Table 4.
Overview of the approaches to handling missing data within the identified RCTs by PROM category
| Questionnaires | EQ-5D-3L index | HUI | OHS | OKS | PDQ | QLQ-C30 | SF-12 | SF-36 | Overall |
|---|---|---|---|---|---|---|---|---|---|
| Number of studies | 72 | 13 | 4 | 9 | 17 | 21 | 25 | 76 | 237 |
| Methods to limit missing data described (%) | 25.0 | 15.4 | 50.0 | 22.2 | 11.8 | 14.3 | 36.0 | 21.1 | 22.8 |
| Differential missingness assessed (%)a | 25.0 | 15.4 | 0 | 11.1 | 11.8 | 14.3 | 28.0 | 18.4 | 19.8 |
| Assumed missing data mechanism | |||||||||
| Not described (%) | 91.7 | 100 | 100 | 100 | 82.4 | 100 | 88.0 | 96.0 | 93.7 |
| Missing at random (%) | 6.9 | – | – | – | 17.6 | – | 12.0 | 4.0 | 6.3 |
| Missing completely at random (%) | 1.4 | – | – | – | – | – | – | – | 0.42 |
| Missing data mentioned in methods/analysis section (%) | 62.5 | 53.9 | 25.0 | 11.1 | 75.0 | 42.9 | 52.0 | 52.6 | 54.2 |
| Analysis population | |||||||||
| Intention to treat (%) | 27.8 | 7.7 | – | 11.1 | 29.4 | 9.5 | 24.0 | 19.7 | 21.1 |
| Modified intention to treat (%) | 54.2 | 46.2 | 50.0 | 66.7 | 47.1 | 59.1 | 48.0 | 46.1 | 50.6 |
| Per protocol (%) | 1.4 | – | – | – | 5.9 | – | – | 1.3 | 1.3 |
| Unclear (%) | 16.7 | 46.2 | 50.0 | 22.2 | 17.7 | 33.3 | 28.0 | 32.9 | 27.0 |
| Primary method of handling with missing data | |||||||||
| Complete cases (%) | 38.9 | 30.8 | 50.0 | 22.2 | 5.9 | 14.3 | 32.0 | 39.5 | 32.9 |
| Last observation carried forward (%) | 11.1 | 7.7 | – | 11.1 | 41.2 | 9.5 | 4.0 | 10.5 | 11.8 |
| Mean imputation (%) | 5.6 | – | – | – | – | – | 4.0 | 2.7 | 3.0 |
| Regression imputation (%) | – | – | – | – | – | – | 4.0 | – | 0.4 |
| Direct likelihood analysis (%) | – | – | – | – | 5.9 | – | – | – | 0.4 |
| Repeated measures model (%) | 8.3 | 15.4 | – | 11.1 | 17.7 | 14.3 | 20.0 | 25.0 | 16.9 |
| Multiple imputation (%) | 15.3 | 15.4 | – | – | – | – | 16.0 | 5.3 | 8.9 |
| Unclear (%) | 20.8 | 30.8 | 50.0 | 55.6 | 29.4 | 61.9 | 20.0 | 17.1 | 26.2 |
| Justification provided for primary method of dealing with missing data (%) | 13.9 | 15.4 | 25.0 | 0 | 11.8 | 0 | 8.0 | 5.3 | 8.9 |
| Sensitivity analysis was performed (%) | 25.0 | 23.1 | 25.0 | 0 | 17.7 | 19.1 | 32.0 | 19.7 | 21.9 |
| Potential influence of missing data on results mentioned in discussion (%) | 18.1 | 15.4 | 25.0 | 0 | 17.7 | 14.3 | 16.0 | 14.5 | 15.6 |
aThe studies considered differences between those with complete and missing data in terms of participant (baseline) characteristics