week 7: infectious disease research - study designs & measures of association

what is a study design?

• plan for selecting participants and deciding how + when to collect info to answer a research question

• exposure: factor being studied

• outcome: health event or condition being measured

exposure vs. outcome examples

• exposure may be harmful, protective, or unrelated to outcome

• exposure ex: broad-spectrum antibiotic use, influenza vaccination, injection drug use

• outcome: c. diff infection, influenza infection, hep c infection

descriptive vs. analytic studies

descriptive studies describe patterns, analytic studies examine associations

descriptive studies: main question, purpose, example

• mq: how common is condition? who is affected, where, when?

• p: describe disease patterns by person, place, and time

• ex: # of influenza cases by age grps

analytic studies: main question, purpose, example, types

• mq: is an exposure associated w/ an outcome

• p: examine associations

• ex: compare influenza occurrence in vaccinated + unvaccinated grps

• two types: observational, experimental

observational (type of analytic study)

• researchers measure exposures + outcomes without assigning exposure

• cross-sectional, cohort, and case-control studies

• ex: clinicians choose antibiotic a or b during routine care. researchers compare recovery btwn grps

experimental (type of analytic study)

• researchers assign intervention + measure outcomes

  • ex: randomized controlled trials (RCTs)

• ex: researchers randomly assign antibiotic a or b + compare recovery btwn grps

cross-sectional study

• researchers sample ppl from defined population

• researchers assess exposure + current outcome during same assessment period

  • ex: during one-day survey, UM nursing students report their usual hand hygien frequency (exposure) + current respiratory symptoms (outcome)

• key limitation: researchers usually cannot determine which came first - exposure or outcome

• overall: snapshot of exposure + current outcome

common measures of cross-sectional studies

• point prevalence

• prevalence ratio and prevalence difference

point prevalence (cross-sectional study measure)

• proportion w/ outcome at specified point in time

• ex: percentage of surveyed UM nursing students who have respiratory symptoms on survey day

prevalence ratio + prevalence difference

• compare prevalence btwn exposure grps

• ex: compare respiratory symptom prevalence btwn students reporting less frequent + more frequent hand hygiene

cohort study

• researchers select + group participants by exposure w/o assigning exposure

• researchers compare risk or rate of developing outcomes over defined period

  • ex: researchers compare risk of developing influenza (outcome) over 12-week period btwn nurses who were vaccinted before that period + those who were not (exposure)

• key limitation: grps may differ in age, underlying conditions, or other factors that may affect results

• overall: cohort studies compare subsequent outcomes btwn exposure grps

two types of approaches for cohort studies:

• prospective studies follow participants into future

  • occurs after study begins

• retrospective reconstruct events from past

  • already occurred before study begins

• both examine outcomes after exposure

example of prospective (approach to cohort studies)

• researchers identify vaccinated + unvaccinated nurses now

• then follows them for next 12 weeks to compare risk of developing influenza

example of retrospective (approach to cohort studies)

• researchers use records to identify vaccinated + unvaccinated nurses at start of completed 12 week period

• then compare their risk of dev influenza during that period

common measures of cohort studies

• incidence risk

• incidence rate

• calculate each measure within vaccination grp, then compare grps using ratio or difference

incidence risk: denominator + comparison measure (common measure of cohort studies)

• d: people initially at risk

• cm: risk ratio and risk diff

• % of nurses who dev influenza during 12 week period

incidence rate: denominator + comparison measure (common measure of cohort studies)

• d: total person-time at risk

• cm: rate ratio and rate diff

• new influenza cases per 1000 person-weeks at risk

case-control study

• researchers select participants by outcome: cases have outcome + ctrl do not when selected

• ctrls come from same population as cases

• researchers compare prior exposure btwn cases + ctrls

  • ex: researchers select nurses who dev influenza during a 12 week period (cases) + sample of nurses from same hospital who did not ctrls. they compare influenza vaccination before that period

• key limitation: poor ctrl selection + inaccurate exposure info can distort the result

case-control studies: common measures

• exposure odds

• odds ratio (OR)

• why not risk or rate? researchers select participants by outcome; typical caes-ctrl sample counts do not directly provide population risk, incidence rate, or prevalence

exposure odds (common measure of case-ctrl studies)

• # prev exposed ÷ by number of unexposed, calculated separately for cases + ctrls

• ex: odds of having received influenza vaccination (exposure)

odds ratio (OR) (common measure of case-ctrl studies)

