Delfini Evidence-based Practice & Evidence-based Medicine Glossary
|
About PMID Numbers: We frequently utilize a PMID number in place of a citation. Where PMID numbers are available, enter that number into the PubMed search box to retrieve that citation and listing. A B C D E F G H I J K L M N O P Q R S T U V W X Y Z Absolute
risk reduction (ARR) – A measure of outcomes which reflects
the actual percent difference in study outcomes between groups.
See a visual description here. Accuracy – The ability to correctly identify that which one intends to identify. Active agent — Active drug (i.e., not the placebo). Adjustment - Statistical means for minimizing differences in study and comparison groups (e.g., standardization and regression analysis).Adverse event (ADE) — see Safety Aim, study — see Objectives Allocation concealment — see Concealment of allocation Alpha error — see Error A priori – In advance. For example, issues such as study questions, outcome measures, subgroups for analysis and p-values should be determined in advance of the actual study (i.e., determined a priori). Ascertainment bias - See bias Assessment bias - See bias Association – Statistical relationship between two or more events, characteristics or other variables. Attrition bias - See bias Baseline characteristics — Key characteristics of interest at the start of a study which are potential prognostic factors. These usually include demographic information such as age, gender, race and other characteristics related to the area under study such as disease severity, baseline measurements for clinical items of interest, etc. Collecting and reporting on baseline characteristics helps to establish how similar the study groups are, describes the actual population studied and helps to measure improvements during the course of the study. Benchmark – A standard or point of reference used in measuring and/or judging something such as quality or value. Beta error - see Error Bias – Study processes which may result in, or lead to, conclusions differing from truth in a systematic way (meaning not due to chance). Bias may occur at various study stages such as assigning subjects to study or comparison groups, intervention or exposure, performance, provision of services or conduct of processes affecting subjects, data collection, subject follow-up, measurement, analysis, interpretation and/or publication of data. Bias frequently occurs as a result of some inequality between the study and the comparison group. The inequality may be important because it may affect the study outcome.
Blinding –
An important study procedure to keep secret certain study procedures
such as which is an active drug and which is a placebo. Bias can
result when study subjects and those involved in study procedures
know treatment assignment of individual subjects. Blinding is
a method to help avoid the introduction of this kind of bias.
Double-blinding refers to when neither patient nor persons performing
the intervention or exposure — or working with the subjects'
data — know if the patient is in the study group or the
comparison group. Case — Research subject who is a member of the group possessing the characteristic under study such as an exposure, risk factor, outcome or receiving the intervention in an experimental study. Case-control study – An observational study in which subjects with outcomes of interest (cases) are compared to those without the outcome (controls). Histories are examined to attempt to determine exposures. Case series – A group of patients receives an intervention and outcomes are assessed. There is no comparison group. Case series is almost never “evidence.” All-or-none results are required, which rarely happens. Case study — An observational of one person. Also referred to as a case report. A case study is not evidence. Chance — Research
results happening by "accident," meaning that they did
not happen because of truth or because of bias. Likelihood is
assessed by using appropriate statistical tests. To critically
appraise, examine p-values and confidence intervals with associated
point-estimates (none of which can address bias, which is a systematic
error - meaning an error not due to chance). Class effect - Class effect refers to a concept that certain drugs share sufficient similarities that they can be thought of as a class. It is a determination that a set of agents with similar chemical structures, mechanisms of action and pharmacological effects has similar therapeutic and adverse effects. There are no universally accepted criteria for defining class effect. (For all intents and purposes, class effect treats the agents as if they are the same clinically, but some other factors, such as cost, may vary.) Clinical practice guidelines – Clinical practice guidelines are systematically developed statements to assist practitioners and patients in choosing appropriate healthcare for specific conditions. They should be based on valid and useful evidence, and they should be critically appraised before adoption. Clinical significance — Research should benefit patients in ways that are clinically significant (i.e., that matter to them). Clinically