ATC 420 - Research Designs

validity

accuracy

reliability

repeatability

validity and reliability

study can be valid but not reliable

VICE VERSA

generalizability

large enough sample so results can apply to broader/more general population

5 steps of evidence based practice

1. ask

2. acquire

3. appraise

4. apply

5. audit

ask

convert information need into an answerable clinical question

acquire

to obtain

appraise

critically appraise the evidence for validity, impact, and applicability

apply

integrate the evidence into your clinical decision making

audit

evaluate steps 1-4 and seeks ways to improve

scientific method

1. make observation

2. review of literature

3 state problem

4. ID variables

5. hypothesis

6. research design

7. IRB approval

8. collect data

9 statistic analysis

10. conclusions

11. apply findings

12. assess application

ID variables

independent

dependent

control

independent variable

manipulated in experiment

dependent variable

outcome on manipulated variable

control variable

the equalizer and delimitation

IRB approval

institutional review board approval

null hypothesis

statement of no difference

A = B in study (statistical)

alternate hypothesis

statement predicting what will happen

qualitative research designs

1. case report

2. case series

3. cross sectional

4. case-control

5. cohort studies

qualitative

"why"

observation, words, symbols, etc.

quantitative

"how many/much"

always a number result

measurable and testable

statistical analysis

mixed methods

combination of qualitative and quantitative

take quantitative date and ask qualitative questions

non-experimental

no comparison group

quasi-experimental

control group to be compared

experimental

measures outcomes for both groups

explicit comparison group

relationship

correlation

cause and effect

t test

prevalence

frequency (observation)

compare

pre vs post

A vs B

regression analysis

predict

randomization

increases level of evidence

recall/memory

decreases level of evidence

large n

larger population is better for study

clinical research

may be unethical (placebo)

weakness of research designs

subjects representative of whole target population

qualitative factors

no cause and effect

no randomization

no equal start

quantitative research designs

1. simple RCT

2. between subject RCT

3. between subject pre-post RCT

4. within subject RCT

5. within subject pre-post RCT

levels of evidence

the LOWER the number the BETTER the quality of evidence

bench/expert

neither qualitative or quantitative

bench

lab study done on animals or cells

(unreliable since not done on humans)

expert

presentation from an specialist in the field

(no study, just facts from an expert, neglects conflict of interest)

case report

observational

qualitative

one subject

case series

case report with more people

qualitative

no control group (everyone receives manipulated variable)

cross sectional

observational

qualitative

to understand prevalence

snapshot of a particular period of time

case-control

observational

qualitative

outcome is already known

compares subjects with and without (control group)

looks in past

relies on recall/memory

confounding variables not considered

ONLY QUALITATIVE WITH CONTROL GROUP

cohort studies

prospective and retrospective

observational

qualitative

may rely on recall/memory

difficult to populate

needs larger subject size

confounding variables not considered

prospective cohort

observes a population and what things influence outcomes

(conducted presently and looking at future)

retrospective cohort

takes a group and sees if there are any trends from the past that would cause ______

(conducted after and looks into the past)

simple RCT

experimental research design

quantitative

researchers manipulate independent and observe dependent

control group vs treatment group

simple RCT disadvantages

clinically preventing someone from getting something they need

OR

giving something that results in bad outcomes

improve simple RCT

add a control group as a baseline

between subject RCT

compares control vs experiment group

between subject pre-post RCT

testing before and after manipulation

pre acts as control (baseline)

between subject threats

boredom

crossover

fatigue

within subject RCT

subject is being compared to oneself

within subject pre-post RCT

crossover study

everyone serves as own control group

washout period

both groups doing the same thing

within subject threats

order effect (learning effect and randomization)

clinically have to give something up thats potentially working to do the other part of study

any learning

cannot be unlearned

meta-anaylsis

makes any research design stronger

(one step up)

strongest research design

meta-analysis of RCT

research design order

FINER

Feasible

Interesting

Novel

Ethical

Relevant

feasible

possible, able to be done

interesting

for researcher to study/to read

novel

new and different

ethical

morally right

relevant

important or pertinent

p levels

statistical significance

sound be set to <0.05 (95%)

probability results would occur if null hypothesis is true