Table 3.
By algorithm rule: agreement between algorithm- and expert-assigned modules, with an assessment of the degree of information loss when the two approaches were discordant.
| Proportion of jobs in strata (%) | ||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Exposure screening question with yes responsea | No. of occupation keywords identified | No. of industry keywords identified | Algorithm rule numbers | No. of jobs | % of all jobs | Exact match | No info. loss or INDSOL/ SOL discordance | Low, info. loss | Medium or high info. loss | |||||||||
| Yes | 0 | 0 | 1 | 100 | 9.0 | 58 | 30 | 5 | 7 | |||||||||
| No | 0 | 0 | 2 | 311 | 32.5 | 48 | 32 | 5 | 15 | |||||||||
| — | 0 | 1 | 3 | 229 | 12.4 | 54 | 30 | 1 | 15 | |||||||||
| — | 1 | 0 | 4 | 265 | 20.5 | 67 | 18 | 1 | 14 | |||||||||
| — | >1 | 0 | 5–8 | 26 | 1.7 | 8 | 23 | 4 | 65 | |||||||||
| — | 0 | >1 | 9–12 | 15 | 0.9 | 40 | 20 | 0 | 40 | |||||||||
| — | 1 | 1 | 13 | 241 | 19.8 | 73 | 10 | 2 | 15 | |||||||||
| — | 1 | >1 | 14–18 | 25 | 1.2 | 36 | 24 | 4 | 36 | |||||||||
| — | >1 | 1 | 19–22 | 34 | 1.4 | 50 | 21 | 6 | 24 | |||||||||
| — | >1 | >1 | 23–24 | 4 | 0.4 | 50 | 0 | 25 | 25 | |||||||||
| — | Missing | Missing | Missing | 1 | 0.3 | 100 | 0 | 0 | 0 | |||||||||
| Overallc | Reliability subset | 1251 | 58 | 23 | 3 | 16 | ||||||||||||
| All jobs (estimated) | 11409 | 68 | 18 | 2 | 12 | |||||||||||||
aScreening questions were not used except in rules #1 and #2. No indicates that a ‘no’ response was received to all three exposure screening questions. Yes indicates that at least one screening question had a ‘yes’ response.
bSee Supplementary Table S1, available at Annals of Occupational Hygiene online for more detail on each rule’s criteria and resulting action.
cSee Table 2. Provided here for comparison purposes.