tx grp don't do activity required of them, patient in control grp performs the activity instead)

Intent to Treat Analysis

maintains original randomization -\> potential confounders remain equally distributed across grps (best strategy)

Treatment Received Analysis

Cons of RCTs

\- artificial setting

\- limited scope of potential impact

\- adherence to protocol

\- ethical dilemmas (informed consent, data & safety monitoring, monitoring for side effects, protecting interest of participants, trials in special populations)

Screening Test

a test that checks for specific diseases in people who do not have the disease yet, types include:

\- mass: applied to whole population

\- multiple/multiphasic: more than one test

\- targeted: applied to grp of exposures (ex. occupational/environmental epidemiology)

\- case-finding/opportunistic: when patients consult w/ a provider for another purpose

Validity

extent to which a method measures what it's supposed to

\- face validity: extent to which a measure appears to measure what it's supposed to measure

\- construct validity: whether a particular measure relates as it should to other measures (ex. 2 different depression measures, if they correlate with each other = construct validity)

\- criterion validity: how well one measure predicts an outcome for another measure; a test has this type of validity if it is useful for predicting performance/beh in another situation)

Reliability

ability of method to measure consistently & reproduce the same results with repeated measurements

\- test-retest reliability: consistency of participant's responses over time

\- inter-rater reliability: consistency among raters for the same measure

\- internal consistency reliability: consistency among the items in a scale that are supposed to measure the same construct (having the questions be relevant/easy to understand for the reader)

Sensitivity

proportion of people WITH the disease who are correctly identified by the test

\
Sens = TP/(TP+FN)

Specificity

proportion of people WITHOUT the disease who are correctly identified by the test

\
Spec = TN/(TN+FP)

Positive Predictive Value

PPV; proportion of people who have a positive test result who are actually positive

\
PPV = TP/(TP+FP)

Negative Predictive Value

NPV; proportion of all people who test negative who are actually negative

\
NPV = TN/(TN+FN)

Social Determinants

aspects of society and the social environment that impact on health

\- general socioeconomic, cultural, & environmental conditions

\- social & community networks

\- individual lifestyle factors- age, sec, & constitutional factors

Health Disparities

differences in rates of disease occurrence across pop, grps, or places

\- socioeconomic

\- racial/ethnic

\- gender

\- geographic disparities (urban/rural)

\- etc.

Equal Playing Field

\- making sure everyone has equal access & the same opportunities as everyone else

\- not only individual responsibility needed, but responsibility of outside parties too (ex. companies mislabeling their food)

Approaches to Good Health

\- high quality education at a young age

\- strong & secure job positions (hard to take responsibility for health when you don’t know what’ll happen tomorrow)

\- stand the tide of growing inequality

Tenants of Causality

Bradford Hill criteria:

\- strength of association: large associations more likely to be causal & more difficult to think of alt explanations for them

\- consistency: evidence for causation stronger if replicated in diff pops by diff researchers at diff times using diff study designs

\- specificity: 1-1 relationship b/w exposures & outcomes? (ex. HIV specific to causing AIDS)

\- temporality: exposure must precede disease

\- biological gradient: strength of association increases as exposure level increases (exposure increase = outcome increase)

\- plausibility: is there an existing biological/social model to explain the association?

\- coherence: association consistent w/ generally known facts of the natural history & biology of disease

\- experiment: experiment that modifies exposure through prevention/tx/removal should result in less disease, may be infeasible/unethical

\- analogy: association b/w exposure & disease has characteristics/features that are similar to other associations generally regarded as causal