Definition of EBM

• The conscientious and judicious use of current best evidence from clinical care research inthe management of individual patients

Transdisciplinary Model of 

Evidence Based Practices

• clinical decision making should include some best available research evidence; patients characteristics, needs, values, preferences; and resources including providers expertise
• all of these are under the umbrella of environment and organizational context

Dimensions of Any Benefit or Harm

• nature(quality, intensity, time course)
• probability it will occur
• importance to the person
• how can benefits be maximized and harms minimized

New Evidence

• 13,000+ known diseases
• 2,000+ new MEDLINE articles each day
• half life for clinical research is 5 1/2 years(the time it takes for a clinically important change to take place)

5 Steps of EBM

• formulate an answerable question (EBM II)
• Track down the best evidence (EBM I)
• critically appraise the evidence for validity, clinical relevance and applicability (EBM!
• Individualize, based on clinical expertise and patient concerns (EBMII and beyond)
• Evaluate your own performance

Problems with randomized trials

• bias of researcher
• use of control groups
• placebo, current treatment, or biased comparison

Problems with meta-analysis

• missing negative trials
• cochrane group

Criticisms of EBM

• experimental evidence trumps other evidence
• assumes researcher is unbiased
• excludes important information
• each patient is an individual (too many uncommon diseases and varients)
• could limit individual choice

Types of Experimental Studies

• randomized controlled trial (RCT)
• nonrandomized controlled trial

Types of Observational Studies

(non-experimental)

• analytical(comparison group): cohort studies, case control studies, and cross-sectional studies
• descriptive (no comparison group): case series, qualitative studies, cross-sectional studies

Randomized Controlled Trials

• experimental group recieves an intervention
• control group receives no intervention(sometimes they receive usual care or a placebo)
• subjects are allocated at random
• outcomes are measured in both groups adn compared
• experimental treatment is blinded to subjects and clinicians(doubleblinded) to reduce bias
• intention to treat analysis provides conservative estimates

Advantages and Disadvantages of 

Randomized Controlled Trials

• advantages: researcher has control over the experimental variable; way to control for extraneous factors;causation is easier to prove;short duration of study
• desadvantages: artificial environment; problems in generalizability-whou volunteers for studies?

Superiority Vs Equivalence or

Noninferiority Trials

• already an effective treatment and want to show equivalence
• easier to prove equivalence than superiority
• are results valid of did the researcher use poor reasearch methods: easier to find no diff than a significant one, blinding is no longer as effective and intention to treat is not as effective

Cohort Study

(Prospective and longitudinal)

• advantages: only way to establish incidence of a disease, able to look at several risk factors, not limited to 1 disease
• disadvantages: not practical for rare diseases, expensive-many subjects over time, takes a long time to get results
• population sample-> exposed or not exposed

The Framingham Heart Study

• 5,209 men and women, ages of 30 to 62
• framingham, massachusetts in 1948
• physical examinations and lifestyle interviews every two years
• 1971, second generation enrolled
• risk factors are age, total cholesterol, HDL, blood pressure, smoking status and diabetes

Case Control Study

• cases(diseased) and controls (non-diseased)
• advantages: way to study rare diseases, short duration of study, study multiple causes of disease
• disadvantages: recall bias b/c retrospective, limited to study of 1 disease

Observational Studies

• descriptive studies: describethe distribution of disease or risk factors for disease, prevalence studies
• cross-sectional: one point in time, difficult to determine causality
• case series studies
• qualitative studies: geater understanding of context and values

Tertiary Resources

• reference books
• web resources 
• databases

Secondary Resources

• PubMed
• OVID
• Review Articles

Primary Literature

• research journal articles
• thesises or dissertations
• conferences
• reports
• patents

Why use Tertiary Resources?

