Transcript Epidemiology in Medicine

Epidemiology in Medicine
Sandra Rodriguez
Internal Medicine
TTUHSC
Epidemiology and Research
Study of the frequency and cause of
disease in human populations.
Research allows understanding of risk
factors, progression of diseases, treatment
effectiveness, outcomes and highlights
research needs.
After critical evaluation of data, physicians
can make clinical decisions and improve
populations health (EBM).
Types of Statistical Studies
Descriptive
– Take data, arrange them, and present them to
demonstrate associations or to generate a
research hypothesis for further studies
Analytic
– To compare exposures to risk factors with
disease states, allow hypothesis testing and
statistical analysis
– Could be without or with intervention
Types of Statistical Studies
Descriptive
– Case Reports and Series
– Correlation studies
Large sample size to identify associations between
disease and variables
Best to generate research ideas
– Cross-Sectional Studies
Evaluates a group at one point in time
Causal links can be speculated but no conclusions
can be drawn.
Types of Statistical Studies
Analytic
– Case control
Study group analyzed to identify associations
Begins with a population and it’s disease and can
be evaluated retrospective to determine exposure
– Prospective Cohort Study
Disease-free subjects are followed overtime to
identify onset of disease, or incidence
Used to establish relative risk
Types of Statistical Studies
Analytic
– Meta-Analysis
Quantitative analysis of two or more independent studies into
a large one for analysis of variables and results. Gives a
statistic summary and is used to increase knowledge beyond
one study, guide diagnosis and treatments and point toward
research.
– Interventional study
Clinical trial
Ideally randomized, blind, designed to minimize impact of
bias and confounding factors.
Crossover study design.
Reviewing literature
Study design
– Hypothesis: Research and Null
– Theory/natural law
– Sample
Inclusion/Exclusion criteria
Randomization
Matching controls
– Test: Threshold for normal, sensitivity, specificity, use depending
prevalence of disease in your population.
– Measures for assessment results
Accuracy
Precision
Statistical analysis.
Intention to treat.
Validity
Reliability
–
–
–
–
Bias: Selection, Observational
Confounding factors
Efficacy
Effectiveness
Reviewing literature
Statistical power
– Accepting or rejecting the null hypothesis is
the basic thrust of any study
– Power: Likelihood that a statistically
significant difference would be found between
two groups given that a difference truly exists.
Sample size: Rule of 3, if a condition occurs 1 in
to 10, then the population needed is three times
this number (30) for a statistically significant study.
Experiment design
Reviewing literature
Type I error
– Rejecting a true null hypothesis/accepting a false
positive research hypothesis; hence a false positive
result
– Alpha is the frequency of occurrence of a type I error
– The probability of committing a a type I error is the Pvalue
– P-value is the probability that the null hypothesis is
true, and the lower the more significant, <0.05 means
that less than 5% possibility that result is by chance,
<0.01 means that less than 1% possibility that result
is by chance.
Reviewing literature
Type II error
– Accepting a false null hypothesis/rejecting a
true positive research hypothesis; hence
false-negative result
– Beta signifies the frequency of a type II error
occurring
– Protect with statistical power
Statistical Analysis
Noncontinuous
– P-values
– The chi-square which concerns the frequency
of event occurrence
– The Fisher’s exact test which estimate the Pvalue when small samples are used
Statistical Analysis
Continuous
– One sample T-Test
Compares the sample mean value to a known
mean of a standard variable
– Two sample T-Test or paired T-Test
Compares the mean values with two independent
groups
– ANOVA
Compares values in more than two independent
groups, the variation is between and within group
Statistical Analysis
The normal distribution
– Mean: Sum of all values divided by number of
them
– Median: The middle most observation
– Mode: The most frequent observation
– Standard deviation: A measure of spread
1 SD: About two thirds of data
2 SD: About 95% of date will fall within
3 SD: Virtually comprises 100% of the data
Statistical Analysis
Sensitivity refers to ability of the test to correctly
identify patients who have disease.
s= All positive results in disease (TP) x 100
All specimens with disease (TP+FN)
Specificity refers to the ability of a test to
correctly identify patients who do not have a
disease.
sp= All negative results without disease (TN)
All specimens without disease (TN+FP)
x 100
Statistical Analysis
Prevalence
– Proportion of persons with the disease among a
group to whom test is applied
P= Disease present (TP+FN)
All group (TP+FP+TN+FN)
Pretest probability
– % of patients who have the target disorder as
determined before test is performed
Statistical Analysis
Likelihood Ratio summarizes in a single number
the clinical utility of a test, and it is added to the
pretest probability to increase certainty:
– LR of a positive test: s/(1-sp)
2 increases probability of disease by 15%
5 increases probability of disease by 30%
10 increases probability of disease by 45%.
– LR of a negative test: (1-s)/sp
0.5 decreases probability of disease by 15%
0.2 decreases probability of disease by 30%
0.1 decreases probability of disease by 45%
Statistical Analysis
Odds Ratio
– Compare a portion of the affected population with the
unaffected population and is expressed as a ratio
– The OR gives the odds of having a risk factor if the
condition is present as compared to having a risk
factor if the condition is not present
– The higher, the stronger association.
Odds ratio= risk factor with disease present A/C
risk factor without disease
B/D
Predictive values
Positive predictive value
– Describes the probability that a patient who
has an abnormal test actually has the disease
– Directly proportional to the prevalence of the
disease
PPV= Number of true positive results (TP) x 100
All positive test results (TP+FP)
Predictive Values
Negative predictive value
– Describes the probability that a patient who
has a normal test is actually free of disease
– Inversely proportional to the prevalence of the
disease
NPV= Number of true negative results (TN) x100
All negative test results (TN+FN)
Number necessary to treat
NNT is how many patients must receive a
treatment to produce one additional
improved outcome compared to control.
The lower the NNT, the more effective the
treatment.
– NNT: 1/ARR
– ARR: Event rate w/o tx-event rate w tx.
– RRR: Event rate w/o tx/event rate w tx.
Confidence Intervals
Provides an interval that is likely to capture the
population mean with a level of confidence.
95% CI: Indicates that if a test were repeated
100 times, a result within the specified range of
values would be expected 95% of the time. In
studies is presented as mean minus and plus
two standard deviations: “4.5 ( 95% CI, 3.8 to
5.4).
Larger studies typically have narrower CI’s.
Board-type Questions
PSA has a sensitivity of 75% and
specificity of 80%. Prevalence of prostatic
carcinoma in your referral male population
is 10%. If your patient has positive result
on the blood test, what is the chance that
he has prostate carcinoma?
– What they are asking for?
– Calculate TP, FP, FN, TN then PPV, NPV.
Board-type Questions
A 40 years old woman wants to have a
mammogram to make sure that she does
not have breast carcinoma because the
prevalence of breast cancer in her
population is known to be 20 in 1000.
Sensitivity of mammogram test is 70% and
specificity is 90%. If her mammogram is
positive, what is the likelihood that she has
breast cancer?
Board-type Questions
An HIV pregnant patient has 2% risk of
transmitting infection to the baby if
delivered with CS compared to 7% risk if
delivered vaginally. How many c-sections
you would have to do to prevent one HIV
infection?
Board-type Question
Rates of re-admission for HF diminished
from 46% to 41% after close
multidisciplinary monitoring program
initiated upon discharge.
– What is the number necessary to monitor to
prevent one re-admission?
– What is the sample size required to give
power to a study? (Rule of 3).
– What could be the number to start with? Rule
of 3.