Epidemiology Exam 2

cause

characteristic about disease, without it the disease doesn’t occur

association

linkage between variables

exposure

contact with the factors that are linked with a bad outcome

necessary cause

when a factor is required for the disease to occur

(deterministic causality)

sufficient cause

when a factor that is sufficient by itself to guaranteed to produce the disease

(deterministic causality)

necessary and sufficient deterministic causality

when the factor and disease are present, the factor is the only thing needed to cause the disease

very uncommon

sufficient and not necessary deterministic causality

When the factor may be present when the disease occurs, the factor can be one of the causes

necessary and not sufficient deterministic causality

factor must be present for disease to occur, but factor by itself does not guarantee the disease occurrence

not necessary and not sufficient deterministic causality

factor may/may not be present for disease to occur

most common

causal pie model

Pie diagram for a disease arranged by causes (deterministic causality)

• Necessary cause is always present with non-necessary variables (A in pic)

• Component cause: other causes accompanying the necessary cause

• sufficient cause complex: whole pie, all the causes together create a sufficient cause

probability causality

Type of causality where a factor increases the probability of a disease

deterministic causality

Type of causality where a factor directly causes a disease

stochastic model

model for probability causality

associations

when variables are connected positively and negatively

• (+) association → when variables go the same way

• (-) association → when variables have opposite directions

measurable by statistics, tests determine causation

directed acyclic graph (DAG)

Graphing approach for causal relationships

• arc: single-headed arrow that connects variables - direct links

• node: points/variables (letters in pic)

• path: unbroken route along nodes

  • direct: single-headed arrow sequence (A→B→E)

  • backdoor: non-causal pathway (B,D,C all connect to → E, all factors)

clinical trial

Experimental study design that focuses on the individual

• measures incidence - subjects do not have outcome

• used to test therapies

• 2 types: prophylactic and therapeutic trials

prophylactic trial

type of clinical trial

focus is on preventing disease by testing preventative treatment

measures incidence

therapeutic trial

type of clinical trial

focus on testing treatments and outcome

measures prevalence

single blind trial

characteristic of clinical trial

when the participant does not know what control they have

double blind trial

characteristic of clinical trial

when the researchers and participant do not know what control the participants are assigned to

Phase 1 clinical trial

Beginning phase of a clinical trial

testing new volunteers for adverse physiological side effects from treatment

<100 participants

Phase 2 clinical trial

clinical trial phase

a sample from population is taken and tested

tests physiological responses and clinical reactions

100-200 participants

Phase 3 clinical trial

phase of clinical trial

large scale testing in target population, compares incidence against general population

tests efficacy - how the treatment performs under tight circumstances

Phase 4 clinical trial

phase of clinical trial

post-marketing research into a drug

monitors side effects/effectiveness not under controlled circumstances

clinical trial advantages

Control:

• exposures

• time/frequency

• observation period

disadvantages of clinical trials

• artificial experiment

• might have ethical boundaries

• tight adherence protocols

community trials

quasi-experimental trial at the community level

focuses on health/behavioral changes in a target population

measures policy/program effectiveness through incidence

community trial disadvantage

• hard to control: outcome monitoring, intervention delivery

• affected by population dynamics

• hard to randomly assign study units, affects comparability

observational studies

study design that compares disease frequencies

cohort study

Study design where participants share a common exposure/characteristic - individual level

sample from population is followed to see if disease develops - subjects lack disease

measures incidence and changes in markers for disease

• population-based

• exposure based

population-based cohort study

type of cohort study

participants are broadly chosen from a target population

exposures are unknown until baseline measurement are taken

exposure-based cohort study

type of cohort study

participants are specifically chosen by exposure status

non exposed group chosen to align with exposed group by characteristics (not exposure)

prospective studies

studies that take present subjects and follow them for possibility of disease in the future

measures incidence

retrospective study

studies that get a baseline from the past and compare it to present data

measures prevalence

cohort study advantages

• time sequencing

• studies outcomes and rare exposures

cohort study disadvantaging

• long time periods

• costly to follow participants

• participants can be lost to follow up

• hard to follow rare diseases

• selection bias

relative effect measures

divides frequencies to show strength of relationships

• risk ratio

• incidence rate ratio

risk ratio

used for cohort and CT studies, type of relative effect measure

tells if the risk of disease if different in exposed vs. non-exposed

“The exposed have __ times the risk of disease compared to non-exposed".”

