Reading...
Play button
Play button
Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
70 Cards in this Set
- Front
- Back
|
Prevalence
|
the number of people who have an outcome divided by the number of people at risk of having tht outcome, at a specific time period.
|
|
|
Incidence
|
refers to the number of people who will develop the new outcome divided by the people at risk of developing the new outcome over a specific period of time.
|
|
|
Case report(descriptive)
|
detailed report of an outcome(disease etc) of a single person
|
|
|
Case studies (descriptive)
|
detailed report of an outcome of several people
|
|
|
Correlational studies
|
also known as ecologic studies, exposure and outcome are both included however exposure data and outcome data come from:
Two different groups at one point in time One group at two different points in time ex colon cancer and meat consumption study. |
|
|
Probs with correl studies / Ecologic fallacy
|
Inability to link exposure to outcome in same individual means that one cannot be sure that an association exists between exposure and outcome
This inability may suggest an association that does not truly exist (ecologic fallacy) |
|
|
cross sectional studies
|
Most common type of descriptive study
Exposure and outcome data are collected from the same individual at one point in time Many exposure-outcome combinations may be studied at once Also known as “snap shot” studies Frequently used to describe a population, particularly in terms of person, place, and time. |
|
|
prob with xsectional studies
|
One problem is the inability to determine whether exposure preceded outcome
Cigarette smoking and oral cancer example. |
|
|
descriptive studies
|
eg: case studies, correlational,xsectional studies.
Descriptive studies are: Relatively simple and inexpensive Useful for hypothesis generation Less useful for hypothesis testing |
|
|
Analytical
|
Most useful for hypothesis testing
Relatively complex and more expensive May be more time consuming |
|
|
observational
|
Researcher does not manipulate exposures
Less time consuming and expensive Less support for an association |
|
|
experimental
|
Researcher manipulates exposure status
Can be time consuming and costly Strongest support for an association May be subject to ethical constraints |
|
|
case control studies
|
For case-control studies, subjects are enrolled into study according to whether they have an outcome (disease, condition), and then exposure status is assessed. Most common analytical study design
Relatively inexpensive because subjects are not followed over time Well suited for studies of rare outcomes (diseases, conditions) Allows tests of multiple exposures Subject to selection and observation biases. |
|
|
cohort studies
|
For cohort studies, subjects are enrolled into study according to whether they have an exposure, and then outcome (disease, condition) status is assessed.More useful type of analytical study design
Relatively expensive because subjects are followed over time Cost depends on design – retrospective less expensive than prospective Well suited for studies of rare exposures Allows for the testing of multiple outcomes Subject to loss to follow-up bias. |
|
|
selection bias
|
Relation between exposure and outcome among those who are selected into study is different from the relation for those who would have been eligible but were unwilling to participate or were not selected
Occurs when selection of cases and controls into the study is dependent on exposure status Problematic in case-control studies because exposure and outcome have both occurred by the time subjects are selected |
|
|
observation bias
|
Includes recall and interviewer biases
Recall bias occurs when cases remember details about their exposure history differently than do controls Interviewer bias occurs when the researcher consciously or unconsciously asks exposure history questions of one group differently than of the other group |
|
|
loss follow up bias
|
Subjects followed over time may not complete study
Loss of interest Change of residence Morbidity or mortality Bias occurs when loss to follow-up is dependent on exposure status |
|
|
Prospective cohort
|
exposure groups established today.
|
|
|
Retrospective cohort
|
if the exposure groups established in the past , ie everything happened in the past. eg smoking surveys from the past, then looking at the data years later to see who died of cancer, oral cancer.
