This preview has intentionally blurred sections. Sign up to view the full version.View Full Document
Unformatted text preview: Group Randomized Clinical Trials This section of course addresses studies in which the randomization is carried out in groups or clusters, but measurement of effect is carried out at level of individuals within the groups. Such studies are widely used in education. In medical research, often used for health policy or health care delivery interventions. Two excellent textbooks exist: Donner and Klar, Murray. Classic example: NHLBI study of intervention to reduce the delay time in seeking medical attention after heart attack. Rapid Early Action for Coronary Treatment (REACT) study. Motivation for study: • Effective treatments now exist for acute myocardial infarction (AMI):- Coronary reperfusion (by thromboytics or angioplasty).- Arrhythmia control by defibrillation or other means. • Effectiveness depends on rapid access to skilled emergency medical care. • Majority of time lost is from onset of symptoms to presentation in emergency department (ED) in medical facility. • In US, most patients stall, then self transport (not ambulance or EMS). • Can’t just focus on patients with known disease and risk factors; many AMI occur in people with no previous problems. • Theory: need general community awareness. Challenges to study design: you can’t do a mass media campaign at individual level! So the design for this study was at a city-wide level. • Ten pairs of cities in US. • One city in each pair got the intervention: 18-month campaign targeting mass media, community organizations, and professional, public and patient education. • Goal: to decrease time from symptom onset to ED arrival, increase use of EMS. • Note that randomization is by city, but outcome is for each patient who shows up with symptoms of AMI. Other settings where group- or cluster-randomized designs are necessary: • Educational interventions in schools: might have contamination across students and families, or can’t do classroom intervention for some but not other in a class. • Changes in physician or group medical practice: doctors may have similar patients, treat them generally same ways. • Nutrition education: can you change eating habits by having restaurant menu show heart-healthy options? Distinguishing characteristics of group-randomized studies and their impact Several typical features of group-randomized studies influence design, analysis and results of these studies. • Unit of assignment to treatment is a group. These are not random, but have some pre-existing connection among members. • Groups are nested within treatment: All units in a group assigned to same treatment, and several groups to each treatment....
View Full Document
- Winter '06
- Standard Deviation, Randomness, ICC