Evidence-Based Medicine
Terms
undefined, object
copy deck
- Why is EBM needed?
-
-lots of junk science in the market place for medical info
-need to make the best possible clinical decisions often in the absence of a single, definitive best course of action
-problems have led the medical community to develop and emphasize: small area variations analysis, outcomes research, clinical practice guidelines - Define EBM
- conscientious explicit, and judicious use of current best evidence in making decisions about the care of individual patients
- Arguments for EBM
-
-de-emphasizes the institution, unsystematic clinical evidence, and pathophysiological rationale as sufficient grounds for decision making
-many therapies used by physicians not backed by solid evidence
-to much potential for fraud in the medical marketplace - Arguments against EBM
-
-proponents are antagonistic
-no evidence that EBM works
-individual patient should not be looked at as aggregate data
-may cause conditional compassion - define paradigm
- a combination of a set of methods and a phylisophical approach
- what is the guiding paradigm
- best available science at a point in time
- what is clinical epidemiology
- study of the distribution and determinants of health and disease frequency in human populations
- what is the basic tenet of clinical epidemiology
- disease does not happen at random and is quantifiable
- what is clinical epidemiologies aims
- to profile typical and atypical patients, determine history and patterns, surveillance, search for causes, and test new treatments
- list criteria for causality
-
-establish a link between drug X and effect Y
-strength of the association
-consistency of the observed evidence
-specificity of the relationship
-temporality of the relationship
-dose-response relationship
-biological plausibility
-coherence of the evidence
-experimental conformation
-reasoning by analogy - characteristics of descriptive studies
-
-goal is to record events, observations, and/or activity
-does not provide info about causes or clinical efficacy
-does provide initial picture of some clinical phenomenon, and suggests areas of more in-depth research - examples of descriptive studies
-
-case reports
-case series (cluster reports)
-clinical series
-population (prevalence) studies/surveys
-course/program description - characteristics of explanatory studies
- -goal is to compare and explain differences between things, and to shed light on etiology or prognosis of disease
- what are the names of the two approaches to explanatory studies
-
-experimental
-observational - experimental study characteristics
-
-known as control trials, clinical trials, RCT, or intervention studies
-primary feature is that investigator controls intervention, and methods in great detail
-most powerful in terms of causal inferences - observational study characteristics
-
-investigator is passive observer
-study natural course of health events, gather data, classify and sort them
-through making comparisons across groups, investigators try to provide insight as to causes and effects
-less than perfect approach
-never as good as a RCT - differences between cases and controls
-
-a case can be a subject who has the outcome of interest when doing a case-control study
-sometimes a group of people (cohort) that share a characteristic are watched for development of outcomes (cohort/follow-up study)
-key is the directionality of the study
- - differences between retrospective and prospective
-
-retro studies begin and end in the present, but consist of looking back in time for data collection.
-prospective studies begin in the present and measure outcomes that lie in the future
-use retro and pro to describe the time frame for collection of data only - define validity
- validity refers to the degree to which a measurement represents the true value
- define reliability
- reliability relates to the reproducibility of measurements
- define unsystematic variability
-
-errors that ocurr unpredictably and are minor
-they do decrease reliability and validity of a study
-random variations tend to even out (regression to the mean) - types of unsystematic variability
-
-subject variation
-observer variations - define systematic error
-
-more serious than unsystematic variability
-predictable variation known as bias - types of systematic error
-
-investigator/interviewer bias
-subject bias - define investigator/interviewer bias
- those doing the study are looking for an outcome or may have different approaches to asking subjects questions
- define subject bias
-
-pre-existing beliefs
-social desirability bias
-recall bias
-hawthorne effect
-test-retest bias
-rebound bias - list ways to control measurement error
-
-best way is to keep subjects and investigators blind
-establish clear standards for measuement
-train observers
-use multiple observers or data sources, then assess inter-rater reliability (IRR) or to what extent the observers agree (between 0-1)(kappa=correlation coefficient which accounts for chance agreement) - define a target population
- population of people that a study tries to generalize a hypothesis about (all pediatric patients)
- define an accessible population
- portion of target population from which information is attainable from (pediatric patients admitted to hospitals in US between 1995 and 2005)
- define sample population
- portion of accessible population that is used for study (200 pediatric patients from 5 different hospitals throughout US)
- what are 3 basic sampling methods
-
-systematic
-random
-convenience - define systematic sampling
- picking every nth person
- define random sampling
- base picks from a computerized random number chart
- define convenience sampling
- volunteers or recruits
- define fundamental principles of case series
-
-share experiences, new observations
-provides the intial steps toward sophisticated research
-generates testable research hypothesis
-initial (crude) quantification of incidence rates
