Topic C Notes

Página 1

Types of Research and Research Designs

Steps in Research

1. Identify relevant and important topic, review published research literature related to topic

2. Develop research question - clear and simple

3. Use question to make a hypothesis - SMART, defines variables and statistical significance

Hypothesis: prediction of a relationship

Null hypothesis: no relationship in the population of data

PICO Hypothesis: Population, Intervention/Exposure, Comparison, Outcome

4. Prepare research protocol

5. Organize methods and materials

6. Collect and analyze data

7. Study results and make decision

8. Study design and checklists

Página 2

Types of Research and Research Designs

Meta Analysis

Formal, defined system combines results of numerous small studies to increase strength of belief in observed effect - studies have similar design, defined inclusion and exclusion criteria, published, peer reviewed

QUOROM: proposed quality standards for the review and monitoring of systematic reviews and meta-analysis of randomized trials

Companion standards (common evaluation criteria for clinical trials):

-PRISMA: minimum set of items for reporting, focuses on the reporting of reviews evaluating randomized trials

-MOOSE: meta-analysis of observational studies in epidemiology

-CONSORT: common criteria for clinical trials

Research Report

Abstract: condensation of final report with purpose of study, questions asked, scope, method, conclusions

Introduction: objectives, definitions, background, limitations

Review of existing literature: summary of different points of view

Methodology: statement of hypothesis, discussion of methods

Results: specific lab, clinical, objective findings

Discussion: interpretation of the results, comparison with other studies

Conclusion: brief summar of results, with recommendations

Inplications: how the info might be applied in practice

Página 3

Types of Research and Research Designs

Descriptive Research

Descriptive: describes state of nature at a point in time, hypothesizes determinants of a condition, baseline data and monitoring of changes over time, establish associations among factors but does NOT allow causal relationships to be determined

-Qualitative (explore a phenomenon of interest before theory development)

-Case Report

-Surveys

-Correlation or Ecological studies (compare frequency of events in different population with consumption of dietary factors)

Analytical Research

Tests hypotheses concerning the effects of specific factors of interest and DOES allows causal associations to be determined

Clinical trials, follow up studies, case control studies

Página 4

Types of Research and Research Designs

Types of Analytical Research

Experimental: randomization of experimental and control groups, placebo may be in effect, program successful if experimental group has improved more than control group

Clinical Nutrition Studies (Randomized Controlled Trial or RCT): comparison placebo control group is gold standard

-Parallel design: participants randomly assigned to a treatment group and remain there for the study

-Crossover design: participants serve as their own control

-Two Period Crossover: each participant would either receive intervention or control in first period and alternate in second - variability is reduced

Quasi Experimental Design: series of measurements at periodic intervals before the program begins and after the program ends - shows whether changes are a previous pattern or a noteworthy change

Cohort Studies: group of healthy people followed through time to see if they develop a specific disease

Case Control Studies: focus on specific disease and compare to group without the disease but with other similar characteristics

Cross-Sectional Studies: one time data collection counting all of the cases of a specific disease among a group of people at a particular time - snap shot

Página 5

Types of Research and Research Designs

Institutional Review Board (IRB, under the FDA)

Committee that reviews and approves research involving human subjects to make sure it is legal and ethical

Author

Authors of manuscripts in biomedical journals: make substantial contributions to the design or analysis and the interpretation of the data

Página 6

Statistical Evaluation, Interpretation and Application

Relevance or Validity - ability to measure the phenomenon you intend to measure

Internal validity: tests whether difference between two groups is real

External validity: can generalizations be made from study to larger population

Analysis of Variance = ANOVA - is the difference a reliable one that would be repeated

Reliability - Consistency of Reproducibility of Test Results

Test then retest, are results similar

Parallel forms: two separate but similar forms of the same test at the same time

Split Halves: Divide the test in half

Precision: Amount of variation that occurs randomly

Página 7

Statistical Evaluation, Interpretation and Application

Sensitivity and Specificity

Sensitivity: proportion of afflicted individuals who test positive

Specificity: Proportion of non-afflicted identified as non-afflicted

Variables

Nominal: Fit into a category with no specific order

Rank Order

Numerical Discrete: Data with numbers

Numerical Continuous: underlying continuous scales (blood pressure)

Dependent Variables are outcomes, independent variables are what you change

Página 8

Statistical Evaluation, Interpretation and Application

Sampling

Probability: Each segment of population represented in the sample - selects from a larger population with randomization

Non-probability: no way of forecasting that each element will be represented in the sample

Statistical Evaluation, Interpretation and Application

Measuring Data Distributions - Central Tendency

Arithmetic Mean: Simple average

Median: Midpoint

Mode

Página 9

Statistical Evaluation, Interpretation and Application

Measuring Data Distributions - Dispersion

Range

Standard Deviation: 68% of all observations in a normal curve fall within 1 SD of the mean, 95% lie within 2 SD of the mean

Statistical Evaluation, Interpretation and Application

Existence of Relationships

Correlations - closer points are to the line, stronger the relationship

R=1 all dots on line with positive slope, R=-1 all dots on line with negative slope, closer to 0 means no linear correlation

Página 10

Statistical Evaluation, Interpretation and Application

Clinical Significance

Outcome must be statistically significant - the lower the p value, the higher the significance and confidence

Looking for P <0.05

Statistical Evaluation, Interpretation and Application

Other Definitions

Double blind: removes bias from research as neither participant or researcher know

Mortality is rate of death, morbidity is state of disease

NOnparametric test: does not depend on a normal distribution

Dichotomous score: only two results are possible

Chi Square Test: whether or not there is a real difference between categories

T Test: significance between means of two different populations (null against alternative hypothesis)