Comprehensive Study Guide on Evidence-Based Practice and Descriptive Statistics
Fundamentals of Evidence-Based Practice (EBP)
Definition of Evidence-Based Practice (EBP):
EBP is defined as the conscientious, explicit, and judicious use of current best evidence in making decisions about the care of individual patients (Brown, 2023, p.3).
It involves questioning daily clinical activities as healthcare providers and grounding clinical decisions in scientific research.
Evidence-based practitioners are reflective, avoid making clinical decisions based on hunches, and can articulate what interventions they are providing and why.
The Three Pillars of EBP:
External Scientific Evidence:
Derived primarily from research articles published in peer-reviewed scientific journals.
Provides objective information to address clinical problems.
Developed using the scientific method to minimize bias through controlled study designs, objective data collection, and rigorous statistical analysis.
Science is inherently speculative, self-correcting, and subject to change; replication of research is essential. The results of a single study should never be interpreted as definitive proof of an intervention's efficacy, but rather as a starting point.
Practitioner Experience and Expertise:
Clinical expertise is complementary to and equally important as research evidence.
Early EBP models were criticized for ignoring practitioner experience. Current models recognize that research can never keep up with clinical practice or address every daily clinical scenario.
Practitioner experience is essential when scientific evidence is limited (e.g., only one or two existing studies).
Reflective practitioners consider clinical outcomes with clients, analyze what occurred, and apply theories and frames of reference to explain intervention mechanisms or design new interventions.
Reflection becomes structured and explicit through program evaluation methods.
Client Situation, Values, and Preferences:
Client-centered practice views the client as the expert in their own life and respects client autonomy.
A client-centered approach integrates the values, preferences, and experiences of the client, family, and caregivers alongside external evidence and therapist expertise.
Shared decision-making involves presenting research findings and clinical expertise to the client, while the client shares their values and lived experience. Differing viewpoints are discussed to reach a mutual agreement on the intervention plan.
Shared decision-making enhances client knowledge, confidence, and adherence to recommendations and home programs. Increased internet access has significantly expanded client participation in this collaborative process.
Ethical and Professional Practice Justifications:
Ethical Mandate: The American Occupational Therapy Association (AOTA) Occupational Therapy Code of Ethics and Ethics Standards addresses beneficence by mandating that practitioners use, to the extent possible, evaluation, planning, intervention techniques, assessments, and therapeutic equipment that are evidence-based, current, and within the recognized scope of practice.
Interdisciplinary and Stakeholder Influence: Grounding clinical decisions in scientific evidence carries greater weight and credibility when communicating with interdisciplinary team members and external stakeholders.
Professional Credibility and Reimbursement: The use of evidence increases confidence among professional colleagues and clients, and third-party payers are more likely to reimburse evidence-backed treatments.
Quality of Care: EBP directly improves patient outcomes by combining scientific justification, clinical expertise, and client values.
The Five-Step Process of Evidence-Based Practice
Step 1: Formulate a Question Based on a Clinical Problem:
Identify the specific clinical issue and structure a question to establish parameters for searching the literature.
Questions generally fall into five primary categories:
Efficacy of an intervention
Usefulness of a assessment tool
Description of a medical condition
Prediction of an outcome (prognosis)
Lived experience of the client
Step 2: Identify Relevant Evidence:
Search for evidence across three sources: research literature, practice knowledge/clinical expertise, and client experience and values.
Step 3: Evaluate the Evidence:
Appraise gathered research to determine study strength, methodological quality, and validity of conclusions.
Key appraisal considerations include study design, sample size, outcome measures, and clinical applicability to the practitioner's specific client circumstances.
Step 4: Implement Useful Findings:
Synthesize, analyze, and present gathered evidence to the client.
Collaborate through shared decision-making to apply findings into clinical practice.
Step 5: Evaluate the Outcomes:
EBP is recursive, drawing upon its own outcomes to build practice knowledge.
