Three rates, three different questions
Three measures are routinely mixed up in accident statistics. All three use the same set of accidents; only the denominator changes.
- Frequency rate — accidents per one million working hours. Because it measures exposure in hours, it is the least distorted by part-time work and shift patterns.
- Incidence rate — accidents per 100 workers. It describes risk per person and is used when working hours are unknown.
- Severity rate — lost working days per one million working hours. It shows how heavy the accidents were, not how many there were.
The values SGK published for 2025
| Period | Accidents | Frequency rate | Incidence |
|---|---|---|---|
| January-April | 219,590 | 13.39 | 3.01 |
| May-August | 281,415 | 17.19 | 3.87 |
| September-December | 265,034 | 15.59 | 3.51 |
| Year total (4/a) | 766,039 | 15.39 | 3.46 |
| Year total (4/b) | 586 | 0.07 | 0.01 |
In the same table the severity rate is 533.30 for 4/a and 12.77 for 4/b. Fatal occupational accidents in the year number 1,931 under 4/a and 13 under 4/b.
Calculating the rate for your own workplace
The denominator is actual working hours: annual leave, sick leave and overtime all count. On sites with heavy subcontractor presence, leaving subcontractor hours out makes the rate look better than it is. When two periods of the same year are compared, the denominator must be that period's hours; a period's accidents are never divided by the annual hours.
The unit trap
Frequency rates cannot be used directly for country comparisons. ILOSTAT and Eurostat publish rates per 100,000 workers, US labour statistics (BLS) per 100 full-time workers, while SGK's frequency rate is per one million working hours. For example, Türkiye's fatal injury rate is 11.47 (per 100,000 workers, 2024) while the US total recordable case rate is 2.6 (per 100 full-time workers, 2024) — these two numbers are not on the same scale.
All period rates and the sector breakdown are on the Türkiye OHS statistics page; for international rates use the global data and the statistics hub.