Reasons for inaccurate classifications of the regular expression (RE)-based algorithm
| Classified by the RE algorithm* | Classified by manual prescription review* | Reasons for inaccurate classifications†|
| Tier 1B | Tier 1A | Multiple diagnoses incorrectly concatenated together (n=1). |
| Tier 2A | Tier 1A | The infectious disease written after the fifth diagnosis (n=2). |
| Tier 2A | Tier 1B | The infectious disease written after the fifth diagnosis (n=1). |
| Tier 2B | Tier 2A | Multiple diagnoses incorrectly concatenated together (n=4). |
| Tier 3 | Tier 1A | Single diagnosis improperly split (n=1); The infectious disease written after the fifth diagnosis (n=3). |
| Tier 3 | Tier 2A | Multiple diagnoses incorrectly concatenated together (n=5); Single diagnosis improperly split (n=2); The infectious disease written after the fifth diagnosis (n=10); Traditional Chinese used (n=1). |
| Tier 3 | Tier 2B | The infectious disease written after the fifth diagnosis (n=1); |
*Tier 1A: tier 1 diagnoses without uncertainty. Tier 1B: tier 1 diagnoses with uncertainty. Tier 2A: tier 2 diagnoses without uncertainty. Tier 2B: tier 2 diagnoses with uncertainty.
†There was a visit (classified as tier 3 by computer and tier 2A by manual review) in which one diagnosis was divided into two diagnoses and the second part was written together with another one; thus, the total number of incorrect classification for different reasons was 31 in this table.