INBDE Bootcamp High-Yield Research | Bootcamp.

com

Measurements

Screening vs. diagnosis Reliability Validity

Screening - detection of disease targeted at an Reliability - how well research methods reproduce the same Validity - ability to distinguish between health and
asymptomatic population results, multiple times disease

Diagnosis - classification provided to symptomatic ● Intra-examiner reliability - amount of variability made by the True positive (TP) - patient has a disease and the test
patients seeking care same examiner on multiple occasions correctly detects the disease
● Inter-examiner reliability - amount of variability between two
False negative (FN) - patient has a disease and the test
or more examiners incorrectly detects no disease

False positive (FP) - patient does not have a disease

and the test incorrectly detects disease
Sensitivity and specificity Predictive values
True negative (TN) - patient does not have a disease
Sensitivity (Se) - measures the accuracy of the test in Positive predictive value (PPV) - probability a test will and the test correctly detects no disease
detecting disease accurately identify a disease

● Se= TP / (TP+FN) ● PPV= TP / (TP+FP)

Test result
● High sensitivity indicates low numbers of false
negatives
Negative predictive value (NPV) - probability a test will Positive Negative
accurately identify no disease
Specificity (Sp)- measures the accuracy of the test in Yes True False
● NPV= TN / (TN+FN) positive negative
detecting health
Disease (TP) (FN)
● Sp= TN / (TN+ FP)
No False True
● High specificity indicates low numbers of false
positive negative
positives (FP) (TN)

Area under the curve (ROC curve):

● A higher area under the curve signifies there is a
higher true positive rate and therefore a higher test
accuracy
● A lower area under the curve means there is a higher
false positive rate and therefore a lower test accuracy
 INBDE Bootcamp High-Yield Research | Bootcamp.com

Confidence interval Null hypothesis Error

Confidence interval (CI) - provides a range of values Null hypothesis - proposes that there is no significant Type 1 error - false positive result
that a percentage of the population likely falls into difference between the outcomes of different groups
● Rejecting a null hypothesis that is actually true
● For example, if a study has a 95% CI of (3,5), there ● Investigators aim to disprove the null hypothesis ● Probability of making this error is represented by an
is a 95% chance that the true population has the ● If the confidence interval does not include zero (null alpha value which is the p-value for rejecting the null
range of values 3 to 5 hypothesis), the results are statistically significant and the null hypothesis
hypothesis can be rejected ○ P-value: measures probability that the sample data
● If the confidence interval does include zero (null hypothesis), is due to chance, typically is set to 0.05
the results are not statistically significant and the null
hypothesis cannot be rejected Type 2 error - false negative result
○ May occur if the study has a low statistical power
● Failing to reject a null hypothesis that is actually false
● Probability of making this error is represented by the
beta value which is the statistical power
○ Statistical power: probability of observing an effect

Prevalence Incidence Risk predictors

Prevalence - frequency of disease for the population Incidence - rate of developing new disease Risk factors - causally associated with disease

● Prevalence = number of previously diseased ● Incidence = number of new cases / total population Risk indicator - marker of exposure to a risk factor,
cases / total population ● Only involves people who develop a new condition during a indirectly linked to a disease
● Sample population provides an estimate for the specified time period
target population
● Number of people who already have the condition
during a given time

Relative risk Odds ratio Significance

Relative risk - ratio of the risks for an event in the Odds = exposed/unexposed Statistical significance - measured mathematically
exposed group / risks for an event in the unexposed Odds ratio (OR) = odds of exposure in cases / odds of using p values
group exposure in controls = AD/BC
● Conventionally, a value lower than an alpha
● RR = 1: risk predictor is not associated with disease significance level of 0.05 is considered statistically
● OR = 1: exposure is not associated with disease significant and the null hypothesis can be rejected
● RR > 1: risk predictor is associated with an increased
● OR > 1: exposure is associated with increased odds of
risk of disease
disease Clinical significance - measured using numbers needed
● RR < 1: risk predictor is associated with a decreased
● OR < 1: exposure is associated with decreased odds of to treat (NNT)
risk of disease
disease
● NNT signifies the total number of patients needed to be
Risk Ratio (RR) = incidence rate in the exposed treated in order for one patient to be cured from
population / incidence rate in unexposed Cases (disease) Controls (no disease) disease
○ NNT of 1 is the best
Exposed A B ○ Only considered if p < 0.05
Unexposed C D
 INBDE Bootcamp High-Yield Research | Bootcamp.com

