Table 4.
Methods and results of the validation of the outcomes reported in the original studies
| Outcome and study authors | Validation method | # case validations completed/# case validations attempted | % of cases confirmed |
| Anxiety | |||
| Martín-Merino et al, 201018 | GP questionnaire | 135/140 | Among pharmacologically treated: 73.5%; Among not pharmacologically treated: 89.6%. |
| Meier et al, 20042 | Record review | nr/nr | nr |
| Depression | |||
| Becker, 2011 | Sensitivity analysis with different definitions | nr/nr | nr |
| Hagberg, 2016 | Record review | nr/nr | nr |
| Martín-Merino et al, 201018 | GP questionnaire | 135/140 | 89.6% |
| Meier et al, 20042 | Record review | nr/nr | nr |
| Yang et al, 200319 | Record review | 30/nr | 83.3% |
| Anxiety and depression (composite outcome) | |||
| John, 2016 | Compared 12 EHR algorithms to results of the Mental Health Inventory , a subscale of SF-36 | 2799* | Between 61% and 76%, depending on the algorithm. |
| Dementia/cognitive impairment | |||
| Imfeld et al, 2013 and Imfeld et al, 201522 23 | GP questionnaire | nr/120 | Alzheimer’s disease: 79%; Vascular dementia: 74%. |
| Dunn et al, 200524 | GP asked to confirm diagnosis | 50/200 | 100% |
| Dunn et al, 200526 | GP questionnaire | 95/~100 | 83% |
| Strom et al, 201525 | GP questionnaire | 86/100 | 88.4% |
| Strom et al, 201525 | Review of free text | 1047/1048 | 1.5% patients excluded as not having the diagnosis; 42.4% confirmed as having definite memory loss, 36.8% possible memory loss, 3.2% undetermined and 16.0% unknown. |
| Pain | |||
| Hall et al, 20134 | GP questionnaire | 48/54 | 56% |
| Mansfield et al, 201727 | EHR data linked to self-reported pain status collected by postal questionnaire | 1780* | 97% |
| Becker, 2008 | GP questionnaire | 176/200 | 86.4% |
| Self-harm | |||
| Thomas et al 201329 | Comparison of cases of suicide and self-harm identified in CPRD with Read codes, with the cases identified in CPRD data linked to HES data, and published self-harm incidence data. | 74236* | 68.4% |
| Suicide (attempted and completed) | |||
| Hagberg, 2016 | Record review | nr/nr | nr |
| Haste, 1998 | GP asked to confirm suicides | 77% of uncertain deaths/nr | 82% |
| Jick, 1995 | Record review | nr/nr | nr |
| Meier et al, 20042 | Record review | nr/nr | nr |
| Schuerch, 2016 | Outcomes identified in CPRD were compared with those identified in CPRD linked to HES and ONS data. | nr/nr | Compared with CPRD data, the frequency of the outcomes in linked data was approximately three times higher. |
| Yang et al, 200319 | Record review | 30/nr | 83.3% |
| Suicide (completed) | |||
| Arana, 2010 | GP questionnaire and record review | nr/132 | 97% |
| Arana, 2010 | GP questionnaire and record review | nr/86 | 87% |
| Hall, 200928 | GP questionnaire and record review | 33/33 | 21.2% |
| Thomas et al, 201329 | Comparison of cases of suicide and self-harm identified in CPRD with Read codes, with the cases identified in CPRD data linked to ONS mortality data, and national suicide rates. | 1767* | 59.7% for men; 46.0% for women. |
*Validation attempted and completed for all patients identified in electronic health records database.
CPRD, Clinical Practice Research Datalink; EHR, electronic health record; GP, general practitioner; HES, Hospital Episode Statistics; nr, not reported; ONS, Office for National Statistics; SF-36, 36-item Short Form Health Survey.