reallllll stats

1. timeliness

to implement effective control mesures

1.Intervention

effectivness of an intervention to improve health outcome. Randomized

2. Etiology

identify risk factors for disease. Non randomized

2. represenation

to provide acurate picture of disease trend

3 critera for confounder

"1. associated with exposure (without being a consequence of exposure)

2. associated with outcome (independant of exposure)

3. not on causal pathway"

3 principles for case control studys

"1. study base principle

2.deconfounding principle

3. comparable accuracy principle (reduces misclassification bias)"

3. Diagnosis

compare diagnositic tests (by sensitivity, specificity, predictive value and likelihood ratio). Randomized or non randomized. Can be interventional is test is the intervention

3. sensitivity

to allow identification of individuals with disease

4. Screening

compare screening tests to diagnose. Can be randomized but usually non randomized. can be interventional

4. specificty

to exlude people without the diease

5. Prognosis

clinical course of disease. What characteristics are associated with a better outcome. Non randomized

6. Prevelence

proportion of people with a disease. Non randomized

Absolute Mortality

Number of deaths in a population caused by a certain disease

Absolute Mortality

Number of deaths in a population caused by a certain disease

Analysis stage

multivariable regression analysis, stratification

Analysis stage

multivariable regression analysis, stratification

Analytic Bias

Beliefs influence data interpretation

Analytic Bias

Beliefs influence data interpretation

Ascertainment Bias

When the assessor knows exposure status

ascertainment bias

known exposure can lead to a biased outcome classificaiton

Attack Rate

The risk of a disease being developed during an outbreak

Attack Rate

The risk of a disease being developed during an outbreak

Attributable Risk

The incidence in the exposed group - the incidence in the unexposed group

Attributable Risk

"The incidence in the exposed group (risk) - the incidence in the unexposed group

* if we want % divide by incidence exposed"

Attrition Bias

Incomplete outcome data due to losses to follow up

Beneficial Effect of Work

Increase in access to healthcare due to employment

Beneficial Effect of Work

Increase in access to healthcare due to employment

Blinding difficulties

not always possibles when there are obvious diffrences between interventions (so blind data analysis)

BRFSS

"Behavioral Risk Factor Surveillance System

-telephone survey about health in USA"

case control CONS: inefficient (discard lots of potential controls), hard to find controls if you match on too many variables, cannot study the effect of matched variable, overmatching

inefficient (discard lots of potential controls), hard to find controls if you match on too many variables, cannot study the effect of matched variable, overmatching

case control matching

"-balance certain factors

-1 or more controls are selected to match case on certain characteristics"

case control matching PROS:

remove confounders

Case Fatality

The mortality rate divided by the incidence rate

case fatality rate equation

deaths/ cases x100 = a %

Case Reports

Reports which are individual to a single patient, Used to describe unusual symptoms/circumstances

Case Reports

Reports which are individual to a single patient, Used to describe unusual symptoms/circumstances

Case Series

A group of case reports each with a similar treatment, Do not test hypotheses

Case Series

A group of case reports each with a similar treatment, Do not test hypotheses

case-control interview bias:

ask cases and control questions about exposure history differently

Case-Control Studies

Defines a set group for cases which have the disease, and controls which do not have the disease, Can only be retrospective, Generates hypotheses

Case-Control Studies

Defines a set group for cases which have the disease, and controls which do not have the disease and compare exposure history retrospectively. Compare exposure history of those who got the disease and those who didn't

cases control and selection bias

"-exposure can influence detection of cases

-how we select cases for control can influence outcome: selecting cases and controls diffrently"

Causal Pathways

The relationship between variables, Confounders are not in the causal pathways of outcomes

Causal Pathways

The relationship between variables, Confounders are not in the causal pathways of outcomes

CCS CONS:

cannot estimate incidence rate (since we select ratio of cases and controls), can't estimate expusre rate, did exposure precede disease?

CCS PROS

: can study multiple risk factors, good for rare diseases, inexpensive since retrospective

Child mortality rate

number of deaths of children under 5/ 1000 live births

Clinical Equipoise

The question about what intervention is more valuable; only the expert medical community needs to be uncertain

Clinical Equipoise

"The question about what intervention is more valuable; only the expert medical community needs to be uncertain. A clinical trial is used to resolve

-The guiding ethical principal to justify an RCT is that there is general uncertainty in themedical community about the efficacy of a treatment."

clinical equipoise questions

"-how to define experts

-how to quanitfy uncertainty"

Cluster RCT

randomized groups rather then individuals, done for large scale public health

cluster RCT CONS

: experiences of individuals in same group are lielly similar leading to correlated results

cluster RCT PROS:

less time consuming, control for contamination across individuals (one person changing behaviour doesn't affect overall study)

cohort interview bias:

looking harder for outcomes in the exposed group

Cohort Studies

All participants are at risk of developing disease, Participants are chosen based on their exposure, Followed over a long period of time with large groups, Establishes temporal relationships, Done for rare diseases, Can be prospective or retrospective, Generate hypotheses

comparable accuracy principle

exposures should be measured with the same way in cases and controls

Comparator

control group

comparisions

"1. is intervention better then nothing

2. is intervention better then SOC

3. is intervention better then alternative tretament"

Concordant pairs

both case and control are either exposed or unexposed

Concordant pairs

both case and control are either exposed or unexposed

Confounders can

Wipe out associations, Create invalid associations, Reverse the association, Change the degree of association

Confounders can

Wipe out associations, Create invalid associations, Reverse the association, Change the degree of association

Confounding

Refers to the "mixing" of exposure effects with the effects of confounders, Distorts the estimated measure of association, A third factor is responsible for the change of two factors, The effect modifier is the effect of the third variable; confounding is when the third variable explains the effect

Confounding

Refers to the "mixing" of exposure effects with the effects of confounders, Distorts the estimated measure of association, A third factor is responsible for the change of two factors, The effect modifier is the effect of the third variable; confounding is when the third variable explains the effect

cons of fixing comfounding through mathcing

"-time-consuming and costly

-cannot evaluate the effect of the matched variable on disease'"

cons of fixing confounding through restriction

"-reduces # of eligible ppl

-residual confouding (catagory isn't narrow enough)

-can't generalize to other catagory of confounder"

cons of fixing confounding through strificaiton

"-computationally difficult if there are multiple confounders

-cannot be done when the confounder is a continuous variable that cannot be categorized"

CONS of RCS

"-historical effect

-accuracy/ completeness of data"

Coomabted Recal bias

through records/testing, blinding, biomarkers Case-control studies

CR CONS

not generalizable or symtematic

CR PROS

unusual and novel findings

Cross-Sectional Studies

Used to assess the relationship between the exposure and outcome, Identifies the exposure and disease at the same point in time, Quick and cheap to complete, Prevalence of both groups can be calculated

cross-sectional study

"observationsl study where we analyze data from population at certain point of time.

-doesn't mesure change or past history

-mesures prevelence of outcome"

Crossover

Every participant receives both treatments, Still an RCT, but groups switch halfway through study, Not applied to surgical intervention

CS CONS

if retrospective there can be lack of accurate data, selection bias, not generalizable, bad at establishing causation

CS PROS

cases reports but more people so help identify rare conditions and treatments

CSS CONS

can;t be sure if exposure preceded the disease, prevenlant cases may not be representative of all cases

CSS PROS

quick, cheap, exploratory and hypothesis generation, results may suggest the need to case control or cohort

deconfouding principle

expousre is not assocuated with other varriables also associated with the disease

Design stage

randomization, restriction, matching

Design stage

randomization, restriction, matching