Randomization
To be sure that the groups are as similar as possible, patients are put into each study group using an allocation method that is unbiased (no personal preference on anyone’s part is allowed). Random allocation is the best way to do this. Researchers use random number tables, coin tosses, or other, similar methods and set up allocation sequences for patients. This can be done before the study starts or done during the study itself. When a patient agrees to be part of the study and is ready to enter it, they get the next allocation, often in a sealed envelope or via a telephone call to the study office. Allocation by birth date, patient chart number, or day of week seen in the clinic is not random. These methods are, however, better than coin tosses or clinician or patient preference to ensure comparability of groups, as long as the allocation procedure is strictly adhered to; often however, it is not.
Randomization can be done with “parts” of people (for example half of head washed with shampoo A and half with shampoo B to assess softness or the ability to kill head lice, or arms randomly allocated to intravenous insertion of a catheter using the standard landmark guidance procedure or ultrasonography). Most often persons are allocated, but towns, wards, schools, hospitals, bus stops, and so on have been allocated. The research question often dictates the unit randomized.
Groups do not need to be the same size. Often the sizes are similar, but randomization does not guarantee exactly equal numbers. Indeed, some studies purposely involve unequal numbers. For example, Maizel et al.2 studied intranasal lidocaine compared with placebo for migraine pain relief. Twice as many people (58:28) were allocated to receive the intranasal lidocaine as were allocated to receive placebo (intranasal saline solution). These unequal groups were formed because of the ethics and common sense issues of withholding medication from persons with migraine headache pain. The analysis at the end of the study can factor and adjust for the difference in group size.5475
Content on this page was last changed on March 19, 2009.
© 2002 BC Decker Inc. Show Disclaimer
| 5475. | McKibbon, A. PDQ Evidence-Based Principles and Practice, 1999, BC Decker Inc., Hamilton, Ontario. |