significant areas are morbidity, mortality, symptom relief, functioning and health-related quality of life. Results from valid research in these areas should be large enough and useful enough to provide meaningful clinical benefit. Clinical trial — A research methodology which evaluates a drug or other intervention to assess safety and efficacy by evaluating its effect on a group of subjects. The subjects are divided into two or more groups with the study group receiving the drug or intervention and the comparison group or groups receiving a comparison drug, placebo or intervention. Cohort study – An observational study in which exposure occurs in the study group and not in the comparison group. Exposure is not introduced by the investigator, but happens through other means such as natural occurrence or individual choice. Groups are followed over time and outcomes are compared. Cohort studies can be done prospectively where a group of interest and a comparison group are followed forward in time. Cohort studies can be done retrospectively when the group of interest and comparison group are identified from the past, based on an exposure or characteristic, and then information about them is reviewed from that time forward to assess development of outcomes. Comparison groups — Subjects in groups being compared to the group of interest (e.g., intervention group would be “group of interest” and placebo group would be the “comparison group”). Composite endpoints — Refers to an endpoint in which single endpoints are grouped together to form one endpoint such as “cardiovascular events.” Concealment of allocation
– The process used in a randomized controlled trial to hide
the assignment to a study group. The purpose is to ensure that
no one can influence or control which study subject gets assigned
to which group (e.g., assignment made through a call-center, etc.) Confidence interval
(CI) – For a valid study, the range in which the likelihood
of a true value is expected to be, within a given degree of certainty,
usually 95 percent. Confounding – A special type of bias in which another factor associated with the study variable of interest may have "traveled" with that variable and is the true reason for the study conclusion instead of the variable under study. There are known confounders and unknown confounders. Randomization is a method which attempts to minimize confounding by randomly allocating subjects to their groups in hopes that any potential confounders are equally distributed between the groups. Contamination – Any member of a comparison group receiving the intervention under study or being exposed to the variable of interest. Control - Research subject who is a member of the comparison group. Controlled clinical
trial – An experiment in which investigators include a control
group which is used as a comparison group. The comparison group
may receive another treatment, a placebo or “usual care”.
Subjects are assigned to the study or to the control group. Intervention
or exposure is then performed. Patients are followed to determine
outcomes such as improvements or harms. Covariate - A variable that might affect, or possibly predict, an outcome. Critical appraisal – A scientific evaluation of evidence (e.g., research data) to appraise validity (closeness to truth) and usefulness (e.g., generalizability to one's own patients or circumstances, meaningful benefit, etc). Cross-over design - A study design in which patients start out having one intervention and then are “crossed over” to a different intervention. Crossover trial –
A comparison of two or more interventions in which the participants
receive one treatment and then are “crossed over”
to another intervention. Care must be taken with this study design
because effects of the first treatment could contaminate the effects
of the subsequent intervention. Cross-sectional study – An observational study in which a sample from a larger population is examined, at one single point in time – like a snapshot - for prevalence of an exposure or characteristic along with an outcome of interest. Cross-sectional studies are also frequently used to evaluate diagnostic testing. Database research — Statistical relationships between two or more variables are assessed from databases. Frequently database research is used to report outcome measures. Database research, however, is a type of observation and, like all observational research, is highly prone to selection bias, observation bias and confounding — and the likelihood of finding "statistically significant" relationships merely due to chance is high to certain, depending upon the way the analysis is conducted. Definite cause and effect conclusions should not be drawn from database research. Data gathering validity – Data gathering validity refers to the methods used to obtain numerators and denominators during the data gathering phase of performance measurement. Delta — See Equivalence trials and Non-inferiority trials Denominator –
The lower portion of a fraction used to calculate a rate or ratio.
In a rate, the denominator is usually the population (or population
experience, as in person-years, etc.) at risk.