Problems or biases

● Convenience sampling- the study population is chosen from a group that is easy to reach or access
● Incomplete data - differing rates of responses from relative groups
● Hawthorne effects - participants perform differently in a study than if they were not in a study
● Regression to the mean - values may be extremely high or low to begin, but over time they will regress to an average level
● Placebo effect - tendency for people to respond favorably knowing they are receiving treatment
● Publication bias- when the outcome of a study biases whether or not to publish the study

Study designs

Single group study Case study Cross-sectional survey

Studies a group of patients with a disease and Studies a single person, group, or situation involving a rare Studies a random sample (cross-section in time) from a
compares them to a historical control group condition target population

● Easy to administer
● Cannot prove a cause and effect relationship
● Measures prevalence

Case-control study Cohort study Non-randomized clinical trial

Studies a random sample selected based on presence Studies a random sample from a healthy, at risk population and Studies non-random sample groups
or absence of disease follows them over time
● Samples are placed into either one of two control
● Samples are placed into either a case group (people ● Identifies new events of disease groups or one of two different treatment groups
who had an exposure) or a control group (people ● Measures incidence
who did not have an exposure) ● Measured using relative risk
● Individuals are observed to see if the predicted ● Prospective cohorts follow individuals over time
outcome occurs ● Retrospective cohorts collect information about individuals’
● Measured using odds ratios pasts
● This study is most impacted by recall bias (change in
the risk that participants in a study are able to recall
or report information)
 INBDE Bootcamp High-Yield Research | Bootcamp.com

Randomized control trial (RCT) Systematic review / meta-analysis Hierarchy of scientific evidence

Studies two randomly assigned sample groups Studies results from two or more published studies to create a
meta-analysis
● Samples are placed into either a treatment ● A systematic review collects and summarizes data that fits
(intervention) or control group into the specified category
● A meta-analysis uses statistical methods to analyze the
results of all of the gathered studies

PICO Types of sampling Measures of central tendency

PICO - an acronym that is used to help formulate ● Consecutive sampling - selecting subjects who meet the Mean - average of values in a data set, affected by
clinical research questions study criteria until an adequate sample size is obtained outliers
● Voluntary response sampling - involves volunteers who
● P= population, patient, or problem Median - middle score in a data set, least affected by
agree to participate in a study
● I= intervention outliers
● C= comparison ● Snowball sampling - involves an initial subject who then
● O= outcome recommends another subject who meets the study criteria Mode - most frequent value in a data set
● Stratified random sampling - obtaining a random sample
from a population and then stratifying them into subgroups
with similar criteria. Sample groups are then created by taking
a member from each subgroup
● Convenience sampling - inviting participants who are easy to
contact
● Cluster sampling - selecting samples after dividing a
population into subgroups that do not have any overlap in
similarities
● Systematic sampling - selecting participants from a list at
random set intervals
 INBDE Bootcamp High-Yield Research | Bootcamp.com

Types of variables Forest plot Box and whisker plot

● Continuous - numeric variables that are obtained by ● Graphing technique used to summarize information from ● Graphing technique used to represent numerical data
measuring. individual studies in a meta-analysis ● Plots values including the median and quartiles,
○ Example: millimeters ● Width of the horizontal line represents the range of a variable dividing the data set into groups of equal sizes
● Categorical - assigned based on a qualitative being tested (examples: relative risk, odds ratio) within the ● Interquartile range- difference between the first (Q1)
property, not numeric. confidence interval and third (Q3) quartiles. Indicates how spread out the
○ Example: hair color ● If a confidence interval crosses the vertical line of “no middle 50% of the data set is
● Discrete - numeric variables that are obtained by difference” (odds ratio or relative risk of 1), the outcome is not
taking a count from a set of distinct whole values. statistically significant
○ Example: number of teeth in a patient’s dentition
● Ordinal - variables that are obtained by ordering or
ranking.
○ Example: the first, second, or third individual to
finish a task
● Binary - variables that can only take on two possible
values or categories.
○ Example: the presence or absence of oral
cancer
● Independent - variable that is manipulated by the
researcher.
○ Example: different types of analgesics
● Dependent - variable that is measured by the
researcher.
○ Example: reported pain level

Research notation

Example notation: 95%, [s,n] [1.0,500]:[1.4-1.6]

● 95%- represents the level of confidence.

● [s,n]- represents the standard deviation (s) and the
sample size (n)
● [1.0,500]- corresponds with the s and n values of
the study.
○ Standard deviation (s) = 1.0 and sample size (n)
= 500
● [1.4-1.6]- represents the confidence interval of the
data