Prob- limited information, undetected diseases such as if the person dies in an automobile accident but could have had oral cancer |
|
|
Experimental studies
|
also referred to intervention studies, clinica trials. Similar to cohort studies (observational) in that subjects are selected into the study based on exposure status
Different from cohort studies in that researcher controls exposure status Well conducted experimental studies represent the strongest form of epidemiological evidence because effect of covariates is minimized. |
|
|
Randomization
|
Exposures are randomly assigned to the treatment (exposed) and comparison (non-exposed) group
Maximizes the probability that the treatment and comparison groups are similar in terms of covariates • Minimizes the probability that covariates affect the exposure-outcome pathway |
|
|
Blinding
|
• Single blinding: subjects are unaware of treatment assignments
• Double blinding: subjects and researchers are unaware of treatment assignments |
|
|
Placebo
|
• Inactive form of a treatment, resembling the true treatment in as many ways as possible
• Standard treatment o Sometimes you cannot withhold an active ingredient from the comparison group o If this is the case, researchers must use an already established treatment o This results in a less pronounced difference between a treatment and comparison group than when placebos are used |
|
|
Efficacy trials
|
test whether a treatment works in ideal conditions
•Study groups are susceptible to outcome •Groups are observed to ensure that they adhere to treatment protocols |
|
|
Effectiveness
|
Effectiveness trials: test whether a treatment works in normal conditions
|
|
|
Odds ration
|
o Describes the odds of an outcome occurring among an exposed group compared with the odds of an outcome occurring among a non-exposed group
o Used for cross-sectional and case-control studies. Shows the odds of an outcome happening in one group compared to another. |
|
|
Relative risk
|
Relative risks (RRs) describe the risk of an outcome occurring among an exposed group compared with the risk of an outcome occurring among a non-exposed group
Appropriate for cohort and experimental studies |
|
|
Researcher training
|
Belmont report,
Nuremberg report |
|
|
Screening measures
|
oIdentifies individuals who might have an outcome
oDoes not test an association between exposure and outcome Instead, it tests the association between a screening test and a “gold standard” test Measures sensitivity, specificity, positive n negative predictive value. |
|
|
Sensitivity
|
proportion with disease who test positively. a/a+c
|
|
|
Specificity
|
Proportion without disease who test negatively. d/d+b
|
|
|
Positive predictive value
|
Proportion of those who tested positively who truly have disease = a / (a + b)
|
|
|
Negative predictive value
|
proportion of those who tested negatively who do not have the disease; d/d+c
|
|
|
Population
|
group to be reported or researched on.
|
|
|
Samples
|
selected from within a population
Should infer back to the population AKA statistic Multiple samples can be considered per a given population. Samples gives ranges of what the population actually is |
|
|
Statistical inference
|
method to describe a population by using a sample and accepted statistical method. two types: parameter estimating, hypothesis testing.
|
|
|
Parameter estimating
|
sample from within a population is selected and inferences are made about the population based on the sample
-As sample size increases, accuracy increases -The probability sampling method is the key to accuracy with this. |
|
|
Hypothesis testing
|
the results from the sample are used to infer an association between exposure and outcome in the population
-As sample size increases, accuracy increases -The probability sampling method is the key to accuracy with this |
|
|
Sampling
|
Statistical method concerned with the selection of individual observations from a larger study population
Sampling objectives Efficiency Generalizability (external validity) |
|
|
Population
|
is a complete set of persons with a specified set of characteristics
Two main types of populations Target population Accessible population A sample is a subset of the accessible population |
|
|
Target population
|
the complete set of individuals about which the research is concerned (or about which results will be generalized)
|
|
|
Accesible population
|
the subset of the target population that is available for study
|
|
|
Samplying frame
|
It may be impractical and/or scientifically unsound to measure every individual in the accessible population. Sampling frames allow the researcher to identify every individual member of the accessible population and to include any one in the sample
Examples Study of the brushing behaviors of child dental school patients during 2000-2004 |
|
|
Probability samplying
|
individuals in the population have a known probability of being included in the sample; more scientifically acceptable than non-probability sampling
|
|
|
Nonprobability samplying
|
individuals in the population do not have a known probability of being included in the sample; often used for convenience and/or lack of resources
|
|
|
Simple random samplying
|
Suitable applications
Accessible population is fairly homogenous Minimal information about the accessible population is available Usually conducted without replacement (as in drawing marbles from a bag) |
|
|
Stratified random samplying
|
Most appropriate for accessible populations with sub-groups that differ considerably
Process Divide accessible population into strata (must be mutually exclusive and must be collectively exhaustive) Select a sample from each stratum according to simple random sampling method |
|
|
Systematic samplying
|
Defined as selection of every nth element from a sampling frame
n (sampling interval or sampling fraction) is equal to: # in accessible population / # in sample Process Calculate sampling interval Begin process of selecting every nth element at a random beginning point in the list |