Dependent variable
– In a statistical analysis, the outcome variables under
study. Detection bias - See bias Diagnostic testing — see Measures of test function Disease spectrum — A range of symptoms, signs, lab results and results of other diagnostic tests, and rate of disease progression, response to therapy and disease severity. Double-blind – see Blinding Double-dummy treatment design — A treatment design involving two placebos. Control subjects receive one or the other of two placebos, but not both. Study subjects receive the intervention drug. For example, if you were studying an antibiotic being given at two different doses, you might need two placebos because of the weight and size difference of the doses. Drug utilization review (DUR): A system for evaluating appropriateness of drug therapy. Prescribing patterns are evaluated to determine if drugs are being misprescribed, possibly resulting in problems with safety or effectiveness. Effectiveness
– The extent to which a given intervention is likely to
produce beneficial results for which it is intended in ordinary
circumstances. Efficacy –
The extent to which a given intervention is likely to produce
beneficial effects in the context of the research study. Empirical – Empirical results are those which are based on experience or observation. Encapsulation —
A drug blinding method whereby drugs (active or inert) are crushed
and put into gelatin capsules to disguise the active agent as
compared to the placebo. Endpoint — see Outcome measure Epidemiology – A study of the distribution and determinants of health-related states or events in specified populations, and the application of this study to the control of health problems. Equipoise -In healthcare, a state in which it is unknown which interventions will result in best outcomes. Equipotency - See Class effect Equivalence trial - Equivalence trials are usually used to demonstrate that the effects of two treatments do not vary more than a prespecified clinically acceptable amount and can therefore be considered clinically equivalent. Delta is the name given to the range of results within which results are judged to be equivalent. Error —
Estimate of effect – see Measures of outcomes Evidence-based medicine (EBM) – Delfini definition: "The use of the scientific method and application of valid and useful science to inform health care provision, practice, evaluation and decisions." Experimental study
– A study in which the investigator controls an exposure
or intervention, then follows subjects to compare outcomes between
research subjects who are exposed or who receive the intervention
and those who are not exposed or who do not receive the intervention.
Tip: If an intervention is "assigned" through the research,
it is an experiment. If it is chosen, then the study type is observational. Exposed group – A group whose members have been exposed to a supposed cause of disease or health state of interest, or possess a characteristic that is a determinant of the health outcome of interest. External validity
– Whether a study's results are generalizable either to
a patient population (see Population bias) or to "real world"
circumstances (see Intensity bias). Also referred to as First-line therapy: The drugs to be utilized first in treating a patient. Follow-up bias - See bias Foraging tools -
Tools that assist with keeping up-to-date with new information.
Formulary: List of therapeutic agents AQ29available in a particular practice for caring for patients. The term “preferred drug list” is also used.
Formulary system: A system that provides for the processes for establishing and managing the formulary. Generalizability – see External validity Generic substitution: Replacement of one agent with a different agent having the same chemical structure. This may be done when the patent on a brand-name drug expires. Bioequivalence is frequently assumed (i.e., it is assumed that the generic agent is equivalent to the brand-name drug). In some cases the effects of other components of the generic preparation (e.g., the vehicle in a dermatological preparation) may vary and result in outcomes that differ from those reported for the brand-name agent. Gold standard – The intervention generally believed to be the best available against which new interventions should be compared. Halo effect - How an observer's perception, knowledge or recollection, may bias results. Also used as a synonym for Placebo effect when medical attention and services can bias results. Hawthorne effect - How the knowledge of being in a study can influence a subject's behavior (usually favorably) which can bias results. Hazard ratio - The slope of the survival curve to provide estimates of how rapidly subjects are dying in the treated versus the control group. The hazard rate is the probability that if the event in question has not already occurred, it will occur in the next time interval. The time interval is short, so the hazard rate represents an “instantaneous” rate. Hunting tools -
Tools which aid the process for searching to answer a clinical
question. Hypothesis —
Tentative explanation that forms the basis of a research study. Imputation
— As in "imputation of missing variables." Principles
of Intention-to-Treat (ITT) analysis require analyzing all patients
in the groups to which they were assigned. Investigators are to
"impute" or assign outcomes for missing data points.
