Study Designs Flashcards

Question

What helps reduce variation in experimental studies? What is the main pro and con?

Answer 1

controlled settings - PRO: stronger evidence of causation due to it being prospective and fixation of confounders by design or random allocation - CON: limited range of hypothesis due to only being able to investigate one or few factors at a time and ethics (can't test force minors to smoke to investigate the effects of smoking on their health)

Answer 2

experimental studies with strictly controlled "in-house" conditions - PROS: highest level of evidence for causation - CONS: low-to-moderate level of relevance to the real world, expensive, time and labor intensive

Answer 3

experimental studies that account for real-world conditions - PROS: very high level of evidence for causation, highest level of relevance to the real world - CONS: very expensive, time and labor intensive

Answer 4

LAB: homogenous subjects (genetics, age, other confounding factors) are randomly sorted into treatment and control groups and disease status is observed CONTROLLED TRIAL: study subjects are taken from the population and are randomly sorted into treatment and control groups and disease status is observed

Answer 5

- randomized controlled trials ---> keeps it general - [randomized, blinded, two-arm, placebo] [field, clinical, community] trial trialists

Answer 6

study of an intervention (treatment, preventative factor) allocated randomly among animals naturally exposed to disease with the goal to test said intervention 1. comparison group (alternative treatment) - often a control placebo/sham group 2. random allocation - avoids selection bias 3. blinding/masking - avoids information and selection bias

Answer 7

1. Population with condition to begin with 2. Intervention group - treatment, exposure 3. Comparison group - placebo 4. Outcome

Answer 8

1. Hawthorne effect: participants in a trial tend to change many behaviors, regardless of intervention - taking better care of themselves, frequently going to the doctors 2. placebo effect: beneficial event that cannot be attributed to the properties of the placebo itself

Answer 9

owners evaluated patient's lameness as improving, when it was generally the same throughout the study

Answer 10

1. intention-to-treat: all enrolled patients and randomly allocated to treatment are included in the analysis ---> once randomized, always analyzed 2. per-protocol: includes only those patients who completed the treatment originally allocated ---> only those who perfectly followed protocol ---> biased toward better results

Answer 11

intention-to-treat avoids bias - the magnitude of effect depends on the adherence of the effect, depends on the adherence to assigned treatment strategy, making it represent the real world and pulls estimated toward the null hypothesis

Answer 12

results better represent treatment effect, but it is prone to bias with untestable assumptions

Answer 13

compare the risks of treatment and control groups

Answer 14

average number of patients who needed to be treated to prevent one additional bad outcome ---> number of patients that need to be treated for one to benefit compared with a control inverse of risk difference (1/RD)

Answer 15

RD = placebo risk - treatment risk for every _____ patients treated, _____ recovered because of treatment

Answer 16

NNT = 1/RD = 1/(placebo risk - treatment risk) if ____ patients out of ____ benefit from treatment (RD), the NNT for one patient to benefit is about ______

Answer 17

losses from the beginning of treatment to the follow-up period, commonly due to dropouts, death, moved away, and changes in treatment - should be minimal (<20%) - should be close to equal in each group: unbalanced = potential bias

Answer 18

if losses were related to the exposure AND the outcome = BIAS (untestable) - if patients who didn't respond to treatment died or dropped out before the end of the study, the treatment effect could look much better than it actually is

Answer 19

1. randomization - selection and confounding bias (uniformly allocate patients to both groups and evaluate similarities between them that may cause statistical differences, like weight, age, sex, breed) 2. blinding - selection and information bias randomize and blind...everything should be fine!

Answer 20

patients and clinicians are both unaware of treatment allocation to avoid information bias

Answer 21

research synthesis studies that tend to be subjective due to reviewer bias and weighing all studies equally, even if they do not have many participants quantitative summary conducted with subject-matter experts, possibly with pre-conceived notions

Answer 22

obtaining a broad overview of an area, NOT for decision-making

Answer 23

1. specific question 2. written protocol for review (transparent and repeatable) 3. thorough literature review 4. standard data extraction tool 5. does not require subject matter expertise

Answer 24

1. ensures all evidence considered 2. minimizes potential for bias in drawing conclusions 3. identifies gaps in knowledge - precursor of research 4. policy decision

Answer 25

qualitative - weighs studies based on assessment of quality of methods (more evidence = more priority)

Answer 26

becomes a meta-analysis that extracts ORs, RDs, and other measured values and weighs studies based on result precision

Answer 27

1. question formulation 2. literature search 3. relevance screening 4. quality assessment 5. data extraction - QUALITATIVE = systematic review - QUANTITATIVE = meta-analysis with pooled estimates and accounts for heterogeneity

Answer 28

extraction of important data from all relevant primary research studies or studies of acceptable quality to summarize and/or synthesize results 1. PAPER: general study information, like year, author, country, language, funding, and study design 2. POPULATION: sampling frame, sub-populations, population attributes 3. INTERVENTION: treatment, challenge, dosage, administration 4. OUTCOME: types of test, test protocol 5. RESULTS: type of data, effects estimates, statistics

Answer 29

DICHOTOMOUS - OR, RR, RD/ARR CONTINUOUS - mean difference, standardized mean difference

Answer 30

by design and quality identifies gaps in knowledge and common methodological issues if heterogeneity is too large ---> too much variation in outcomes, populations, etc.

Answer 31

statistical methodology used to combine data from two or more studies that have addressed the same question 1. provides an overall estimate of an association 2. explores sources of variation in the observed effect across studies

Answer 32

1. ensures all data considered has appropriate weight 2. gains statistical power, allowing for small, non-significant results to contribute 3. documents variability in study results 4. investigates reasons for variability 5. quantifies impacts of potential bias

Answer 33

QUANTITATIVE OUTCOMES ---> OR, RR, RD, mean difference, standardized mean difference (controlled trials are best)

Answer 34

weighted average treatment effects by their precision (more precise = more weight)

Answer 35

can use a cumulative plot (1+2, 1+2+3, 1+2+3+4....) to see when sufficient evidence overtime was available to make a decision

Answer 36

publication bias - failure to publish results that are not statistically significant or unfavorable to study sponsor funnel plots - X-axis = precision (standard error); Y-axis = treatment effect

Answer 37

1. RANDOM CONTROLLED TRIALS*** - more standardized with clearly defined intervention/comparison groups 2. observational studies - possible, but difficult due to issues with heterogeneity and prone to bias 3. diagnostic test evaluation - estimates DSe and DSp

Study Designs Flashcards

(61 cards)