Practitioners collect data to measure client outcomes, reflect on the experience, refine future clinical decision-making, and restart the cycle.
Formulation of Evidence-Based Questions and Design Hierarchies
Efficacy Questions and PICO Structure:
Designed to guide clinical decisions regarding intervention selection.
Formatted using the PICO structure:
P: Population or Client Problem
I: Intervention
C: Comparison or Control Condition
O: Outcome
The order of words is less critical than including the key components. PIO or PICO structures are utilized when specific approaches are being evaluated. Broad, generalized questions are used when exploring initial intervention options.
Levels of Evidence for Efficacy (Intervention) Research:
Establishing intervention efficacy requires demonstrating a cause-and-effect relationship, which demands high-level experimental designs.
Level I: Systematic Reviews of at least 2 Randomized Controlled Trials (RCTs).
Level II: Individual Randomized Controlled Trial (RCT).
Level III: Nonrandomized Controlled Trial (Quasiexperimental Study).
Level IV: One-group trial with pretest/posttest (Pre-experimental Study).
Level V: Case reports and expert opinion.
Methodological Requirements for Randomized Controlled Trials (RCTs):
An RCT (true experiment) must meet three essential criteria:
Must contain at least two groups: an experimental group and a control/comparison group (control groups may receive no intervention, standard care, or an alternative intervention).
Participants must be randomly assigned to groups.
An intervention must be actively manipulated and applied to the experimental group.
Caveat on Systematic Reviews: Systematic reviews summarizing multiple RCTs provide higher evidence than a single RCT. However, reviews that include lower-level designs are ranked lower. A conducted systematic review does not automatically mean an intervention is effective.
Questions Regarding Usefulness of Assessments:
Evaluated using psychometric methods to test reliability and validity:
Reliability: Examines score consistency and dependability. Tested via test-retest methods (repeated administration) or inter-rater reliability (varying evaluators).
Validity: Evaluates whether an assessment measures its intended construct. Tested by comparing the tool with similar established measures or assessing its ability to discriminate between diagnostic populations.
Sensitivity: The proportion of individuals correctly identified as possessing the condition.
Specificity: The proportion of individuals correctly identified as not having the condition.
Sensitivity of Change: The degree of precision required to detect meaningful clinical change over time.
Questions Describing a Condition:
Address conditions with limited existing information using nonexperimental, observational designs without variable manipulation.
Prevalence: The proportion of individuals within a specific population who have a condition at a designated point in time.
Incidence: The rate or risk of developing a new condition over a specified period.
Utilizes group comparisons of existing populations or survey methodology.
Questions Predicting an Outcome (Prognostic Questions):
Examine associations between factors or predictors and clinical outcomes without manipulating variables.
Participants are grouped by naturally occurring characteristics rather than random assignment.
Utilize correlational, observational, and survey data.
Evidence strength depends on whether data collection is longitudinal or cross-sectional, as well as total sample size.
Correlation indicates association, not causation.
Questions About Lived Experience:
Addressed via qualitative research to explore subjective meaning, human experiences, and underlying themes.
Uses open-ended, exploratory questions formatted using PEO statements (Person, Environment, Occupation).
Search Strategies, Database Systems, and Evidence Retrieval
Electronic Databases Overview:
Structured digital collections of catalogued scientific literature. Databases vary in scope, subject inclusion, and cataloguing criteria.
Most health databases consist of peer-reviewed journals. Search interfaces are often publicly accessible, whereas full-text retrieval may require institutional subscriptions.
Database Profiles:
PubMed: Electronic repository of the U.S. National Library of Medicine. Provides free access to MEDLINE and the Cochrane Database of Systematic Reviews. All indexed journals are peer-reviewed. Utilizes Medical Subject Headings (MeSH), a hierarchical tree-structured vocabulary system. Entering a MeSH term automatically searches synonyms and related descriptors. Keywords can be substituted when specific MeSH terms are unavailable.