Example: worst case scenario in which study subjects with missing
outcomes are assigned as "treatment failures," and comparison
subjects, assigned as "successes." Frequently "last-observation-carried-forward"
(LOCF) is used, however, this has been shown to be a method prone
to bias. Imputing outcomes for missing data points is not a method
for "determining the truth of what may have happened if we
had no missing data," but rather a method to test the strength
of the outcomes (i.e., statistically similar results to those
reported) considering all the data points in such a way that does
not favor the intervention. LOCF would be especially biased in
patients with a progressive illness, for example, because patient
outcomes would appear better than would be expected in a progressive
illness. Incidence –
The proportion of new cases of the target disorder in the population
at risk during a specified time interval. Incidence rate – A measure of the frequency with which an event, such as a new case of illness, occurs in a population over a period of time. The denominator is the population at risk; the numerator is the number of new cases occurring during a given time period. Independent variable
– An exposure, risk factor, or other characteristic being
observed or measured that is hypothesized to influence an event
or outcome (i.e., the dependent variable). Inference, statistical – In statistics, the development of generalizations from sample data, usually with calculated degrees of uncertainty. Intensity bias -
An external validity bias in which the circumstances of the study
differ meaningfully from the circumstances to which study methods
might otherwise be applied (e.g., "real world" settings
as contrasted with the highly controlled environment often found
in controlled trials). Intention-to-treat
– Analyzing results for all patients in the groups to which
they were assigned whether or not they received or completed the
intervention or experienced the exposure. The number randomized
to each group should equal the number analyzed in each group —
and they should be the same people. Intermediate Outcome
Markers – Outcome measures, such as a biologic factor (biomarker)
or lab/imaging test, that are “assumed” to represent
clinical outcomes (e.g., blood pressure used as a surrogate end
point in studies of stroke). Internal Validity
- Closeness to truth within the context of the study (i.e., truth
of the study not taking into account external validity). Assessing
internal validity entails "ruling out" bias, confounding
and chance as possible explanations for an observed association
between an element of interest in a study and resulting outcomes. Interventions — Includes screening, prevention and treatments. Only valid randomized controlled trials with useful results should be used to inform decisions about interventions in these areas. Intervention bias - See bias Lead
time bias - A bias resulting from a disease found through screening
as compared to when it might otherwise have been detected. This
kind of bias can result in a treatment seeming to be very effective
if the lead time is long (e.g., “increased” survival
time). Length time bias
- A bias that can occur when certain characteristics or conditions
under study differ in the speed of progression. This kind of bias
can result in findings favoring screening. An example is tumors.
Faster-growing tumors causing symptoms will be more likely to
be found outside of screening. Slower-growing, asymptomatic tumors
will have a longer duration and be less likely to be found outside
of screening. Thus, screening will identify more slower-growing
asymptomatic tumors which could then result in a conclusion that
screening helps prevent mortality. This kind of bias is most likely
to occur in screening studies and case-control studies where prevalence
cases are included, rather than incidence cases because incidence
cases are assumed to be new starts. Likelihood ratios
(LR) – Likelihood ratios can be helpful for comparing one
test to another, and results can help rule in or rule out a condition.
Line of no difference
– The point at which there is no greater benefit or risk
one way or another (meaning the meeting point for "favors
intervention" versus "favors placebo.")
Matching – A mechanism to reduce selection bias by attempting to make a study group and a comparison group as comparable as possible by matching similar variables such as age and sex, etc. Mean, arithmetic – The average. Measure of association – see Measures of outcomes Measurement bias - See bias Measurement of risk – see Measures of outcomes Measures of outcomes
— Statistics that show the size of differences between the
results from the study groups.
Measures of test
function — Measures to help determine the accuracy and usefulness
of diagnostic tests.