• general information
• learn a new subject or refresh your memory
• \background information about a disease or treatment
• looking for gold-standard diagnosis or treatment algorithm for a disease or condition

Disadvantages to Tertiary Resources

• information can become dated during the publication process
• information may be incomplete
• errors in transcription
• author bias
• lack of expertise of authors
• incorrect interpretation of information by authors

Evaluating Tertiary Resources

• authors experience/expertise
• publication date
• references 
• relevant information
• bias or blatant errors

Drug Monographs

• Lexi-comp online
• micromedex
• access pharmacy
• accessmedicine
• all except accessmedicine have herbal monographs too

MeSH

• Medical Subject Headings
• on pubmed
• lets you search concepts instead of keywords
• generic name of a drug is always the MeSH term
• drugs that have not received a MeSH are designated as a substance name
• can build to searches
• refine your search by pub date, article type, language, gender/age, full text, journal subsets

PP-ICONS

• you use it to create your clinical question
• establish your patient, population or problem, desired intervention, comparison and desired outcome
• makes it easy to evaluate the article to your needs

PubMed Versus Google Scholar

• pubmed: more limits to narrow searches, can use MeSH, search history available to user, can use truncation, save results to my NCBI
• google scholar: sorts search results by relevance, searches full text  of article for keywords when full-text article is available

A study compared 150 adults with a rare disease to 300 disease free adults to examine past experiences that may have contributed to the development of the illness. what study design did they use?

-Cohort

-controlled clinical trial

-case-control

-case series

-perspective

case-control

Simple Random Sample

• each element has an equal probability of being selected

Stratified random sample

• divide patients into males and females and select 10% of each gender
• ensures that both men and women are represented

Cluster Sample

• select 10 anticoagulation clinics in NE OH and then select 50 patients from each clinic

Systematic Sample

• select every 5th patient that walks through the door at your clinic?

Probability

• p=probability of an event and q=1-p(probability of event not occuring)
• adition rule where probability of any 1 of several events occuring is equal to the sum of their individual probabilities, if they are independent events
• multiplication rule where probability of two or more statistically independent events all occuring is equal to the product of their individual probabilities

Types of Data

• nominal(male or female--cannot be ordered)
• ordinal(likert scales, strongly agree to strongly disagree-can be ordered)
• interval(has meaningful intervals btw categories
• ratio(age-has an absolute 0, rations are possible
• discrete(# of patients in a clinic-counts, no fractions)
• continuous(age-infinite number of values)

Measures of Central Tendency

• mode: observed value that occurs most often, best for describing nominal or ordinal data
• median: the value in the middle of the distribution, 50th percentile, best for describing ordinal or interval/ratio level variables
• mean: the average, the sum of all the elements divided by the number of elements
• μ= population mean, Χ= sample mean

Measure of Variability

• range is the difference btw the lowest and highest scores
• variance is the mean of the squares of all the deviation scores in the distribution

Prevalence vs. Incidence

• prevalence refers to the number of people with the diseases at a given time
• incidence refrs to the number of new cases of a diseases within a certain time period

Relative Risk

• a/(a+b)/c/(c+d)

Relative risks and Odds ratios

• less than 1 indicates a protective effect
• a value of 1 indicates no effect
• a value over 1 indicates a risk factor
• if a 95% confidence interval crosses 1, then no effect, either positive or negative

Absolute Risk Reduction

• what is the difference in incidence rates btw those exposed and those not exposed?

Relative Risk

• how many times more likely are exposed persons to become diseased, relatvie to non-exposed persons

Number needed to Treat

• NNT: number of people who would need to be treated with the experimental treatment in order to prevent one bad outcome
• NNH: number needed to harm-number of people needed to treat in order to cause one person harm

Attributable Risk

• what is the incidence of disease that is attributable to an exposure?
• AR=Incidence_exposed-incidence_nonexposed
• example 28% of smokers develop lung cancer and 6% of nonsmokers develop lung cancer so only 22% is the risk of lung cancer attributable to smoker

Validity-Is there bias?

• validity- did the study or test measure what it claimed to test? how accurate is the study or test? is there bias?
• internal-are the results of the study valid for the population studied?
• external-are the results of the study valid for the larger population? are they genralizable?

