Evidence Based Practice Final Exam Review (Comprehensive)

Introduction to Evidence Based Practice

• Evidence Based Practice (EBP) integrates research evidence, clinical expertise, and patient values/circumstances.
• EBP guides treatment, references effective/ineffective practices and optimizes treatment and patient responsiveness.
• The 5 steps of EBP are: formulate a question, search for evidence, appraise evidence, integrate evidence with expertise/values, and evaluate effectiveness.
• The 3 pillars of evidence are: scientific research, clinical expertise, and patient's values/circumstances.
• The 3 sources of evidence are research, clinical experience, and patient preferences.
• Systematic reviews and clinical practice guidelines sit atop the evidence pyramid, followed by RCTs, cohort studies, case-control studies, case series/studies with narrative reviews, expert opinion, and textbooks at the bottom.
• Patient management aspects include screening, examination, diagnosis, intervention, prognosis, and outcome measurement.
• Research evidence is rated based on applicability (relevance) and quality (rigor/validity).
• Barriers to EBP include time, lack of generalizability, research skills, statistical understanding, appraisal skills, resources, and inconsistent culture.
• Shared informed decisions are made collaboratively by therapist and patient, informed by evidence.
• Clinical research involves human subjects; it addresses diagnosis, intervention, prevention, and prognosis.
• Nonclinical research involves healthy humans, animals, cadavers, and cells.
• Systematic Reviews (SRs) combine other studies and are considered secondary studies.
• Randomized Clinical Trials (RCTs) are individual studies often considered high-level evidence for intervention questions.
• Clinical practice guidelines are evidence-based summaries of recommendations, combining research, expertise, and patient perspectives.
• Study appraisal involves determining the applicability and quality of a research article.

Comprehensive Literature Searches

• Prediction in research helps identify expectations for the future using present information within longitudinal study designs.
• Association or correlation does not imply causation in research.
• Randomized controlled trials (RCTs) are the best for determining cause-and-effect.
• Descriptive questions describe the current situation using cross-sectional & longitudinal surveys, clinical notes, etc.
• Predictive questions learn about the future using present information and longitudinal designs.
• Causal questions find treatment targets or compare treatment effects using RCTs and cohort studies.
• PubMed - MeSH (Medical Subject Headings) uses controlled vocabulary to narrow searches with related terms.
• PubMed - Clinical Queries are predefined filters for clinically relevant and disease-specific topics.
• Google Scholar and PubMed can be searched by date to find the newest evidence.
• PEDro (Physiotherapy Evidence Database) is specific to physical therapy research.
• APTA searches can utilize Clinical Practice Guidelines, Clinical Summaries, and the APTA Journal.
• Cochrane Searches focuses on systematic reviews and includes Population, Intervention, Comparison, Outcomes (PICOs).
• The placebo effect is influenced by confidence, beliefs, misinformation, and source of information.
• The nocebo effect is facilitated by emotions and overdramatizing treatments.
• Physical therapists should prevent nocebo effects by refining language and overwhelming with positivity.

Introduction to Study Appraisal

• When selecting literature, make a clinical judgment about which study is "close enough" to your patient population.
• Surrogate outcomes are indirect measures of clinically useful outcomes where the relationship to the major outcome may not always be known.
• Applicability is how well the study relates to your patient population where you consider similarities and differences between the study participants and your patient.
• Realistic interventions focus on the intervention proposed in the study and the feasibility of the intervention using available resources.
• Relevant outcomes determine if the outcome measures used in the study are pertinent to your clinical question.
• Follow-up duration provides information about the long-term effects of an intervention.
• Preservation of original groups matters in intervention studies, analyze participants in their initially assigned groups; changes justified.
• Single-blinded studies involve either participant or clinician/rater being unaware of the treatment assignment.
• Double-blinded studies involve both participant and clinician/rater being unaware.
• Triple-blinded studies involve the participant, clinician/rater, and the statistician being unaware.
• A Sham Group is a control group receiving a placebo or non-active intervention to account for psychological effects.
• The appraisal process includes appraising the study for applicability and quality as well as appraising study results and formulating a clinical bottom line.
• Applicability evaluates a study and determines what specific people might benefit from the intervention.
• Selection bias occurs when certain participants are preferentially chosen for a study.