Median – The data midpoint. Meta-analysis –
A quantitative technique for summarizing results of more than
one study using predetermined criteria. The goal is to provide
a summary estimate of effect based on the scientific weight of
the studies. Meta-analysis can be achieved through use of the
results of individual studies or actually pooling data from those
studies. Mode – The most frequently occurring value in a dataset. Monograph - A written review and analysis, often of a single agent, containing a list of usage recommendations. Multivariable analysis - A statistical method for determining the specific contributions of various factors to a single outcome. This allows investigation of a single variable while controlling for the effect of other variables. Multivariable analysis determines the independent contribution of each independent variable to the dependent variable (i.e., development of coronary heart disease).
Multivariate analysis - A statistical method for determining the specific contributions of various factors to a more than one outcome. Narrative review: An article in the medical literature summarizing other studies for a given topic, characterized by a lack of a transparent, scientific, and systematic approach; thus, the summary is highly likely to be misleading. Instead, systematic reviews should be sought (and appraised). Negative likelihood
ratio – The number of times the percent of false negatives
occurs over percent of true negatives.
Negative predictive
value (NPV) – Of all testing negative, percent who do not
have a condition, based on that population's prevalence. Non-inferiority trial - Non-inferiority trials aim to show that an intervention is not inferior to a comparison intervention by more than a prespecified clinically acceptable amount (Delta). Judgment is required to establish what is meant by a clinically acceptable amount. Non-parametric - Non-parametric refers to instance in which the distribution is unknown. Parametric refers to instances in which parameters of the distribution of the variable of interest is known (e.g., knowledge that a distribution of a variable is expected to follow a bell curve. Null hypothesis –
The first step in testing for statistical significance in which
it is assumed that the exposure is not related to a disease. This
is done for mathematical purposes to level the playing field to
begin with the assumption that study outcomes are the same as
what would have occurred by chance. Number-needed to
diagnose (NND) — From Bandolier: "For any chosen clinical
endpoint the NNT is the reciprocal of the fractional improvement
in a treated group minus the fractional improvement in an untreated
group NNT = 1/(fraction improved with active - fraction improved
with control)" or NND = 1/[Sensitivity - (1 - Specificity)].
Primarily useful for comparing the NND values between different
tests. Best outcome is as close to 1.0 as possible. (http://www.jr2.ox.ac.uk/bandolier/band27/b27-2.html) Number-needed-to-treat
(NNT) – The number of patients who need to be treated in
order for one patient to benefit within the study time period.
NNT is the reciprocal of the ARR. Numerator –
The upper portion of a fraction.
Objective — Goal or aim of the study. Observation bias - See bias Observational study
– Epidemiological study in which observations are made,
but investigators do not control the exposure or intervention
and other factors. Changes or differences in one characteristic
are studied in relation to changes or differences in others, without
the intervention of the investigator. Observational studies are
highly prone to selection bias, observation bias and confounding.
Tip: If an intervention is "assigned" through the research,
it is an experiment. If it is chosen, then the study type is observational. Odds - The likelihood of an event occurring compared to not occurring--e.g., odds of two to one mean that likelihood of an event occurring is twice that of not occurring. Odds ratio (OR) –
A point estimate used for case-control studies which attempts
to quantify a mathematical relationship between an exposure and
a health outcome. Odds are used in case-control studies because
the investigator arbitrarily controls the population; therefore,
probability cannot be determined because the disease rates in
the study population cannot be known. The odds that a case is
exposed to a certain variable are divided by the odds that a control
is exposed to that same variable. Odds are often used in other
types of studies as well, such as meta-analysis, because of various
properties of odds which make them easy to use mathematically. Open label study
- Not blinded as to the intervention. Outcome — see Outcome measure Outcome measure –
What we are interested in studying (e.g., mortality, use of rescue
medications in asthma patients, incidence of no flares in atopic
dermatitis). Overdiagnosis - A finding of a disease at an asymptomatic stage in a patient who would not have become symptomatic or harmed by the disease. Override - Process of setting aside a prescriber’s choice of a medication and usually substituting another medication. Parametric - Parametric refers to instances in which parameters of the distribution of the variable of interest is known (e.g., knowledge that a distribution of a variable is expected to follow a bell curve. Non-parametric refers to instance in which the distribution is unknown. Pathophysiology - Disruption of normal biochemical or physical function by a disease or disorder. Pearson correlation
co-efficient - An analysis method to measure the extent of the
linear relationship of two variables to determine how independent
and dependent variables change together (e.g., salt intake and
blood pressure). Performance bias - See bias Performance measure – A quantitative assessment of a process of care, outcome or service usually used in quality improvement work. It consists of a denominator (e.g., population of interest), a numerator (i.e., a count of events of interest occurring within the denominator) and a frequency (i.e., the specified interval for measurement). May target various levels or units such as a system, specialty group or individual. Usually expressed as a rate, ratio or percentage. Sometimes used interchangeably with “quality indicator.”