absolute effect measures

risk differences to measure impact

• risk difference

• population risk difference

• etiological fraction

incidence rate ratio

used for cohort and CT studies, relative effect measure

compares incidence between exposed and non exposed groups

requires person-time

risk difference

used for cohort and CT studies, absolute effect measure

shows excess number of cases in exposed group

etiologic fraction

used in cohort and CT studies, absolute effect measure

shows how much exposure is a cause of disease in exposed group

population risk difference

identifies number of cases that would be eliminated if exposure was eliminated

I_p = overall incidence in population

“(exposure) results in excess number of cases of (disease) in total population

case-control study design

observational study design that compares people who have a disease (case) to those who do not (control) - individual level

retrospectively uses data to make observations in present - measures prevalence

population-based, control group should match case group closely

case-control advantages

• quick

• cost -effective

• good to measure rare diseases

case-control limits

• unclear temporal relationship with exposure

• requires recalling because its retrospective

  • uses an indirect estimate of risk

odds ratio

relative effect measure for case-control and cross sectional studies

measure association between exposure and disease

cross sectional

study design on the individual

uses survey to estimate prevalence, exposure and disease measures, to describe the magnitude/distribution of disease, and show risk factors over time (usually not done in the same person over time)

used to compare trends and plan health services

cross sectional advantages

• low cost

• easy to generalize

cross sectional limits

• does not show incidence

• hard to determine temporality from exposure (collected after exposure)

• self-reporting

ecological study design

study design for the group level

uses secondary data → census, government data, and investigations

used to test/develop a hypothesis and determine causation

ecologic fallacy

observation at the group level may not be true at the individual level, so it is hard to make inferences for the individual from ecologic studies

external validity

the ability to generalize a study to a different target population or a general population

validity outside the study

internal validity

validity inside a study

when there are correct measurements of the exposure, outcome, and associations

“truth”

desired outcome of the study → results are true

hope to show probabilistic causality - shows causation

“chance”

when a study’s results show causation but it happened by random chance

due to errors - often from statistical testing errors

type 1 error

type of error that shows a result

does not align with reality - not true

type 2 error

type of error to that does not show the correct truth that aligns with reality

random error

error around a value of a parameter

• sampling error: wrong people in sample

• poor precisions: data not measured precisely

• validity in measurements: variability around results due to conflicting results

reducing random error

• increasing sample size

• raising the number of measurements

systematic error

bias - error in study design, data collection, data interpretation, and reporting

• selection bias

• information bias

• observer bias

• confounding

selection bias

systematic error type

exposure/disease is different in participants than the ideal sample

observer bias

systematic error type

when participants perform differently because they know they’re being watched by observers

information bias

systematic error type

when the measurement/information about the disease/exposure is reported inaccurately

reducing systematic error

• using objective definitions in the study

• having higher participation rates

• using memory aids - not based on one persons recall

• blinding

• standardized rules for research team

• standardized procedure for data collection/analysis

• clear/understandable questions

confounder

a variable that that may distort associations or provide an alternate explanation for association btw disease and exposure

confounder criteria

• must be a risk factor for the disease

• must be associate with exposure

• cannot be an intermediate step in the causal pathway (exposure cannot cause the confounder)

controlling the confounding in study design

1. randomization: equal distributions of confounders in each group

2. restriction: restricting participants to align with a known confounder

3. matching: matching subjects in groups according to known confounders

randomization (confounding) advantage/limits

advantages: convenient and addresses unknown

limit: need control over exposures to assign groups

• need to have a high sample size

restriction (confounding) advantage/limits

advantage: have complete control over the known

limit: do not have control of unknown

matching (confounding) advantage/limits

advantage: requires fewer subjects

limit: costly because extensive searching

controlling confounding in analysis

• stratification: performing analysis in strias (levels)

• multivariate: advances statistical analysis