Critically Appraise the Results of an Intervention Research Study

• Descriptive statistics summarize data with typical values and variability (e.g., mean, median, mode, standard deviation).
• Inferential statistics use probability to interpret differences and make conclusions about populations (e.g., t-tests, ANOVA, p-values, confidence intervals).
• Clinical importance of results is determined by the magnitude of change on measures valuable to the patient, not solely by statistical significance.
• The mean is the sum of observations divided by the number of observations.
• The median is the point below which half the observations fall, best describing skewed data.
• The mode is the most frequent score.
• Alpha Value (α) is typically 0.05 (5%), the probability of concluding there is a difference when none exists (Type I error).
• Effect size represents the magnitude of the difference between treatments.
• When appraising results, read tables/figures before the text, then the results section, look for confidence intervals and clinically meaningful statistics, then read the authors' conclusions.
• A normal distribution is a bell-shaped curve where the mean is in the middle, and standard deviations capture known percentages of the sample.
• Non-normally distributed data (skewed) is common in clinical populations.
• Groups in an intervention study should be as similar as possible at baseline; randomization helps.
• Reliability of measures is consistency in score production.
• Intra-rater reliability is the consistency of one therapist over time.
• Inter-rater reliability is consistency between different therapists.
• Validity of measures is the ability to measure what it is intended to measure.
• The null hypothesis states there is no difference between groups.
• Inferential statistics test the probability of rejecting the null hypothesis.
• Intraclass Correlation Coefficient (ICC) measures association and amount of association for continuous data.
• Spearman's rho measures association of two measures for ordinal data.
• Kappa (κ) measures the association of two measures accounting for chance agreements in nominal data.
• Kendall's tau measures association of two measures for nominal data.
• Cronbach's alpha indicates internal consistency of an instrument.
• Standard Error of Measurement (SEM) measures the variability of the standard deviation.
• The confidence interval (CI) is a range of values likely to include the true population mean.
• P-values are the probability that the observed difference is due to chance.
• Chi Square (X²) analyzes nominal (categorical) data, comparing observed and expected frequencies.
• A T-test compares the means of two groups.
• Analysis of Variance (ANOVA) compares the means of more than two groups.
• Repeated Measures ANOVA (Univariate) compares multiple measures from the same subjects over time.
• The F Value is the statistic produced by ANOVA.
• Intention-to-Treat (ITT) Analysis analyzes all subjects in their initially assigned groups and helps account for attrition in RCTs.
• Effect Size (Cohen's d) measures the magnitude of the difference between two means relative to the variability (large > 0.8, medium 0.5-0.8, small 0.2-0.5).
• Number Needed to Treat (NNT) is the number of patients you need to treat with a new intervention to see one additional positive outcome compared to the control.
• A lower NNT is generally better.
• Key appraisal questions when interpreting results: Were groups similar at baseline? Were outcome measures reliable and valid? Were CIs reported? Were descriptive and inferential statistics applied? Was there a treatment effect, and was it clinically relevant?