Pharmacy & Therapeutics Committee: Committee charged with making formulary management decisions, along with performing other formulary management functions. Pharmacy benefit manager (PBM): A company that manages pharmacy benefits and formulary management for health care systems and/or insurance companies. Phase (clinical) studies –
Placebo – An
inactive substance or procedure administered to a patient for
comparison to the intervention. Used in clinical trials to blind
people to their treatment allocation. Pivotal trial - A controlled trial to evaluate the safety and efficacy of a drug in patients and usually the basis for the New Drug Application (NDA) filing with the FDA. Point estimate – see Measures of outcomes Population – For research studies, the term "population" may refer to a group of people living in a specific geographic area or it may mean a unique set of individuals who share some characteristic such as exposure to a disease or a group of patients of a single clinician, as examples. In statistics, this can mean a population of data. Population bias
- An external validity bias in which the population under study
differs meaningfully from the population to which the results
might be applied. Positive likelihood
ratio (LR+) – The number of times the percent of true positives
occurs over percent of false positives. This represents the change
from pre-test odds to post-test odds.
Positive predictive
value (PPV) – Of all testing positive, percent who have
the disease, based on that population's prevalence. Post-test odds for
positive test — The odds of a person having a condition
if the test is positive. Post-test odds for
negative test — The odds of a person not having a condition
if the test is negative. Post-test probability
for a positive test — After learning test result, the probability
that a person testing positive has the condition. Post-test probability
for a negative test — After learning test result, the probability
that a person testing negative has the condition. Power - See Statistical power. Precision – The ability to provide sufficient detail, such as small incremental units, to be useful. Predictive value
– A diagnostic measure of function which is used to address
the probability of having or not having a condition. Predictive
value is affected by prevalence. Pre-test likelihood
— Pre-test probability of having a condition based on the
prevalence of the condition in the population studied. Pre-test odds —
Pre-test odds of having a condition based on the prevalence of
the condition in the population studied. Prevalence –
The number or proportion of cases or events or conditions in a
given population. Prevalence rate – The proportion of persons in a population who have a particular disease or attribute at a specified point in time or over a specified period of time. Prior authorization: Requirement that a clinician obtain approval before a drug can be dispensed and/or covered. Probability - The likelihood of an event occurring expressed as a number between 0 and 1. It is measured by the ratio of the event to the total number of possible events. Example: the probability of flipping a coin and coming up with heads is .5. Proportion –
A type of ratio in which the numerator is included in the denominator.
The ratio of a part to the whole, expressed as a “decimal
fraction” (e.g., 0.2), as a fraction (1/5), or, loosely,
as a percentage (20%). Proxy outcome markers – see Intermediate outcome markers. Publication bias – The bias toward the publishing of studies showing statistically significant “positive” results. P-value –
The probability that the results observed in a valid study could
have arisen by chance. The p-value of <.05, often set as the
point of statistical significance, is conventional and simply
means there is less than a 1 in 20 chance that the findings observed
are due to chance. Qualitative research - Any research that utilizes non-quantitative methodology, e.g., case studies or narrative description or interviews. Quality indicator – Oftentimes used interchangeably with “performance measure.” Quality indicators are specific and measurable elements of health care that can be used to assess the quality of care. Randomization – Mechanism to assign, by chance, “intervention / no intervention” assignments to study participants in randomized controlled trials.