Reliability

• how precise are the results?
• do you get similar results if you measure more than once? is the study or text precise in the measurement
• test/retest reliability
• repeatability and reproducibility
• precision of measure

Journal Article Evaluation

• relevancy to your question
• authors
• publication date
• abstract

United States Preventive Services Task Force

(USPSTF)

• conducts sytematic reviews of the evidence
• estamates the magnitude of benefits and harms 
• makes recommendations
• grades strength of evidence: A(strongly recommend), B(recommends), C (no recommendation), D(recommends against), I(insufficient evidence to recommend for or against)

Screening Tests

• identify individuals with a disease before the presence of symptoms
• early treatment can save lives
• increases the number of false positives
• screening tests should be sensitve and specific, acceptable to the target population, have minimal risks, used when early interventions can reduce morbidity and mortality

The Cage: Example of a diagnostic questionnaire

• Cut down
• annoyed
• guilty
• eye opening
• this all is for drinking

Basic Structure of Diagnostic Studies

• series of patients
• index test
• reference(gold) standard
• compare the results of the index test with the reference standard, blinded

Sensitivity (Sn) and Specificity(Sp)

• sensitivity refers to the proportion of persons who truly have the disease who have a positive test result a/a+c
• specificity refers to the proportion of persons who dont have the disease who have a negative test result d/b+d

Positive Predictive Value

(PPV)

• refers to the probability that a person who tests positive really has the disease
• a/a+b

Negative Predictive Value

(NPV)

• refers to the probability that a person who tests negative is really disease free
• d/C+D

Relationship btw PPV and NPV

and Sn and SP

• they are inversely related: as PPV goes up NPV goes down
• as Sn goues up then Sp goes down
• vice versa is same for both

Physiological parameters

• appers to be a graph
• tends to have to curves that stand for people who have the disease and dont have the disease according to a test
• cross over for false positives and negatives

ROC Curves

• method of finding the best cut point
• method of comparing tests
• test A is best, greater area under the curve
• best cut point is at top of curve

Parallel Testing

• multiple tests are done and a positve result in any one of the tests is considered positive for a disease
• increases the sensitivity-fewer false negatives
• used when rapid assessment is necessary
• used to rule out disease (SnOut)

Serial Testing

• given a positive result on one test, another test is conducted
• increases the specificity-fewer false positives
• useful when no one test is high in specificity
• used to rule in disease (SpIn)
• used to rule in breast cancer
• causes fewer false positives and more false negatives because the mammogram has to be positive to continue

Likelihood Ratios (LR)

• probability of a particular test result for a  prson with the disease divided by the probability of that test result for a person without the disease
• probability of a positive test result: LR+= sensitivity/1-Specificity)
• probability of a negative test result: LR-=(1-sensitivity)/specificity
• if LR=1 the test is of no use in dividing those with and without
• if LR+ >1 then there is a higher probability that those with the disease will score at that particular level than those without the disease
• IF LR- <1 then there is a lower probability that those with the disease have a negative result than those without the disease

Where Would I find a Systematic Review/Meta-Analyses

• The cochrane Database of Systematic reviews
• the database of abstracts of reviews of effect
• PubMEd
• journals with meta-analysis are JAMA, lancet, chest, New England journal of medicine

Why Do a Meta-Analysis?

• increase statistical power for primary end-points and subgroups
• resolve uncertainty when reports disagree
• improve estimates of effect size
• answer questions not posed at the start of individual trials
• chance, true effect, bias, error effect it

Forest Plot

• also known as a blobbogram
• quickly visualize the results of individual stidies and possibly for pooled data
• 1 forest plot per outcome
• cochrane collaboration
• quick test for heterogeneity

L'Abbe Plot

• quickly shows the amount of contribution of individual studies to the outcome
• sample size is proportional to circle

Cochran's Q Test

• traditional test for testing heterogeneity
• generates a probability(X² distribution)
• P<0.10-heterogeneity present
• if p>0.10 but the ratio of cochran's Q/df>1 means possible heterogeneity
• cochran's Q/df<1-no heterogeneity
• can be underpowered or overpowered
• low poper-> possible type II error

I² Statistic

• percentage variability in results across studies due to differences in treatment effect and not chance
• I² =0% - chance explains variability present
• I²< 0.25 - low heterogeneity
• 0.25 to 0.5 - moderate heterogeneity
• 0.5 < - high heterogeneity

Sensitivity Testing

• do the decisions made while doing the systematic review have a great effect on the results of the systematic review
• robustness of results
• "file drawer effect"

Etiology of Disease

• what causes a disease?
• multiple predictors of a disease
• study of causation
• what predicts positive health behaviors?
• what predicts positive health outcomes?