Clinical Intervention Studies & Interpreting Outcomes 1 & 2, Sampling

• Key data concepts include normal/skewed distributions, descriptive statistics (mean, SD, median, CI, range), and types of data (nominal, ordinal, Likert, Rankin, continuous/ratio, interval).
• A Likert Scale measures attitudes/opinions with a range of responses.
• A Rankin Scale (Modified Rankin Scale, mRS) assesses disability/dependence after stroke or neurological conditions.
• Reliability types include intra-individual/subject, intra-rater, and inter-rater.
• The Intraclass Correlation Coefficient (ICC) helps quantify Reliability.
• Hypothesis testing (NHST) is Null Hypothesis Significance Testing.
• Type I error (α) rejects the null hypothesis when it is true (false positive).
• Type II error (β) fails to reject the null hypothesis when it is false (false negative).
• Inferential Statistics include T-tests, paired samples t-tests and ANOVAs.
• T-tests compares two groups of continuous data.
• Paired samples t-test compares change within an individual.
• ANOVA compares multiple groups.
• Ensure groups are similar at the start of the study by looking at baseline characteristics.
• Within-group change looks at changes within a single group over time.
• Between-group difference looks at differences observed between two or more groups.
• Statistical Significance looks at the probability that the observed effect is not due to chance (often p < 0.05).
• Clinical Meaningful Difference looks at a difference large enough for patients to consider important.
• The goal of sampling is to recruit a representative sample that mirrors the population except for size.
• Random Sampling means every individual in the population has an equal chance of being selected.
• Consecutive Sampling means everyone meeting inclusion criteria at a certain place during a defined period is invited until the sample size is reached.
• Selection Bias occurs when certain members of the population are preferentially recruited.
• Generalizability looks the extent to which study results from the sample can be applied to the broader population.

Outcomes

• Clinical Outcome Studies compare existing clinical outcomes to determine if a new outcome is easier or more efficient.
• Determine if a new clinical outcome is as good as an older, established one in measuring the intended construct.

Alternative Studies, Prognostic Studies, Diagnostic Studies

• In Single-Subject Design (SSD), a participant is intensely studied with repeated baseline measurements, intervention with data, and possible post-intervention measurements. Change during intervention compares to baseline.
• Case studies entail retrospective details of one case and intervention, lacking systematic control of SSDs.
• Quality of SSR requires valid measurements, control of extraneous factors, with blinding of evaluators as well specific and well-described intervention.
• Analysis of SSR includes graphs and statistics analyzed with visual inspection.
• Qualitative Research should focus on experience, culture, and health to understand processes and can generate hypotheses.
• Purposive Sampling selects participants specifically for their ability to contribute..
• Prognostic Studies address questions about the impact of an event on long-term outcomes and helps guide discharge planning.
• Risk factors such as severity, gender, age, environment, and comorbidities influence prognosis.
• Prognostic Studies include cohort studies, case-control studies and cross-sectional studies.
• Prognostic studies show association, not necessarily causation when cohort studies are better for causality.
• Applicability of Prognostic Studies depends on the similarity of the study sample (condition, age, comorbidities, therapy access) to your patient.
• Diagnostic studies are used to determine how common a disease/condition/injury is in a population at a specific time (determined by cross-sectional or retrospective studies).
• Incidence is how common it is for individuals to develop a disease/condition/injury over a period of time (determined by prospective studies).

Confounders, Causation and Correlation, and Research Integrity & Week 13 - Research Integrity

• Correlation indicates an association between two variables as one changes and the other tends to change as well.
• Quantifying Correlations (R value ranges from -1.0 to +1.0.
• Directed Acyclic Graphs (DAGs) display theoretical relationships to identify potential confounders to adjust for.
• Assessing Risk of Confounding looks for adjustment confounders and variables that can cause both exposure and outcome.
• Institutional Review Board (IRB) is required for human subject research and consists of experts in clinical practice, research, ethics, statistics, and law.
• Informed Consent states explanation of the study and opportunity to read documents as well as participants can withdraw anytime without penalty.
• Research Pre-Registration enhances transparency, helps recruit participants, and improves research integrity.
• Conference Abstracts are submitted for conference presentations and vary in length, reviewed by the committee, and compiled in conference proceedings.
• Survivorship Bias occurs when analysis excludes those who failed or dropped out, leading to overly optimistic results.
• Attrition Bias occurs through disproportionate dropout between groups and skews comparisons.
• Conflict of Interest entails personal factors that could influence study integrity.
• Reporting Bias occurs when data gets held back or is selectively published data to favor the researchers' interests.
• Outcome Switching is the retrospective changing the primary outcome after data collection.
• P-Hacking means data analysis is manipulated until a statistically significant p-value is obtained, often without a sound scientific basis.