Randomized controlled
trial (RCT) – An experiment in which investigators control
an exposure or intervention, and subjects are randomly assigned
to the study group or to the control group or groups. Intervention
or exposure then is performed. Patients are followed to determine
outcomes such as improvements or harms. Random sample – A sample derived by selecting individuals such that each individual has the same probability of selection. Rate – Derived
by dividing a numerator by a denominator. The numerator is a subset
of the denominator. Example: The percentage of diabetic eye exams
for Type I diabetics was 80% for our clinic. Ratio – A numerator
and denominator (the numerator is not required to be a subset
of the denominator). Example –The ratio of women to men
in these studies was 1 to 5 (which can also be expressed as 1:5).
Or, the rate and ratio for diabetic eye exams in our clinic would
be 80/100. Rebate: A cost offset that pharmacy systems receive when purchasing large quantities of drugs. AQ30Selection of drugs for large bulk purchase is based on the drugs’ volume of use. Recall bias –
Inaccuracies resulting from data collected from study participants
who are asked to retrospectively self-report on study items of
interest. This kind of bias is most likely to occur when certain
behaviors are hard to track (e.g., diet) or are often hidden for
some reason such as being socially sensitive (e.g., sexual behaviors,
smoking behavior, etc.). Regression model – A method for estimating the relationship of a dependent variable to a variety of independent variables. Regression-toward-the-mean – Returning to the “average” state. Meaning that extreme test values are statistically likely to move to an average over time. When patients present with extreme values and then seem to have improvement, it may be falsely attributed to an intervention when it is truly due to regression-to-the-mean. A comparison group with no intervention can help expose this effect. Relative risk (RR)
- A point estimate in which the risk of some health-related event,
such as disease or death, is compared between a study group and
a comparison group. Relative risk is expressed as the number of
times one group may be at risk over another. A relative risk of
less than 1 is a lower risk. Relative risk reduction
(RRR) – A point estimate in which the percent reduction
in events in the intervention or exposed group is compared to
the comparison group — relative risk reduction is the proportional
difference in size between outcomes. For example, if we had 15%
mortality in a comparison group as compared to 10% mortality in
the intervention group, since 10 is one-third smaller than 15,
the RRR would be 33%. Relative risk reduction begs the question
"relative to what?" Because RRR overestimates the actual
difference, it should always be used in conjunction with another
point estimate such as absolute risk reduction or number-needed-to-treat
when communicating study results. Representative sample – A sample whose characteristics correspond to those of the original population or population of interest. Retrospective study – A study method to look at cause and effect after outcomes have already occurred. Generally this refers to a case-control study; however, cohort studies can use a retrospective method. Risk – The probability that an event will occur. Risk adjustment – The process of adjusting performance rates or other outcomes of care to level the playing field due to differences in health status between populations. Risk factor – An aspect of personal behavior or lifestyle, an environmental exposure, or an inborn or inherited characteristic that is associated with an increased occurrence of disease or other health-related event or condition. Risk ratio – see Relative risk Risk stratification – The process of, or result of, separating a sample into subsamples based on health status or risk factors such as age, comorbidities, etc. Safety
— Safety in research studies pertains to harms of various
interventions. Sample – A
selected subset of a population. A sample may be random or non-random
and it may be representative or non-representative. Sampling error - See chance Second-line therapy: Therapy for patients failing on drugs established as “first-line” therapy. Selection bias – See bias Sensitivity (SN)
– Correct identification by a screening test or case definition
as having disease – of all those with a disease, the percent
testing positive (true positives). Sensitivity is derived from
calculations based on people who are known to have the condition.
