Jan 02, 2015
Terminology 101: Randomization in RCTs
Randomization in RCTs: The process of assigning study participants to groups whereby each study participant has an equal opportunity to be assigned to either of the study groups
Source: Gordis, L. (2014). Epidemiology. (5th ed.). Philadelphia: Elsevier Saunders
To determine whether the intervention causes the outcome in a randomized controlled trial (RCT), researchers must ensure that the intervention and control groups are similar in all baseline characteristics (both known and unknown) that could influence the outcome. Randomization, or random assignment, is used to avoid the possibility of a systematic bias in the assignment of participants to groups. Without randomization, for instance, healthier-looking patients might be assigned to the intervention group and sicker-looking patients to the control group to favour the treatment effect. In a properly randomized study, every participant has an equal opportunity to be assigned to either of the study groups. Randomization is a defining feature of RCTs and it must never be subverted.
There are three approaches to randomization in RCTs. The first, simple random assignment, is similar to flipping a coin: participants are assigned to a study group using a computer-generated sequence. An advantage of this technique is that it is not possible for the administrator to predict the next assignment. However, it tends to produce imbalanced group sizes when used in small samples, and thus it is best used in relatively large single-site studies.
In the second approach, permuted block randomization, a computer generates blocks of equal numbers of As and Bs, whereby A and B represent the control and intervention groups. For instance, if the sample size of the study is 160 participants, 40 blocks of four participants each could be generated — two participants from each group. Each block will then have six possible permutations (ABAB, BABA, BBAA, AABB, ABBA, BAAB). Block randomization tends to generate equal sample sizes between the groups, and thus it is especially useful in multi-site studies. Its major disadvantage is that administrators can predict the assignment at the end of a block if they know the number of participants per block.
The third approach, stratified randomization, should be used when researchers suspect that one or more baseline characteristics may be important to understanding the outcome and therefore must be equally represented in both groups. If the intervention may be more effective if used for the mild form of a disease than the severe form, for instance, then researchers must make sure that there are equal numbers of mild cases in both groups. Simple and block randomization do not always succeed in equating the groups on all characteristics. In our example, stratified randomization would involve generating two separate block sequences, one for the mild form of the disease and another for the severe form. Although stratification adds credibility to the randomization process, it also adds complexity. Thus, it should be performed only if warranted for clinical or theoretical reasons.
Randomization is a defining characteristic of RCTs that is intended to eliminate confounding bias so that causality can be ascertained. Readers of RCT reports should evaluate whether the method of randomization chosen by the researchers is appropriate to the study.
NurseONE.ca resources on this topic
Crosby, R. A., DiClemente, R. J., & Salazar, L. F. (Eds.). (2006). Research Methods in Health Promotion.
Maltby, J., Williams, G. A., McGarry, J., & Day, L. (2010). Research Methods for Nursing and Healthcare.
Mateo, M. A., & Kirchhoff, K. T. (Eds.). (2009). Research for Advanced Practice Nurses: From Evidence to Practice.