|
|
External validity
|
Assessment of the applicability of research findings from the study sample to a larger population
Yields results that are worthwhile to a larger group, thus increasing utility Directly related to sampling method and indirectly related to response rates |
|
|
Response rates
|
Defined as the percentage of persons who participated in the research divided by the total number of persons in the sample
Goal response rates range from 70%-85% depending on type of study and accessibility of population Bias occurs when non-respondents affect the true association between exposure and outcome External validity decreases with low response rates |
|
|
Central tendancy
|
Central tendency refers to a measure of the “middle” of a set of data
There are several measures of central tendency which depend on level of measurement Most common measures in healthcare research Mode Mean Median |
|
|
Mode
|
Defined as the most frequently occurring value within a set of data
Distributions may have more than one mode (2 modes=bi-modal, 3+ modes=poly-modal, etc.) |
|
|
Mean
|
Also known as arithmetic mean or x
Equal to the sum of all observations divided by the number of observations in the data set Sensitive to extreme values |
|
|
Median
|
Middle number of a set of ordered values
Represents the value below which 50% of the distribution falls (50th percentile) Equals [(n + 1) / 2]th largest observation when n is odd Equals mean of (n/2)th and [(n/2) + 1]th largest observations when n is even Less sensitive to extreme values than the mean |
|
|
Measures of variability
|
Variability refers to statistical dispersion or how spread out the values in a data set are
Most common measures in healthcare research Variance Standard deviation Range |
|
|
Variance
|
Variance is the square of the distance of each data point from the mean (s2)
|
|
|
Standard deviation
|
Standard deviation is the square root of the variance (s)
|
|
|
Range
|
max - min
|
|
|
Norminal
|
Variables contain 2 or more categories that lack numerical context
Examples = gender, race, political party (all given numerical assignments having no meaning) Mathematical operations cannot be done Can be used to classify data into categories Only modes can be used to describe the data gathered |
|
|
Ordinal
|
there is an implied order, usually numbers are attached to them. Variables contain 2 or more categories that may be ranked relative to one another
Examples = attitudes (1=agree, 2=neutral, 3=disagree) (all given numerical assignments having no meaning) Mathematical operations cannot be done Can be used to classify data into categories and rank order Modes and medians can be used to describe the data gathered |
|
|
Interval
|
Appropriate for variables containing two or more categories that differ from one another according to a set interval, but not in absolute terms (zero point is arbitrary)
Examples Fahrenheit/Celsius temperature Year date Modes, medians, and means can be used to describe the data gathered |
|
|
Ratio
|
Variables contain 2 or more categories that differ from one another according to a set interval AND in absolute terms
• Zero point exists and is not arbitrary Example = age, number of teeth All mathematical operations can be performed on this Can be used to classify data into categories, rank order, and equal intervals Modes, medians, and means can be used to describe the data gathered RatiO data has a zerO point |
|
|
Linear regression
|
o y = a + bx
y = dependent variable a = intercept b = slope x = independent variable o The slope (b) has the same meaning (and ranges) as the correlation coefficient |
|
|
Multiple linear regression
|
o Sometimes there are covariates that complicate a study, but must be considered (controlled) when collecting the data
This gives a more realistic relationship when comparing two variables o As the number of covariates increases, the actual effect of the variable is more clearly seen |
|
|
Confidence intervals
|
oThe range within which the true magnitude of an association lies given a specified level of assurance (95% confidence interval)
oGives some sense upon repetition of an experiment what the range of the correlation coefficients are |
|
|
Chance
|
Findings occurred not because of an association of exposure and outcome, but because the stars were aligned properly
The p-value (probability value) is assessed to determine the role of chance • If p ≤ 0.05 than results obtained are statistically significant, showing a specific relationship between variables • Example: p = 0.12 means that 12% of the understanding from the research is due to chance |
|
|
Type I and Type II (chance)
|
-• Type I: statistically significant associations are found (p ≤ 0.05) but there is really none at all
oThe null hypothesis was true, but results caused you to reject it •Type II: the null hypothesis is false, but it was not rejected |
|
|
Confounding
|
Observed association between exposure and outcome is affected by covariates
•Covariates are independently associated with exposure and outcome Can effect the exposure and/or outcome Multiple ways to control for confounding Stratified analysis Restriction Matching Adjusted analysis Random assignment of treatment (randomization in experimental studies). |
|
|
Bias
|
Causes observed estimates to be smaller/larger than the true estimates
Can be: •Selection bias (case-control) •Observation bias(case-control) oRecall bias oInterviewer bias •Loss to follow-up bias (cohort and experimental) Bias is not the same as error •Errors become biases when they are associated systematically with exposure/outcome in a way that increases/decreases the magnitude of the association •Error does not always cause bias If an error occurs across the board in an equal manner it is not biased |
|
|
Causation
|
Strength between E and O is measured by: strength of association, temporal relation, biological plausibiliy, dose response, reversibility, consistency, analogy, specificity.
|
