NPTE Research and EBP

systematic review

randomized control trials --> meta analysis

Four steps to EBP

1. Clinical problem identified and answerable research questions formulated.

2. A systematic review is conducted and best evidence is collected.

3. The research evidence is summarized and critically analyzed.

4. The research evidence is synthesized and applied to clinical practice.

Descriptive research (what is it?)

Involves collecting data about conditions, attitudes, or characteristics of subjects or groups of subjects.

Determines and reports existing phenomena to CLASSIFY and IDENTIFY.

Collected by: questionnaire, survey, observation, interview

Descriptive research (examples)

- Case studies/reports

- Developmental research (differentiate individuals at different levels - age, growth)

- Longitudinal studies - over time

- Normative research - investigates standards of behavior (gait characteristics)

- Qualitative research - seeks facts or causes of social phenomena/complex human behavior; uses inductive reasoning to develop concepts, insights, and understanding from patterns in data

Correlational research (what is it?)

Attempts to determine whether a relationship exists between two or more quantifiable variables and to what degree.

Limitations:

- cannot establish cause and effect relationships

- may fail to consider all variables that enter into relationship

Correlation coefficient

1.00 = positively correlated

0.00 = no relationship

-1.00 = inversely related

Correlational research (examples)

- Retrospective

- Prospective

- Descriptive - investigation of several variables at once

- Predictive - useful to determine predictive models

Experimental research

Attempts to define a cause and effect relationship through group comparisons

Cohort design

Quasi-experimental design; subjects are identified and followed over time for changes/outcomes following exposure to an intervention; lacks randomization and may not have a control group

Within-subject design

Repeated measures; subjects serve as their own controls; randomly assigned to treatment or no treatment blocks

Single subject experimental design

Involves a sample of one with repeated measures and design phases:

A-B: pre-treatment/baseline phase, then tx phase

A-B-A: baseline, tx, then baseline phase

A-B-A-B: baseline, tx, baseline, tx

Ex post facto research

Casual-comparative research; attempts to define a cause-and-effect relationship through group comparisons where the cause or independent variable has already occurred (gender)

Epidemiology

The study of disease frequency and distribution in a community.

Nominal

Classifies variables or scores into two or more mutually exclusive categories based on a common set of characteristics.

Examples: gender, tall or short.

Description: frequency, percentage

Relationship: chi square

Differences: One group (chi square), 2 groups(independent = chi square; paired = McNemar test), more than 2 groups (independent = chi square; paired = Cochran's Q)

Ordinal

Classifies and ranks variables or scores in terms of the degree to which they posses a common characteristic.

Examples: MMT, GPA

Description: Median, mode, range

Relationship: Spearman's Rank Order, Kendall's Tau Correlations

Difference: 2 groups (independent = mann-whitney U; paired = Wilcoxon Signed Rank), more than 2 groups (independent = Kruskal-Wallis; paired = Friedman ANOVA)

Interval

Examples: (IQ test, temperature scales)

Description: Mean, median, mode, SD, skew of the distribution

Relationship: Pearson's product moment correlation, partial correlation, multiple correlation, multiple regression)

Difference: Two groups (independent = t-test; paired = test-test), more than 2 groups (independent = ANOVA; across time = repeated ANOVA)

Ratio

Classifies and ranks variables or scores based on equal values and a true zero point.

Examples: weight, height, goniometry

Types of sampling

Random: Individuals in pop. have equal chance of being selected.

Systematic: Take individuals at specified intervals (every 10th person)

Stratified: Individuals selected from identified subgroups based on some predetermined characteristics (weight, height, gender)

Internal validity

The degree to which the observed differences on the dependent variable are the direct results of manipulations of the independent variable and not some other variable.

External validity

The degree to which the results are generalizable to individuals (general population) or environmental settings outside of the experimental study.

Face validity

The assumption of validity based on the appearance of an instrument as a reasonable measure of a variable; may be used for initial screening of a test instrument by psychometrically unsound. It refers to the transparency or relevance of a test as it appears to test participants - it "looks like" what it is supposed to measure.

Content validity

The degree to which an instrument measures an intended content area. For example, a depression scale may lack content validity if it only assesses the affective dimension of depression but fails to take into account the behavioral dimension. Determined by expert judgment.

Concurrent validity

The degree to which the scores on a test are related to the scores on another criterion test with both tests being given at relatively similar times; usually involves comparison to the gold standard.

Predictive validity

The degree to which a test is able to predict future performance.

Construct validity

The degree to which a test measures an intended hypothetical abstract concept (non-observable behaviors or ideas). You might think of construct validity as a "labeling" issue. When you implement a program that you call a "Head Start" program, is your label an accurate one? When you measure what you term "self esteem" is that what you were really measuring?

Threats to validity

- Sampling bias: The researcher uses a sample of convenience (volunteers, available groups) instead of random selection

- Failure to exert rigid control over subjects/conditions

- Pre-test influences scores on post-test (learning effect)

- Measurement tool not accurate

- Subjects respond to tx different because of pre-test

- multiple tx interference or carry-over effects

- experimenter bias

- Hawthorne effect

- Placebo effect

Hawthorne effect

The subject's knowledge of participation in an experiment influences the results of a study; threat to validity

Test-retest reliability

The degree to which the scores on a test are stable or consistent over time; a measure of instrument stability

Split-half reliability

The degree of agreement when a test is split in half and the reliability of the first half is compared to the second half; a measure of internal consistency of an instrument

Sensitivity

A test's ability to correctly identify the proportion of individuals who truly have a disease or condition (TRUE POSITIVE).