Sensitivity is especially useful when it is important not to miss
a disease. High sensitivity is considered to "rule in"
a condition. Specificity (SP)
– Correct identification by a screening test or case definition
as not having disease – of all those without a disease,
the percent testing negative (true negatives). Specificity is
derived from calculations based on people who are known not to
have the condition. Specificity is especially useful when it is
important to avoid false positives. High specificity is considered
to "rule out" a condition. Standard deviation (SD) – A widely used measure of dispersion of a frequency distribution, equal to the positive square root of the variance. Standard deviation can tell you how widely or how tightly variables are distributed around the average. One standard deviation accounts for approximately 68 percent of the distribution. Two standard deviations account for approximately 95 percent. Statistical power
- The ability to reliably determine a statistically significant
relationship between interventions if one exists. Power comes
into play in planning and interpreting studies. Investigators
want to know the risk of drawing erroneous conclusion about efficacy
and so they do power calculations to estimate the number of subjects
needed in a study to show a statistically significant result if
one exists. It is conventional to accept a 5% risk of concluding
that an outcome is truly different in the study groups when they
are not (Type I or Alpha error). It is also conventional to accept
a 20% risk of concluding that an outcome is not statistically
significant when in truth there may be a true difference (Type
II or Beta error). Power can be determined by considering Alpha,
Beta, the event rate in the comparison group, and what is judged
to be a clinically significant difference between the groups.
As a practical matter, many users focus on the confidence intervals.
For valid studies, consider what you judge to be a reasonable
range for clinical significance – this need not be hard
and fast. For statistically significant findings, is the confidence
interval wholly within bounds for clinical significance? For non-significant
findings, is the confidence interval wholly beneath your limit
for clinical significance? A yes to these two questions means
likely conclusive findings for valid studies. No, means findings
are inconclusive. Statistical significance
– The extent to which study results are unlikely to be due
to chance. Expressed as a p-value and it is typically set at <.05
due to convention. Statistical significance may be set lower,
but must be determined in advance of the study. Subgroup analysis — Analysis of research results on subsets of patients in a study, such as a subgroup of Caucasian women in a study of asthmatics of all races, ages and genders. Subgroups for study must be determined in advance of a study — otherwise the results are highly likely to be due to chance. Superiority trial - Clinical trials which are conducted to determine if one intervention is superior to another. Surrogate end points – See Intermediate outcome markers. Systematic review
– Studies that examine more than one study on a given topic
in a systematic way using predetermined criteria. The goal is
to provide a summary estimate of effect based on the scientific
weight of the studies. Therapeutic substitution: Substitution of a drug with an agent having a different chemical structure but similar clinical benefits. Trend - A term used to describe observed results in several studies which consistently go in the same direction, but are not statistically significant. Triangulation issues: A catchall term for all the various considerations that need to be made in making a decision. Type 1 error - see Error Type 2 error - see Error Two-by-Two Table — Table to report results and from which statistics are calculated. See examples below: Intervention Example
Diagnostic Testing Example
Variable – Any characteristic or attribute that can be measured.
Related: Dependent variable, independent variable Variance – A measure of the spread of a variable about its mean value. In a data set, a single point lies above or below the mean for the entire dataset. A deviation score is the measure of how much each point lies above or below the mean for the entire dataset. Variance is the mean of all the deviation scores. It is a way of mathematically stating how far deviations are from expected results. Washout period – Used in a cross-over trial to discontinue patients on a treatment before the second treatment is given to reduce the likelihood that the first treatment will affect the outcome even after it is stopped. |
Resources:
Glossary
.......
Home
What's
New
Intro
Links
Services
Resources
Success
Stories
Notices
About
Us
About Our Work
Testimonials
Other
Site
Navigator
Contact
Us
Site
Search
.......
delfini.org
Also
creators of
medicalleaders.org
Quick Navigator to Selected Resources:
Delfini SERVICES
We offer a wide range of supportive services with the ultimate goal of increasing health care organizations’ and clinicians’ abilities to provide high quality, evidence-based care and to improve medical decision-making for organizations, leaders, teams, clinicians and patients.Delfini About Our Work