CINAHL (Cumulative Index to Nursing and Allied Health Literature): Focuses specifically on nursing and allied health disciplines. Catalogs occupational therapy journals absent from PubMed. Uses its own top-of-page heading system. Not all CINAHL content is peer-reviewed.
Cochrane Database of Systematic Reviews: Part of the Cochrane Library. Specializes in high-quality systematic reviews evaluating efficacy and diagnostic questions. Abstract access is free; full text requires subscription. Unsuitable for qualitative or descriptive questions.
Google Scholar: Not recommended as a primary database for EBP because it lacks filtering options for specific study designs and alters result ranking algorithms.
Database Search Mechanics:
Phrase Searching: Enclose exact phrases in double quotation marks (e.g.,
"Toni Morrison") to restrict results to contiguous words.Truncation/Wildcard: Use an asterisk (
*) attached to a word root to retrieve variant endings.Boolean Operators:
AND: Yields results containing both search terms (narrows search).OR: Yields results containing either search term (broadens search).NOT: Excludes specific terms from the search.
Search Filters and Access:
Filters include full-text availability, publication date ranges, species, language, participant age groups, and study design.
Methods to access full text include institutional database links, direct journal searches, interlibrary loan (ILL), research librarians, and professional organizations.
Critical Appraisal, Source Credibility, and Publication Dynamics
Source Credibility Classification:
Primary Sources: Direct, original research reports, data collections, or official government statistics uninterpreted by third parties.
Secondary Sources: Documents that summarize, analyze, or interpret primary sources.
Website Evaluation: Web resources must be evaluated by verifying organizational authority (e.g., recognized professional associations versus commercial vendors), presence of editorial review boards, transparency in the 'About Us' section, and reliance on peer-reviewed citations over anecdotal testimonials.
Public Press and News Media: Mass media reporting frequently overemphasizes low-level evidence, overstates clinical effectiveness, and presents distorted summaries of scholarly studies.
Scholarly Publications and Quality Metrics:
Impact Factor: Calculation based on the annual average frequency with which articles from a specific journal are cited in other publications. Criticized for favoring journals publishing long review articles, hot research topics, or high self-citation rates.
Peer-Review Process: Editorial evaluation where independent field experts appraise submitted manuscripts for rigor. While providing quality control, peer review does not guarantee total accuracy or eliminate statistical errors or data fraud.
Biases in Publishing:
Research Funding Bias: Sponsorship by organizations with financial stakes (e.g., pharmaceutical or medical device manufacturers) correlates with higher rates of favorable findings. Studies sponsored by commercial entities frequently utilize ideal study conditions or weaker comparison controls.
Publication Bias: Academic journals demonstrate preference for publishing positive, statistically significant results over null findings, biasing published literature.
Duplicate Publication: The unethical practice of republishing identical dataset findings across multiple papers to give the appearance of independent replication.
Rapid Evidence Synthesis Tools:
Critically Appraised Topics (CATs): Synthesize and appraise multiple research studies regarding a specific clinical question. Less methodologically rigorous than systematic reviews, but designed for quick clinical application.
Critically Appraised Papers (CAPs): Succinct summaries of a single study that provide a structured framework evaluating methodological rigor to deliver the clinical 'bottom line.'
Structure and Components of Research Articles
Title: Accurately reflects study variables and scope.
Authorship: Authors are ordered by extent of contribution. The primary or lead author contributed most to writing and research execution. Affiliations and lead author contact details are provided.
Abstract: Summary of the study, subject to strict publisher word limits.
Introduction: Establishes problem background, clinical significance, and an up-to-date literature review combining current citations with classic seminal studies. Concludes with a explicit purpose statement, research questions, or hypotheses.
Methods: Detailed procedural account enabling exact replication. Specifies study design, participant selection, inclusion/exclusion criteria, informed consent, Institutional Review Board (IRB) approval, appraisal of assessment psychometrics, intervention protocols, and analytical statistical tests.