The sensitivity describes the ability of a diagnostic test to identify true disease without missing anyone by leaving the disease undiagnosed. Thus, a high sensitivity test has few false negatives and is effective at ruling conditions "out" (SnOUT).

Specificity

A test's ability to correctly identify the proportion of individuals who do not have a disease of condition (TRUE NEGATIVE)

The specificity describes the ability of a diagnostic test to be correctly negative in the absence of disease without mislabeling anyone. Thus, a high specificity test has few false positives and is effective in ruling conditions "in" (SpIN).

Predictive value

A test's ability to estimate the likelihood that a person will test positive (or negative) for a target condition.

LOER: 1

Grade A

1a = SR of RCTs (with homogeneity)

1b = RCT (narrow confidence level, treatment effects)

1c = all or non case series (overwhelming evidence of substantial treatment effect)

LOER:2

Grade B

2a = SR of cohort studies (with homogeneity)

2b = individual cohort study or low quality RCT (small N)

LOER: 3

Grade B

3a = SR of case-control studies (with homogeneity)

3b = Individual case control study

LOER: 4

Grade C

Case series, poor quality cohort studies and case-control studies, largely descriptive studies

LOER: 5

Grade D

Expert opinion without explicit critical appraisal, or case on physiology, bench research, or first principles; observations not made on patients

Standard deviation

Determination of variability of scores (differences) from the mean.

Most scores are near the mean, within one SD; ~ 68% of scores fall within 1 SD; 95% of scores fall within 2 SD; 99% of scores fall within 3 SD

Alpha level

Preselected level of statistical significant; usually 0.05 or 0.01.

Indicates that the expected different is due to change; for example, at 0.05 there is only a 5% chance that the differences are due to chance (and not the tx)

Allows rejection of the null hypothesis

Degrees of freedom

Based on number of subjects and groups. Allows determination of level of significance based on consulting appropriate tables for each statistical test.

Standard error

Expected chance variation among the means, the result of sampling error.

The standard error (SE) is the standard deviation of the sampling distribution of a statistic

Type 1 error

Incorrect rejection of a true null hypothesis (a "false positive"); detecting an effect that is NOT actually present.

Type 2 error

Failure to reject a false null hypothesis (a "false negative"); failing to detect an effect that is present.

How do you decrease type 1 and type 2 error?

Increase sample size, using random selection, and having valid measures.

Parametric statistics

Testing is based on population parameters. For ratio or interval data.

Assumptions:

1. Normal distribution exists.

2. Random sampling is performed.

3. Variance in the groups is equal.

T-test

Parametric test uses to compare two independent groups created by random assignment and identify a difference at a selected probability level.

Paired t-test

Compares the difference between two matched samples. Example: Subjects are tested prior to a treatment, say for high blood pressure, and the same subjects are tested again after treatment with a blood-pressure lowering medication. By comparing the same patient's numbers before and after treatment, we are effectively using each patient as their own control.

ANOVA

Analysis of variance; parametric test used to compare THREE or more independent treatment groups or conditions at a selected probability level.

One-way ANOVA

Compares multiple groups on a single independence variable.

Example: Three sets of posttest scores are compared from three difference categories of elderly (young, old, and old and frail elderly).

Factorial/multifactorial ANOVA

Compares multiple groups on two more independent variables.

Example: Two groups of injured patients (those with several ankle sprain and moderate ankle sprain) and a control group are compared for muscle activation patterns and sensory perception in each limb.

Analysis of covariance (ANCOVA)

Parametric test used to compare two or more treatment groups or conditions while also controlling for the effects of intervening variables (covariates)

Example: Two groups of subjects are compared on the bases of gait parameters using two different types of ADs; subjects in one group are taller than the subjects in the second group; height then becomes the covariate that must be controlled during statistical analysis

Nonparametric measures

Testing not based on population parameters. For ordinal or nominal data.

Less powerful than parametric tests, more difficult to reject the null hypothesis.

Chi square test

Nonparametric test of significance used to compare data in the form of frequency counts occurring in two or more mutually exclusive categories

Pearson product moment coefficient (r)

Used to correlate continuous data with underlying normal distribution on internal or ratio scales. A measure of the linear correlation (dependence) between two variables X and Y, giving a value between +1 and −1 inclusive, where 1 is total positive correlation, 0 is no correlation, and −1 is total negative correlation.

Example: the relationship between proximal and distal development in infants is examined.

Spearman's rank correlation coefficient (r or rho)

Nonparametric test used to correlate ordinal data. Pearson's product for RANKED variables.

Example: The relationship of verbal and reading comprehension scores is examined.

Point biserial correlation

One variable is nominal and the other is ratio or interval.

Example: The relationship between elbow flexor spasticity and side of stroke (left or right) in stroke patients.

Rank biserial correlation

One variable is nominal and the other is ordinal.

Example: The relationship between gender and functional ability.

ICC

Intraclass correlation coefficient: a reliability coefficient based on an analysis of variance

Common variance

A representation of the degree that variation in one variable is attributable to another variable.

Determined by squaring the correlation coefficient. Example: Coefficient of 0.70 has a common variance of 49% meaning that the variation in one variable can be explained by the other 49% of the time.