Health & Medical Question

Chapter 6: Practical Application of the
Principles of Epidemiology to Study
Design and Data Analysis
PHC 231: Introduction to Hospital Epidemiology
1
Overview – Making Sense of Data
• Bias – “any ‘error in the conception and design of a study—or in the
collection, analysis, interpretation, reporting, publication, or review
of data—leading to results that are systematically (as opposed to
randomly) different from truth.’” (Abramson, 2012, p. 95)
•
•
•
Information bias
Selection bias
Confounding
• How can we minimize bias?
•
•
Methods employed during study planning and conduction stages
Methods employed when handling study results
PHC 231: Introduction to Hospital Epidemiology
2
Descriptive Studies – Information Bias
• Information bias can be caused by:
•
•
•
•
•
Failure to define appropriate working definitions for variables
Any deficiencies in data collection
Missing data
Deficiencies in data recording or management
Changes that occur over time in definitions, case-notification systems, or casefinding methods (for longitudinal studies)
• Incomplete or inaccurate medical records
(Abramson, 2012, p. 97)
PHC 231: Introduction to Hospital Epidemiology
3
Information Bias, cont.
• Consider:
•
•
•
•
Study objectives – what knowledge is the study planning to yield?
Conceptual definition – dictionary definition
Working definition – definition expressed using method of examination
How valid was the measurement?
•
•
Compare findings with a criterion or “gold standard.”
How reliable was the measure?
•
•
High reliability does not mean the measurement has high validity
Low reliability is a sign of questionable validity
(Abramson, 2012, p. 96)
PHC 231: Introduction to Hospital Epidemiology
4
Selection Bias and Sampling Variation
• Selection bias can occur:
• When sampling is inappropriate or sample coverage is incomplete
•
Computer programs can compute sample size necessary for precise
results
• When subjects are lost
PHC 231: Introduction to Hospital Epidemiology
5
Planning a Descriptive Study
•
Methods to minimize bias during planning stage of descriptive studies:
•
•
•
•
•
Clear operational definitions for variables and categories.
Valid measurement methods and employ in a standardized manner
Built-in quality-control measures
Data cleaning
Other considerations:
•
•
•
Use of a sample that is large enough and represents the population
Consideration of tracking procedures if study is longitudinal
Consideration of the usefulness of study results
(Abramson, 2012, p. 98)
PHC 231: Introduction to Hospital Epidemiology
6
Analysis of a Descriptive Study
•
•
•
•
•
Frequency distribution of each variable in the total study sample or its
subgroups is tabulated.
Rates or proportions are computed for a “yes-no” or other categorical
variables.
Measures of central tendency and dispersion are computed for metric
(noncategorical) variables.
One-sample significance tests can be used to compare the conformity of the
rate, proportion, mean, or median with an expected or hypothetical value.
Two sample significance tests can be used to make comparisons with
findings found elsewhere
(Abramson, 2012, p. 99)
PHC 231: Introduction to Hospital Epidemiology
7
Analytical Observational Studies – Information Bias
•
Information bias can be caused by the same factors as in descriptive studies,
plus:
•
•
•
Differential misclassification – a difference in measurement validity between two groups
Diagnostic suspicion bias – occurs when “information about the disease comes from a
subject, interviewer, or examiner whose report about the presence of the disease is
colored by knowledge that there has been exposure to the risk factor and who is more
likely to report the disease if there has been exposure.” (Abramson, 2012, p. 101)
Exposure suspicion bias – occurs when “the information about the exposure comes from
a subject, interviewer, or examiner whose report about the presence of the exposure is
colored by knowledge of the presence of the disease.” (Abramson, 2012, p. 101)
PHC 231: Introduction to Hospital Epidemiology
8
Selection Bias and Sampling Variation
• Selection bias can be caused by the same factors as in descriptive
studies, in addition to the following special issues:
•
Case-control studies – many studies have inappropriate controls
•
•
It is difficult to find controls that are both convenient and without bias
Cohort studies – loss of members can cause selection bias
PHC 231: Introduction to Hospital Epidemiology
9
Confounding
•
Confounding effects/bias – “the distortion of an association between
variables by the influence of another variable.” (Abramson, 2012, p. 102)
•
Extraneous variables can have a confounding effect on an independent variable A and a
dependent variable B if:
•
•
•
•
It influences B and
It is associated with A in the population, but not because it is affected or caused by A
Confounder must have strong associations with A and B for the confounding
effect to be important
Methods to control for confounding include:
•
Restriction of study to a homogenous group, matching, stratification,
neutralization of the confounder by a statistical technique, use of a propensity
score, and use of an instrumental variable
(Abramson, 2012, p. 102)
PHC 231: Introduction to Hospital Epidemiology
10
Modifying Effects and Intermediate Causes
• Effect modifier – occurs when an association between some factor
and an exposure outcome is stronger depending on a third variable
(e.g., sex)
• Practical implications:
•
•
Identification of high-risk groups
Development of new healthcare procedures
• Stratification is used to detect and measure effect modification
(Abramson, 2012, p. 103)
PHC 231: Introduction to Hospital Epidemiology
11
Planning an Observational Study
•
•
•
Observation study planning is guided by the same factors as descriptive
study planning, in addition to some other special issues.
It is advisable to blind anyone involved in the study that could influence
findings.