Results: Objective reporting of quantitative or qualitative findings in narrative, graphic, and tabular formats without editorial speculation.
Discussion: Synthesizes results, addresses research hypotheses, provides evidence-grounded interpretation, contextualizes findings within broader literature, acknowledges study limitations, and outlines clinical implications.
References & Acknowledgments: Full citations following specific style manuals (e.g., APA, AMA) and formal recognition of non-author contributors.
Taxonomy of Research Methods, Designs, and Variables
Classifications of Scientific Research:
Experimental vs. Nonexperimental: Experimental research manipulates independent variables to establish cause-and-effect; nonexperimental research observes naturally occurring phenomena without intervention.
Quantitative vs. Qualitative: Quantitative research collects numerical data to perform statistical hypothesis testing; qualitative research collects narrative data to analyze underlying meaning and themes using inductive reasoning.
Cross-Sectional vs. Longitudinal: Cross-sectional research collects data at a single time point across groups; longitudinal research collects data across multiple time points over extended durations (months, years, or decades).
Basic, Applied, and Translational Research:
Basic Research: Investigates fundamental theoretical questions to expand knowledge, often yielding serendipitous findings.
Applied Research: Directly investigates practical healthcare interventions and clinical problems.
Translational Research: Bridges basic laboratory science with applied clinical research ('bench-to-bedside').
Experimental Research Designs:
True Experimental (RCT - Level II): Random assignment to experimental or control groups with active manipulation of the independent variable.
Quasiexperimental (Nonrandomized Controlled Trial - Level III): Evaluates cause-and-effect using pre-existing groups without random assignment. Susceptible to selection bias and confounding participant attributes.
Pre-experimental (Pretest-Posttest Single Group - Level IV): Evaluates a single cohort before and after intervention without a control group. Offers weak evidence.
Quantitative Hypothesis Testing and Errors:
Null Hypothesis (): States that no statistically significant difference or relationship exists between variables.
Directional vs. Non-directional Hypotheses: Directional hypotheses specify the expected direction of effect; non-directional hypotheses state an expectation of difference without specifying direction.
Quantitative data can only support or fail to support a hypothesis; it can never prove or disprove it.
Alpha Level (-value): The decision threshold set at , meaning there is less than a probability that observed results occurred by chance.
Type I Error (False Positive, ): Occurs when the researcher accepts the research hypothesis (rejects ), but the hypothesis is actually false. Controlled by setting strict alpha levels.
Type II Error (False Negative, ): Occurs when the researcher rejects the research hypothesis (fails to reject ), but the hypothesis is actually true. Commonly caused by inadequate sample size.
Variables in Research:
Categorical Variable: Classifies cases into discrete, non-numeric categories (e.g., diagnostic labels).
Continuous Variable: Measured along an infinite numerical scale with equal incremental distances (e.g., blood pressure, weight).
Independent Variable (IV): The variable manipulated or used to group participants.
Dependent Variable (DV): The outcome variable measured to observe the effect of the IV.
Control Variable: Factors held constant throughout the study.
Extraneous Variable: Uncontrolled background factors that can distort findings and must be tracked or statistically controlled.
Descriptive and Predictive Nonexperimental Designs:
Epidemiological Research: Measures condition distribution within populations.
Incidence: Rate of new cases within a defined population during a specific time period.
Prevalence: Total number of existing cases in a population at a specific point in time.
Group Comparison Studies (Ex Post Facto): Cross-sectional observational designs comparing pre-existing clinical cohorts (e.g., individuals with a disability versus healthy controls) without IV manipulation.
Survey Methodology: Questionnaires administered via mail, electronic platforms, phone, or face-to-face. Advantages include low cost and broad sampling; disadvantages include low response rates, self-report inaccuracies, and response bias.
Correlational and Regression Predictive Research:
Evaluates relationships between variables without implying causation.
Multicollinearity: High correlation between predictor variables that inflates variance.
Multiple Linear Regression: Models multiple continuous predictor variables against a continuous outcome variable.