Logistic Regression

• predicting a binary(dichotomous) outcome
• predicts the probability of the outcome variable
• examples: what predicts a 10% weight loss?
• regression coefficients can be transformed into odds ratios
• the odds of the occurrence of an event

Bonferroni Method

• adjustment to p values when several statistical tests are being performed
• if p is set to 0.05 then 1 time out of 20 we will get a false positive result
• we have 6 independent variables and 15 possible interactions between those 6 variables

Multilevel Modeling

• we surveyed care managers at area agencies on aging
• attitudes toward advance care planning
• each case manager responded to 3 hypothetical vignettes:

Descriptive Statistics

• serves to organize, summarize and describe data

Inferential Statistics

• serves to make inferences or generalizations about a total set of individuals or events on the basis of data from a smaller group

Hypothesis Testing

• state null(H0) and alternate(H1) hypotheses
• choose a significane levl, α(0.05, 0.01,0.001)
• determine the critical region and the non-rejection region, based on the sampling distribution
• make a decision and state the conclusion

Type I Error

• incorrect rejection of a true null hypothesis
• a false positive
• usually concludes that a relationship exists when it doesnt
• aplha
• the smaller the value the less chance that we are making this error

Type II Error

• failure to reject a false null hypothesis
• it is a false negative
• example would be a blood test failing to detect the disease it was designed to detect
• beta

Empirical Rule

• 68% of the distribution lies one standard deviation from mean
• 95% lies 2 standard deviations
• 99.7 lies 3 standard deviations
• this is in a normal distribution

Central Limit Theorem

• distribution of means from samples will always tend to be normal
• closer to normal as number of samples increases
• mean of these means is equal to the population mean
• standard deviation of this distribution is the standard error of the mean

The Z-Score

• estimate of the distance of a value from the population mean(number of std deveations)
• difference btw sample and population/standard error of the mean
• want to find if its less greater or inbtw

Critcal Values of Z

Variability

• variability means that the patterns we notice may be illusory and we may miss the real patterns
• variability means that our conclusions will always be tentative
• variability means that we'll have to deal with uncertainty

Power

• 1-β
• most studies aim for 80% power(at least 80% of the time you will find a sginificant difference if there is in fact a difference
• how many people do you need depends on your expected difference(how small), depends on the variation in measurements, on the sig level, sample size, and on one or two-tailed test

Two additional commonly used tests 

for mean differences

• T-test(n less than 30)-one samples, two samples, paired(before and after)
• ANOVA(three or more groups

Paired T-Test

• allows us to analyze one group at two different time points
• was the intervention effective
• difference of means within the individual
• difference btw 2 dependent means
• 2 observations per person
• formulas are slightly different

ANOVA

• used when longitudinal data are collected
• each person has repeated measures
• controls for dependence of measures
• similar to paired t-test

Parametric tests

• normally distributed dependent variable
• ratio or interval level data
• measures are independent
• usually used to examine means(sometimes difference in variance)

Nonparametric Tests

• make no assumptions about normal distribution
• more conservative than parametric tests
• used for analysis of medians
• used for ranked data
• used for ordinal data

Wilcoxon Rank sum Test

• nonparametric test
• when the assumptions for a t-test do not hold
• also called mann-whitney u test
• ranks scores from lowest to highest
• ranks are analyzed as though they were original observations
• null hypothesis: means of the ranks are equal

Chi-Squared

• nonparametric proportions
• do the proportions of observations falling in different categories differ significantly from the proportions that would be expected by chance
• test statistic: X² calc
• use of contingency tables
• X² calc=Σ[(O-E)²/E]

Correlation-Pearson's r

• measure of bivariate association(direction and strength relationship)
• no association=0
• perfect positive association=1
• perfect negative association=-1
• association does not mean causation
• cannot claim causation in a cross sectional study, need longitudinal data
• scattergram of no association r=0, dots evenly spaced everywhere 

Curvilinear Relationship

• this is a pearsons r relationship
• where the scattergram goes up then comes back down like a curve(arch)

R², The coefficient of Determination

• always between 0 and 1
• percent of variation in one variable explained by a change in the second variable
• high R² means one variable can be used to predict the other variable
• for example if there was a correlation of .50 between class attendance and test scores, then 25% of the variance in test scores would be predicted by class attendance

Ordinary least squares(OLS)

Regression Line

• line of regression with weird error bars