Case-control studies:
•
•
•
Advantages: easier, faster, less expensive than cohort studies; require smaller samples
for uncommon outcomes
Disadvantages: address only one outcome; provide no direct measures of risk; and
subject to recall bias and exposure suspicion bias;
Cohort studies
•
Disadvantages: subject to diagnostic suspicion bias and follow-up bias (Abramson, 2012, p. 103)
PHC 231: Introduction to Hospital Epidemiology
12
Analysis of an Observational Study
•
Exploration of associations between variables
•
•
•
•
•
•
•
Examination of frequency distributions of all relevant variables
Identification of gaps, patterns, and inconsistencies
Rates of proportions computed for “yes-no” or other categorical variables
Bivariate analyses – the examination of relevant associations between pairs
of variables
Multivariate analyses – examine associations that involve more than two
variables
Determination of statistical significance and estimates of confidence
Interpretation
PHC 231: Introduction to Hospital Epidemiology
13
Interpretation
•
•
•
Is an observed association a true one? Is it an artifact caused by some form
of bias? Is an absence of associations caused by bias?
Address the possibility of other factors that might result in misleading
associations
Strength of association
•
•
Stronger associations are more important
Strength appraisals should be done after addressing the possibility of confounding and
effect modification.
• Statistical significance – perform significance tests when needed
PHC 231: Introduction to Hospital Epidemiology
14
Interpretation, cont.
• Casual inferences – for an association to be considered causal, two
criteria must be met:
•
•
The assumed cause must precede the assumed outcome.
The observed association must not be wholly attributable to selection bias.
• Additional criteria:
•
•
•
Statistical significance
Strength of the association
Dose-response relationship – correlation between the amount, intensity, or
duration of exposure to the “cause” and the amount or severity of the “effect”
PHC 231: Introduction to Hospital Epidemiology
15
Interpretation, cont.
•
•
•
•
•
Time-response relationship – if the incidence of the “effect” peaks some time after
a brief exposure to a “cause” and then decreases, this supports the case.
Predictive performance – if the study results provide new knowledge supporting
an a priori hypothesis concerning a predicted effort, this supports the case
Specificity
Consistency
Coherence
PHC 231: Introduction to Hospital Epidemiology
16
Ecological and Multilevel Studies
• Ecological studies – studies in which subjects are groups or
populations, not individuals
•
Advantages: researchers can look at factors that affect an entire group, or a
particular floor or ward within a hospital
• Multilevel studies – studies that use group-based and individualbased variables and examine associations of effects
•
Example: study that examined hospital-associated infections in 60 wards in a
Finnish hospital. Some findings were that infections were higher when staff
worked longer hours, experienced more stress, and had poor team relations.
(Abramson, 2012, p. 107-108)
PHC 231: Introduction to Hospital Epidemiology
17
Program Reviews
• Program reviews – studies that evaluate specific healthcare
programs
•
•
•
•
Descriptive epidemiological studies
Concerns: welfare of the patients, community, or population receiving care;
healthcare program outcomes
Not concerned with: assumptions on which the healthcare program under
review is based
Areas of consideration: program requisiteness, program outcomes, process,
structure, efficiency, and differential values
(Abramson, 2012, p. 108-109)
PHC 231: Introduction to Hospital Epidemiology
18
Trials
• When possible, trial findings should be expressed in terms of impact
• Clinical trials – “evaluate therapeutic, preventive, rehabilitative, or
educational procedures applied to individuals” (Abramson, 2012, p. 109)
•
•
•
Usually parallel studies
Either externally controlled or self-controlled
Avoid selection bias, confounding, and information bias (if well-designed)
• Program trials – evaluate group-level programs
•
Difficulties in planning to avoid bias – researchers often don’t have influence
over a decision to run a program, and how to choose cases and controls
(Abramson, 2012, p. 110)
PHC 231: Introduction to Hospital Epidemiology
19
Practical Implications
• Study findings may be a catalyst for action
• Whose role is it to recommend action?
•
•
Most epidemiologists feel a responsibility to recommend action, and to
improve health or healthcare through their work.
Alternate view: Health policy-making requires different skills and experience
than epidemiology and should be done by experts in that field.
(Abramson, 2012, p.
PHC 231: Introduction to Hospital Epidemiology
20
References
Abramson, J.H. (2012) Practical Application of the Principles of
Epidemiology to Study Design and Data Analysis. In C.G. Mayhall
(Ed.), Hospital Epidemiology and Infection Control (pp. 95-113).
Philadelphia: Wolters Kluwer Health/Lippincott Williams & Wilkins.
PHC 231: Introduction to Hospital Epidemiology
21
College of Health Sciences
Department of Public Health
Summer course
PAPER ASSIGNMENT COVER SHEET
Course name:
Introduction to Hospital Epidemiology
Course Code:
PHC 231
CRN:
50519
Assignment title or task:
(You can write a question)
Please prepare yourself by the following:
• Read Chapters 6 in Mayhall (2012)
• View video available on BB
• Read other resources available on BB
Then:
Write down and discuss the Practical Application of the
Principles of Epidemiology to Study Design and Data
Analysis
Focus on the following:
– Sence of data
– Descriptive epidemiology
– Bias and Sampling Variation
– Planning analysis and confounding factors
• Notice : References are essential
Student name:
Student ID:
Submission date:
Instructor name:
Dr.Ahmed Salih
Grade:
……. Out of 10
Instructions for submission:
•
•
•
•
•
•
•
Assignment must be submitted with properly filled cover sheet (Name, ID, CRN,
Submission date) in word document, Pdf is not accepted.
Length of the write-up should be 200-500 words.
Text size 12-Times New Roman with 1.5-line spacing.
Heading should be Bold
The text color should be Black
Do proper paraphrasing to avoid plagiarism with proper references/sources.
References must be in APA format

Purchase answer to see full
attachment