Multiple Logistic Regression: Models predictor variables against a categorical binary outcome variable, outputting odds ratios and confidence intervals.
Prognostic Observational Designs:
Case-Control Study: Retrospective, cross-sectional observational design that compares individuals with a condition (cases) to individuals without (controls) to identify past exposure risk factors.
Cohort Study: Prospective, longitudinal observational design that follows groups with and without a hypothesized risk factor over time to measure outcome development.
Hierarchy of Evidence for Descriptive/Predictive Studies:
Systematic Reviews of Prospective Cohort Studies
Individual Prospective Cohort Study
Retrospective Cohort Study
Case-Control Design
Expert Opinion, Case Study
Descriptive Statistics, Measurement Scales, and Data Visualization
Levels of Measurement:
Nominal Scale: Categorical data classified into mutually exclusive, exhaustive categories without hierarchical ordering (e.g., gender, injury category).
Ordinal Scale: Categorical data organized into a hierarchical rank order where numerical distances between ranks are unequal or unknown (e.g., Likert scales, functional independence levels, pain rating scales).
Continuous Scale: Numerical data with equal interval distances between scale points.
Interval Scale: Continuous scale lacking an absolute zero point (e.g., temperature in Fahrenheit).
Ratio Scale: Continuous scale possessing a true absolute zero, where zero indicates total absence of the measured variable (e.g., weight, height, volume).
Frequency Distributions and Graphic Representations:
Frequency () and Percentage (): Raw counts of occurrences paired with proportional distribution calculations.
Frequency Table: Tabular display organizing raw frequencies and relative percentages across discrete categories.
Bar Graph: Graphic display using separated bars to show frequencies or percentages across nominal or ordinal categorical data.
Histogram: Graphic display using contiguous adjacent bars to represent frequency distributions of continuous variables across score intervals along the x-axis.
Pie Chart: Circular graphic displaying relative proportional percentages of a categorical variable.
Line Graph: Graphic display connecting discrete data points with line segments to plot continuous variable trends over time.
Central Tendency, Variability, Normal Distribution, and Skewness
Measures of Central Tendency:
Identifies the single central score representing the midpoint of a data distribution.
Mean (): The arithmetic average calculated by summing all individual scores () and dividing by the total sample size ():
Uses every score in the distribution; highly sensitive to extreme score outliers.
Median: The exact middle score dividing a rank-ordered distribution into equal upper and lower halves ( percentile).
The preferred measure of central tendency when distributions contain extreme outliers or pronounced skewness.
Mode: The most frequently occurring score in a distribution.
The only measure of central tendency applicable to nominal categorical data.
Influenced by chance; cannot be utilized in inferential statistics.
Measures of Variability:
Characterizes the spread or dispersion of individual scores around the center.
Range: The arithmetic difference between the highest and lowest scores.
Standard Deviation (, , or ): The average distance that individual scores deviate from the distribution mean.
A small indicates tightly clustered scores around the mean; a large indicates wide variance.
If all scores are identical, .
Standard Error of the Mean (): An estimate of the standard deviation of sample means across an entire population. While describes dispersion within a sample, describes sampling error relative to the broader population.
Normal and Skewed Probability Distributions:
Normal Distribution: A theoretical, perfectly symmetrical bell-shaped curve where the mean, median, and mode are identical.
The Empirical Rule ( Rule): In a normal distribution, specific proportions of data fall within defined standard deviation intervals from the mean ():
of data fall within ( on each side of the mean).
of data fall within (adding per side).
of data fall within (adding per side, with in the outer tails).
Positively Skewed Distribution: Data tail extends to the right (positive direction) due to a cluster of low scores and a few high outliers. Score order along the x-axis: .
Negatively Skewed Distribution: Data tail extends to the left (negative direction) due to a cluster of high scores and a few low outliers. Score order along the x-axis: .
Bimodal Distribution: A distribution containing two distinct frequency peaks.
