Consistently-measured outcomes

Consider whether outcomes were assessed similarly in the people being compared.

If a possible outcome of a health action is assessed differently in the different groups in a comparison of health actions, differences reported in that outcome may be due to how the outcome was assessed rather than because of the health actions carried out by people in each group.


In a comparison of health actions (things people do for their health), if the people assessing the outcomes believe that a particular health action works and they know which participants took that health action, they may be more likely to record better outcomes in those participants. One way of preventing this is to keep outcome assessors unaware of (“blind” to) which people have been allocated to which health action.

Systematic differences in outcome assessment (“measurement bias”) can make the effects of health actions appear either larger or smaller than they actually are. Blinding is less important for “objective” outcomes, like death, than for “subjective” outcomes, like pain.

Although it is not always possible to blind healthcare providers or participants in randomized trials, it generally is possible to blind outcome assessors. However, for some outcome measures, such as patient-reported outcomes, this is not possible if the patients participating in a trial cannot be blinded. It is also sometimes possible to blind outcome assessors in non-randomized studies. When blinding is not possible, it is important to consider that the results of the comparison may be misleading.


A fair comparison of laser surgery with non-surgical treatment for people with angina (chest pain caused by reduced blood flow to the heart). The severity of angina one year after the treatment was assessed both by investigators who were aware which treatment the participants had received (i.e., unblinded) and by trained interviewers who were not aware which treatment the participants had received (blinded). Comparison of all the researchers’ assessments showed that the non-blinded investigators’ assessments favoured the laser surgery over the non-surgical treatment more often than the blinded interviewers’ assessments did. The researchers estimated that 28% of the apparent angina improvement was due to bias resulting from unblinded outcome assessment.


Remember: Be careful about relying on the results of comparisons if outcomes were not assessed in the same way in the different comparison groups.

Back to Top