• exposure odds among cases ÷ exposure odds among ctrls

• ex: odds of prior vaccination (exposure) among nurses who dev influenza (cases) relative to odds among nurses who did not (controls)

cohort study: starting point, assessment, ex, useful for, common measures

• sp: defined group, classified by exposure

• a: whether exposed + unexposed participants develop outcome

• ex: compare new UTIs (outcome) among patients w/ and w/o catheter (exposure) at baseline

• uf: measuring incidence; studying rare exposures or multiple outcomes

• cm: risk or rate ratios + differences

case-control study: starting point, assessment, ex, useful for, common measures

• sp: cases + ctrls, selected by outcome

• a: whether cases + ctrls had exposure before outcome

• ex: compare prior catheter use (exposure) among patients w/ UTI (cases) and w/o UTI (ctrls)

• uf: studying rare outcomes or several possible prior exposures

• cm: odds ratio

randomized controlled trial (RCT)

• researchers randomly assign participants by chance (randomization) to an intervention grp or comparison grp

• randomization helps balance other factors that may affect outcome

• researchers follows both grps + compare outcomes over defined period

  • ex: researchers randomly assign nurses to receive new influenza vaccine (intervention) or standard influenza vaccine (comparison). they compare new influenza cases over 12 weeks (outcome)

• key limitation: missing outcome data (ex: loss to follow up) can bias results

• assign intervention by chance, then compare outcomes!

common measures of RCT: frequency measure; denominator + comparison measure

• incidence risk: ppl initially at risk

  • cm: risk ratio + risk diff

• incidence rate: total person-time at risk

  • cm: rate ratio + rate diff

• compare incidence risk and incidence rate within each assigned grp (ex: new vaccine vs. standard vaccine), then compare grps using ratio of diff

RCT vs. cohort

• both compare outcomes over time

• main diff is whetehr researchers randomly assign intervention

RCT: study type, researcher’s role, assessment, common measures

• st: experimental

• rr: randomly assign intervention

• a: outcomes after assigned intervention

• cm: incidence risk + rate, ratios + diff

cohort studies: study type, researcher’s role, assessment, common measures

• st: observational

• rr: record exposure w/o intervention

• a: outcomes after observed exposure

• cm: incidence risk + rate; ratios + diff

picture of ways to compare grps

picture of study designs + common measures

picture of comparing prevalence - ratio and diff:

calculating prevalence

picture comparing risk - ratio and diff:

picture of calculating incidence risk:

picture comparing incidence rates - ratio & diff:

picture of calculating incidence rates:

picture of odds ratio ex + interpretation

picture of calculating odds ratio

a clear interpretation of a finding identifies:

• measures of association: risk, rate, prevalence, or odds

• exposure

• grps: who is being compared?

• outcome: what condition was measures?

• time: when or over what period?

association: meaning + example

• observed statistical relationship btwn variables

• vaccinated nurses had lower influenza risk than unvaccinated nurses

correlation: meaning + example

• type of association describing how 2 numerical variables vary together

• hospitals with higher staff vaccination coverage had lwr staff influenza rates

causation: meaning + example

• changing an exposure changes the likelihood of outcome

• causal claim: vaccination reduces influenza risk

association, correlation, and causation as a whole:

• association + correlation describes relationships

• neither alone establishes causation

association doesn’t prove causation; what are the three factors to consider?

• confounding

• bias

• chance

confounding: meaning + example

• another factor related to both exposure and outcome distorts their association

• vaccinated nurses also practiced other infection-prevention measures more consistently, which could contribute to their lower risk

bias: meaning + example

• systemic error in participant selection or measurement

• unvaccinated nurses were tested more often, so more infections were detected

chance: meaning + example

• random variation in findings

• more infections occurred in one group by chance, especially in small sample

research ethics, what to consider?

• ethical review

• informed choice

• voluntary participation

• privacy and confidentiality

• safety

• fair selection

ethical review

obtain institutional review board (IRB) review when required

informed choice

• explain study’s purpose, procedures, risks, and potential benefits

• obtain informed consent when required; for minors, obtain parental/guardian permission and child assent as required

voluntary participation

respect right to decline or withdraw without penalty

privacy and confidentiality

protect participants and their personal information

safety

minimize risk, monitor safety as appropriate, and use an ethically appropriate comparison group

fair selection

select participants fairly + provide additional protections when needed