Comprehensive Review Exercises, Application Problems, and Answers
Class Exercise 4-2 (Length of Stay Data Analysis):
Dataset: (
Calculations:
Mode: (occurs 3 times).
Median: (the 8th rank-ordered score).
Mean:
Optimal Central Tendency Measure: The Median () is the best representation of central tendency because the extreme outlier () artificially elevates the mean to , which exceeds 13 of the 15 individual data points.
Class Exercise 4-4 (Distribution Shape Identification):
Scenario 1: Most students perform very well on an exam (receiving A's and B's) with very few low grades.
Answer: Negatively Skewed Distribution (tail extends toward the lower scores on the left).
Scenario 2: Most students receive C's, a smaller number receive B's and D's, and very few receive A's and F's.
Answer: Normal Distribution (symmetrical bell curve centered at C).
Scenario 3: Many students receive A's and C's, with fewer students receiving intermediate grades.
Answer: Bimodal Distribution (two distinct frequency peaks).
Review Question Set:
Question 1: What is the median for the following data set: ?
Options: A) 10.5 | B) 10 | C) 11 | D) 11.5
Answer: A) 10.5
Explanation: With scores, the median is the average of the 4th () and 5th () rank-ordered scores: \frac{10 + 11}{2} = 10.5$.\n\n * **Question 2:** When a study describes a sample as 62% female, this percentage is an example of what statistic?\n * *Options:* A) Mode | B) Range | C) Standard deviation | D) Frequency\n * **Answer:** **D) Frequency**\n * *Explanation:* Reporting raw counts or percentages of categorical data represents a frequency statistic.\n\n * **Question 3:** The mode is the most useful measure of central tendency when the data is:\n * *Options:* A) Skewed | B) Includes a few outliers | C) Is categorical | D) Is normally distributed\n * **Answer:** **C) Is categorical**\n * *Explanation:* The mode is the only measure of central tendency applicable to nominal categorical data.\n\n * **Question 4:** What type of graph is best to use when you want to illustrate changes over time?\n * *Options:* A) Bar graph | B) Histogram | C) Pie Chart | D) Line graph\n * **Answer:** **D) Line graph**\n * *Explanation:* Line graphs visually plot continuous variable movements and longitudinal trends across time intervals.\n\n * **Question 5:** In a normal distribution, the percentage of people that have scores within one standard deviation below and one standard deviation above the mean is:\n * *Options:* A) Approximately 68% | B) Approximately 34% | C) Approximately 27% | D) Approximately 96%\n * **Answer:** **A) Approximately 68%**\n * *Explanation:* According to the Empirical Rule, \bar{x} \pm 1s68\%34\%34\% above).\n\n * **Question 6:** In which of the following samples would you expect to find the largest standard deviation for a set of scores measuring cardiovascular endurance?\n * *Options:* A) A group of long-distance runners | B) A group of sedentary adults | C) A group of pickleball athletes | D) A group that includes sedentary and active adults\n * **Answer:** **D) A group that includes sedentary and active adults**\n * *Explanation:* Standard deviation measures variability. Combining non-athletes and active athletes produces the widest range of endurance scores, maximizing SD.\n\n * **Question 7:** If an exam were very easy, and most students received very high grades and very few people received low grades, the frequency distribution would be described as:\n * *Options:* A) Bimodal | B) Negatively skewed | C) Positively skewed | D) Normally distributed\n * **Answer:** **B) Negatively skewed**\n * *Explanation:* High score clusters pull the distribution mass to the right, leaving a elongated tail pointing left toward the low scores.\n\n * **Question 8:** In a hypothetical normal distribution, the mean is 10 and the standard deviation is 3. What is the score for the person who scores -1 standard deviation below the mean?\n * *Options:* A) 13 | B) 10 | C) 7 | D) -3\n * **Answer:** **C) 7**\n * *Explanation:* Calculating -1\,SD\bar{x} - 1(SD) = 10 - 1(3) = 7$$.