23. Quasi-Experimental Study

Question 1

what is a quasi-experimental study?

Answer 1

a study that aims to eval interventions but does not utilize a randomized control group

• tests descriptive causal hypotheses about manipulable causes
• QE data = summary data before and after a non-randomized intervention

Question 2

when do you do a quasi-experiment?

Answer 2

• chosen when randomized controlled trial or cluster-randomized controlled trial are not possible (time and money and ethics)
• frequently used in quality improvement

Question 3

most basic type of QE study?

Answer 3

baseline data period (O1)

• > implement intervention (x)
• > collect same data as during baseline (after) (O2)

Question 4

threats to internal validity in QE studies?

Answer 4

• selection bias: difference in group characteristics that could produce effect (when you choose a control group) (severity of illness, hospital staff practices may differ, eg MICU vs SICU)
• historical bias: contemporaneous events that produce effect (other interventions)
• regression: if selected for initial extreme measurements, follow-up measures have a tendency to be less extreme and this can be confused with treatment effect (extreme measures come back to the norm)
• maturational effects: temporal and seasonal changes (solution is to get a year’s worth of data before and after or to compare the same calendar time periods)

Question 5

improving QE studies: regression, historical bias, and maturational effects

Answer 5

regression can be reduced by not selecting groups based on extreme measurements (eg very high MRSA rates)

historical bias can be reduced by preventing other events from occuring

maturational effects can be reduced by getting a year’s worth of data before and after to or compare the same calendar time periods

Question 6

does cause precede effect in RCT or QE?

Answer 6

both (if controlled by researcher)

Question 7

does cause covary with effect in RCT or QE?

Answer 7

both (if statistically confirmed)

Question 8

are alternate explanations implausible in RCT or QE?

Answer 8

RCT (this is ensured by randomization, the lack of this is a threat to internal validity in QE study)

Question 9

Heirarchy of QE study designs (A<b></b>

Answer 9

A. QE designs without control groups (most often used)
B. QE designs with control groups but no pretests
C. QE designs that use control groups and pretests
D. Interrupted time-series designs

Question 10

heirarchy of QE designs without control groups (A2-A6)

Answer 10

A2. one group pretest-posttest design (O1 X O2)

A3. one group pretest-posttest design with double pretest (O1 O2 x O3)

A4. One-group pretest-posttest design using a nonequivalent dependent variable (O1a, O1b) X (O2a, O2b)

A5. Removed treatment design (O1 X O2 O3 X O4