7.7.4  Data extraction for ordinal outcomes

Ordinal data, when outcomes are categorized into several, ordered, categories, are described in Chapter 9, Section 9.2.4, and their meta-analysis is discussed in Chapter 9, Section 9.4.7. The data that need to be extracted for ordinal outcomes depend on whether the ordinal scale will be dichotomized for analysis (see Section 7.7.2), treated as a continuous outcome (see Section 7.7.3) or analysed directly as ordinal data. This decision, in turn, will be influenced by the way in which authors of the studies analysed their data. Thus it may be impossible to pre-specify whether data extraction will involve calculation of numbers of participants above and below a defined threshold, or mean values and standard deviations. In practice, it is wise to extract data in all forms in which they are given as it will not be clear which is the most common until all studies have been reviewed, and in some circumstances more than one form of analysis may justifiably be included in a review.


Where ordinal data are being dichotomized and there are several options for selecting a cut-point (or the choice of cut-point is arbitrary) it is sensible to plan from the outset to investigate the impact of choice of cut-point in a sensitivity analysis (see Chapter 9, Section 9.7). To do this it is necessary to collect the data that would be used for each alternative dichotomization. Hence it is preferable to record the numbers in each category of short ordinal scales to avoid having to extract data from a paper more than once.  This approach of recording all categorizations is also sensible when studies use slightly different short ordinal scales, and it is not clear whether there will be a cut-point that is common across all the studies which can be used for dichotomization.


It is also necessary to record the numbers in each category of the ordinal scale for each intervention group if the proportional odds ratio method will be used (see Chapter 9, Section 9.2.4).