-prototypical descriptive methods - define fundamental principles of case-control studies
-
-begin with an outcome of interest and look back in time to compare similarities between subjects
-normally done all retrospectively - application of case reports and series
-
-exceptions to the rule
-natural history of disease
-health services planning
-therapy: OK for feasibility, or potential effectiveness. NOT OK for efficacy or effectiveness statement UNLESS there is a dramatic (slam-bang) effect - problems with case reports and series
-
-"n of 1" is anecdotal info
-no controls of any kind
-small possibility of sampling variation explaining observed outcome (chance)
-investigator may have been looking for the cause (bias)
-alternative explanations (confounders) - Advantages of case-control studies
-
-relatively easy and quick
-inexpensive
-good for studying rare diseases or outcomes
-good for studying multiple exposures
-analysis of results is straight forward - Disadvantages of case-control studies
-
-sampling method employed
-normally only supplies suggestive, and not definitive causal information
-diagnostic bias
-reporting bias
-case definition (needs to be clear and specific)
-any source of control selection can be problematic and often need to use multiple control groups
-possibility to over match
-ascertaining exposure subject to bias (recall and social desiribility)
-researcher bias as information bias
-data sources may not be valid and/or reliable - what is the structure of a cross-sectional study design
-
-investigator makes all measurements at a single point in time
-well suited to the goals of describing variables, their distribution patterns, and correlates
-useful for examination of association, but is difficult to choose/specify which variable is the cause and which is the effect - what are predictor variables
-
constitutional factors such as age, gender, race
thse cannot be changed or altered by other variables - what are outcome variables
- variable that you believe might be influenced or modified by some treatment or exposure.
- what are the stengths of X-S studies
-
-fast and inexpensive
-no problem with loss to follow up
-gives estimate of the prevalence of one or more risk factor
-convenient for initial investigation of networks of causal links
-good first step to cohort study - what are some disadvantages to X-S studies
-
-sampling
-questioning biases
-response biases-non-responders may differ in some way that is related to the outcome
-time order problems
-not good for studying rare diseases; if small sample of population is used - what is the structure of a cohort study design
-
-involves classifying subjects by exposure status
-follow them in groups over time
-two goals are to describe incidence of certain outcomes over time, and to analyze associations between risk factors and those outcomes
-may be done retro or prospectively - what are the 6 steps in designing a case-control study
-
step 1: specify research question
step 2: specify (sample) cases
step 3: draw sample of controls
step 4: measure predictor variables
step 5: compare cases and controls (prior exposure patterns)
step 6: draw appropriate conclusions - what are the 6 steps in designing a X-S study
-
step 1: specify research question
step 2: specify target and accessible (research) population
step 3: draw samples
step 4: measure predictor and outcome variables
step 5: analysis: prevalence (absolute or relative)
step 6: draw appropriate conclusions - what are the 6 steps in designing a cohort study design
-
step 1: specify research question
step 2: assemble suitable cohort
step 3: measure predictor variables
step 4: follow cohort and measure outcomes
step 5: analysis: incidence, relative risks (within, across cohort(s))
step 6: draw appropriate conclusions - strengths of cohort study design
-
-good for quantifying incidence and investigating potential causes of a condition
-good for studying rare exposures and/or multiple outcomes
-times-sequence is clearly established
-if prospective: maximum control over measurement, historical data problems (recall bias, time-sequence issues, missing data) minimized - disadvantages of cohort studies
-
-selection of subjects may not represent your population
-LTFU
-subjects may change status over time
-surveillance bias
-often require large sample sizes (especially if studying rare or fatal outcomes)
-association found in cohorts may be due to confounding
-prospective ones are costly and time consuming
-retrospectively done ones contain historical problems, and time-sequencing questions - how do you minimize LTFU
-
-make repeated contacts with subjects
-trying to track down lost subjects
-reporting LTFU rate and assessing impact on results - when do you use a cohort study design
-
-best design for accurately describing incidence and natural history of condition
-often only way to establish time sequence
-only way to study some fatal diseases (survivor bias enters into X-S study)
-multiple, even previously unseen or unknown outcomes can be studied (if pro design) - probability sampling types (4)
-
-simple random-table of random numbers
-systematic- every nth person
-stratified-devide list into groups (by race, gender, age) then draw random samples from each stratum
-cluster- random sampling of population groupings, then within groupings (random sampling of counties, then RPH's within the counties - non-probability sampling types (3)
-
-consecutive sampling- all eligible subjects over a period of time
-convenience sampling- most easily available members of research population
-judgemental sampling- hand-picking from research population those subjects deemed most appropriate for the study - which sampling method is best
-
-depends on research question, goals, resources
-consecutive sampling OK in clinical research (if not feasible-draw a random sample of subjects from a consecutive series, or use another method
-when possible preferable to use probability sampling methods