Presented by John Ruggiero, MPA, PhD Vice President, Education and Outcomes Institute for Continuing Healthcare Education Philadelphia, PA Adjunct Professor.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Presented byJohn Ruggiero, MPA, PhD
Vice President, Education and OutcomesInstitute for Continuing Healthcare Education
Philadelphia, PA
Adjunct Professor of Graduate BiostatisticsDrexel University
College of Nursing and Health ProfessionsPhiladelphia, PA
John Ruggiero has no interest in selling technology, a program, product, and/or service to CME professionals. There are no
financial disclosures to report.
STATISTICS?! WHY SHOULD I BE INTERESTED IN THAT? Statistics create order from chaos Statistics empower one to consider and
complete the larger picture Statistics help us become better citizens Statistics create outcomes—historical
reports that evoke decisions
A BIT OF HISTORY
THE REALITY
Every person in an organization should understand his/her individual and organizational expected goal(s) for success.
The information should be used to measure and improve effectiveness.
LEARNING OBJECTIVES
At the end of this lecture, learners should be
able to:
1. Generally explain the terminology used with statistics
2. Analyze the information presented3. Discern the relevant information
from the irrelevant information
KEY COMPONENTS TO A STUDY Population ( µ ) v. Sample ( x ) Collection of Data Independent variable v. Dependent
variableExample: Correlation between education
and a commitment-to-change performance
DESIGNING A STUDY
DESCRIPTIVE STATISTICSDescribing a situation –
The collection of data occurs before the analysis
INFERENTIAL STATISTICSHypotheses describe a situation – The researcher makes educated
guesses, collects the data and then analyses whether or not the hypotheses were correct
SAMPLING: HOW TO GET THE DATA Random Sampling
Samples are chosen without rhyme or reason Systematic Sampling
Samples are chosen by every kth number Stratified Sampling
Samples are divided into groups and then randomly chosen from those groups
Clustered SamplingSamples are chosen from a specific cluster for
purposes of the study design
MAKING OBSERVATIONS Collect the data
○ Rating learning objectives○ Rating faculty○ Was the education fair and balanced?
Measure your Central Tendency and DispersionMean, Median, and ModeStandard Deviation
EXAMPLE 1
Using a scale of 1-4 (ORDINAL DATA):1. Learning Objective was not met
2. Learning Objective was partially met
3. Learning Objective was met
4. Learning Objective exceeded expectations
The mean of learning objective 1 is collected among 10 learners from a small regional dinner activity
EXAMPLE 1 ANALYSIS Using a scale of 1-4 (ORDINAL DATA):
1. Learning Objective was not met
2. Learning Objective was partially met
3. Learning Objective was met
4. Learning Objective exceeded expectations
At a mean of “3”, learners from this activity believed that Learning Objective 1 was met. The margin of error is (±1.247).
OBJ 1
0
5
10
15
20
1 2 3 4
OBJ 1
70%
10%
0%
20%Physicians
Nurses
Pharmacists
Other
HISTOGRAMS
A histogram, or normal distribution curve, is used to graphically represent the normalcy of the mean related to all other data values from a study
ENGLISH!
Most of your data centers around the
mean
Extraneous data falls here
Histogram on Practice Change
02468
1012
1 2 3 4
Bins
Series1
DOSE OF REALITY
Pretty charts, animated graphs, and clean presentations don’t mean a thing, unless…You measure the results against a previous
educated guessThe study can be repeated
HYPOTHESES:THE EDUCATED GUESS Alternate v. Null
H1 (Alternate)○ The research hypothesis○ An observed effect is genuine – there is a
definite change
H0 (Null)○ There is no change to the study
HYPOTHESES EXAMPLE (Figures are fictitious). Let’s assume that the
national mean for victims of domestic violence is reported at 7%. This can be assumed because a cluster sample of 1500 people who had entered a medical facility emergency room in the past 12 months was completed.
My educated guess is that 7% is too low. I therefore believe that after educating targeted emergency room medical staff, and agreeing that every patient (regardless of visitation cause) is directly asked if they are a victim of domestic violence, the national mean will increase.
HYPOTHESES EXAMPLE Cont’d H0 (Null)
µ = 7%
H1 (Alternate)µ ≠ 7%
This is a two-tailed testRule of thumb: When hypotheses are written
with equality statements (= to) a two-tailed test can be assumed. When hypotheses are written with inequality statements (>,<) a one-tailed test can be assumed.
HYPOTHESES TESTING Declare an alpha (α) level of significance
○ Usually .01, .05 or .10○ This becomes known as the Critical Value○ Critical Value is compared to the z table. z-score
is then identified Recognize the p-value
○ The probability of getting values of the test statistic as extreme, or more extreme than, that observed if the null is true.
Statistical Significance○ If the p-value is less than the alpha, or when
completing a test value, the value falls within the Critical Value (beyond the z-score), one rejects the null hypothesis
P-value is = .05 while the observed value is 1.645. If the result is greater than 1.645, or in the critical region (5%), then you reject the null.
RELATING HYPOTHESES TO THE CME INDUSTRY
How could you use alternate and null hypotheses to assist you with one of your CME initiatives?
CONFIDENCE INTERVALS
Can the study be repeated at a specific level of confidence?Statisticians will usually choose either 99%,
95%, or 90% confidence percentages○ Example: If you claim 95% confidence with
your results, you are basically saying that no matter how many times you repeat a study, 95% of the time the mean and all other results will be similar.
Why this is important for CME
CONFIDENCE INTERVAL TESTS CI Test of the Means
○ (n > or = 30)
T-test ○ (n < or = 29)
CORRELATION & REGRESSION Correlation is the association between
two quantitative variables Association is linear The correlation coefficient is measured
on a scale that varies from + 1 to -1. Symbol for correlation is r
Correlation Graphs
Le, C.T. (2001).
Correlation Example
Le, C.T. (2001).
Regression (The Trend)
Ruggiero, J. based on Le, C.T. (2001).
LOOK FAMILIAR?
Survival Curves Survival curves illustrate prognosis. The
percentage of patients reaching an endpoint (for example: death, recurrence of disease, or cure) is plotted on the y (vertical) axis against time on the x (horizontal) axis.
The Kaplan-Meier method is preferred unless there is an extremely large number of patients being studied