Evidence Based Practice Final Exam Review

Study Appraisal Process

• The appraisal process includes evaluating a study's applicability, assessing its quality, analyzing the results, and formulating a clinical bottom line.
• Applicability assesses who might benefit from an intervention by considering the similarities and differences between study participants and the individual patient.
• Factors to consider when determining the applicability of a study includes similarities and differences between the study participants and your patient, the intervention proposed, and the feasibility of the intervention for you and your clinic.
• A surrogate outcome is an indirect measure of a clinically relevant outcome, for example, cholesterol levels as a surrogate for heart attacks.
• The follow-up duration of a study indicates the long-term effects of an intervention.

Descriptive vs. Inferential Statistics

• Descriptive statistics summarize information about groups.
• Inferential statistics draw conclusions about differences between groups.
• Measures of central tendency, like mean, median, and mode, describe the "average" or "most typical" values.
• The median is reported for variables with skewed distributions (non-normally distributed data).
• Standard deviation (SD) measures the average variation of individual scores from the mean of a data set.
• Assessing baseline similarity between groups in intervention studies minimizes bias and ensures results are due to the intervention.

Scales & Confidence Intervals

• A Likert scale measures attitudes, opinions, or perceptions using ratings.
• The Rankin Scale assesses disability or dependence, especially in stroke and neurological conditions.
• A 95% confidence interval estimates that the true population mean lies within the indicated range with 95% certainty.
• Inferential statistics uses data from a sample to draw conclusions about whether a treatment is effective for an entire population.
• An independent t-test compares the means of two independent groups to determine if there is a considerable statistical difference, when the data is continuous.

Clinical Outcome Studies

• Clinical outcome studies generally aim to determine if new clinical outcomes are easier or more efficient than existing ones.
• Statistics used to assess agreement between scores for reliability and validity differ if the measures are dichotomous (kappa) or continuous (intraclass correlation coefficients).
• Minimum Clinically Important Difference (MCID) is expressed in units of measure or as a proportion of change from baseline.
• Cohen’s effect size thresholds (0.2, 0.5, 0.8) may not accurately reflect how patients perceive small, medium, and large effects.
• Distribution-based methods for defining MCID indicate "minimum detectable change," while anchor-based methods relate measure changes to patient-reported overall change.

Single-Subject Design (SSD)

• Single-subject design involves intensive monitoring of one participant, by repeatedly measuring outcomes during baseline, intervention, and post-treatment periods.
• Case studies are retrospective and lack the controlled intervention and experimental controls found in SSDs.
• Repeated measurement is important because it measures the fluctuations in the outcomes of interest before intervention begins.
• Repeated measurements can alter a participants performance, either positively or negatively.
• A celeration line is a "best-fit" line through data points in each phase, often determined by the median value for each phase.
• The slope of the celeration line indicates the rate of increase or decrease in the data.
• P-values are not the common expression of statistical significance in SSD due to the common features of SSD data sets.
• A "two-standard-deviation band" analyzes statistical significance alongside the celeration line; significance is indicated if two data points in the intervention phase are above or below the bands.

Qualitative Research & Sampling

• Qualitative research focuses on questions of experience, culture, and social/emotional health.
• Purposive sampling selects participants based on their ability to inform understanding of a topic, rather than random selection.

Longitudinal, Cross-Sectional, Cohort, & Case-Control Studies

• Longitudinal studies will follow patients over time to assess relationships between factors and outcomes.
• Cross-sectional studies provide a snapshot of patients at a single time point to show how factors and outcomes align.
• Cohort studies prospectively track a group to see who develops a specific outcome.
• Case-control studies retrospectively investigate the causes of an outcome by comparing those with the outcome (cases) to those without (controls).
• Prognostic research questions primarily focus on the impact of a disease or event on long-term patient outcomes.
• Factors influencing answers to prognostic questions include the severity of the patient's problem, gender, age, home environment, and comorbidities.
• In a longitudinal study, the sample participants must not already have the study outcome.
• Prognostic studies typically require large samples due to the numerous factors that can contribute to prognostic statements.
• Objectivity in prognostic studies is improved when "measurers" are blinded to the study purpose or participant group status.
• Prognostic studies should detail the monitoring of participants, to check on compliance or to maintain continued participation.

Prevalence & Incidence

• Prevalence is determined from cross-sectional or retrospective studies and looking back at medical records.
• Incidence of a condition is determined using prospective studies.

Sensitivity & Specificity

• Sensitivity indicates how well a test correctly identifies individuals who HAVE a condition.
• High sensitivity means a negative result is likely correct, and the person probably does not have the condition (helps rule out the condition).
• Specificity indicates how well a test correctly identifies individuals who DO NOT have a condition.
• Excellent specificity (90%+) means trust in a negative result is high, indicating the condition is truly absent.
• Lack of specificity results in greater uncertainty following a negative test if you have the condition.
• Poor specificity requires high sensitivity to ensure those with the condition are still identified by the test.

Confounders

• Confounders are variables that can influence the relationship between an exposure and an outcome by influencing both.
• Randomization helps ensure confounders are balanced between groups, reducing the risk of bias.
• The two main ways to deal with confounding are randomization of the exposure and statistical adjustment (controlling) of confounders.

Correlation vs. Causation

• Correlation indicates the association between variables, measurements, or outcomes.
• A correlation test results in a value "R," ranging from -1.0 to +1.0.
• An R value of 0 indicates no correlation.
• A positive R value indicates a positive association, meaning as one value goes higher, so does the other value.
• Correlation indicates an association between variables, while causation implies that one variable directly causes a change in another.

IRB & Research Dissemination

• IRB approval is required for human subjects research to ensure ethical standards are met.
• Informed consent ensures participants are aware of and agree to the study and its risks.
• The two traditional forms of dissemination are conference presentation and peer review journal article.
• Poster presentations allow for individual interaction, while podium presentations involve a focused presentation to an audience.