SG-SST indicators under Resolution 0312: calculations, data and automation
Automating an SG-SST indicator starts with agreeing what is counted, where the data comes from and who reviews it. A correct formula can support the wrong conclusion when it mixes periods, excludes workers or changes its denominator each month. Design the data record before designing the dashboard.
The Explora entry on OSH indicators explains the distinction between lagging and leading measures. This article focuses on a narrower task: translating Colombia’s minimum indicators into a data process that can be reviewed and maintained in software.
The six formulas and their minimum frequency
Article 30 of Resolution 0312 of 2019 (in Spanish) establishes these measures. OA means occupational accident and OD means occupational disease. The expressions below summarise the calculations.
| Indicator | Calculation | Minimum frequency |
|---|---|---|
| Accident frequency | OA during the month / workers during the month × 100 | Monthly |
| Accident severity | (OA disability days during the month + charged days during the month) / workers during the month × 100 | Monthly |
| Proportion of fatal OA | Fatal OA during the year / total OA during the year × 100 | Annual |
| OD prevalence | New and existing OD cases during the period / average workers during the period × 100,000 | Annual |
| OD incidence | New OD cases during the period / average workers during the period × 100,000 | Annual |
| Absence for medical reasons | Days absent due to occupational or non-occupational medical incapacity during the month / scheduled working days during the month × 100 | Monthly |
The same provision defines charged days and the population to include. Occupational disease measures count cases recognised as occupational. Document these rules in the configuration, together with the method for consolidating people and periods. Replacing the denominators with hours worked would produce different indicators.
Create a data specification for each indicator
Before importing historical records, assign a source to every variable: accident records, medical absence records, the working population or work schedules. Identify who can correct it and the check required for acceptance. A field named “headcount” is inadequate if some departments include contractors while others include only direct employees.
- Define the identifier used to detect duplicate accidents.
- Separate the event date, entry date and calculation period.
- Document how averages and scheduled working days are obtained.
- Record the reason, author and date of corrections.
- Distinguish a confirmed zero from missing information.
This approach draws on digital OSH evidence: every result needs an understandable route back to its source records. Aggregated dashboards should avoid exposing identifiable clinical information to people who do not need it.
A worked example for testing calculations
In an illustrative example, a business records two OA in a month and validates a population of 200 workers for that period. Frequency is 1 OA per 100 workers. With six OA disability days and no charged days, severity is 3 days per 100 workers. These are different measures and their units must remain visible.
To test absence, assume 22 medically certified absence days and 4,400 scheduled working days: the result is 0.5%. These numbers test the procedure; they are neither Sabentis results nor industry benchmarks. When a denominator is zero, the system should identify a result that cannot be calculated and request review instead of displaying a favourable zero.
Close the period while preserving corrections
Establish a closing sequence: collection, validation by the responsible person, calculation, analysis and approval. If a later correction arrives, retain the earlier version and explain the variation. This makes it possible to distinguish a real improvement from a change in data quality.
Also test a month with joiners and leavers, an absence spanning two months and an accident entered late. These scenarios expose decisions that demonstrations with perfect data tend to hide. Apply the chosen method consistently and check it against the relevant regulatory criteria.
Turn a result into a preventive decision
Indicators require interpretation: exposure, operational changes and reporting quality can explain their movement. Connect the analysis to the annual SG-SST work plan and senior management review. Add leading indicators for Vision Zero to observe preventive action before harm occurs.
When evaluating Sabentis OSH software, use this example to examine sources, permissions, traceability and presentation. Automation adds value when the team can explain both the figure and the decision based on it.
Checklist before releasing the SG-SST dashboard
A short sample-based review can reveal discrepancies hidden by a chart. Use the same cut-off period to reconstruct results and record unresolved data issues before presenting the dashboard to management.
- Select a non-zero result and rebuild it from source records, without treating the existing calculated total as supporting evidence.
- Compare the included population with the period’s validated source and resolve duplicate people or missing employment and service relationships.
- Test a period without events and another with incomplete information; the dashboard should make that distinction clear to its readers.
- Check a late correction and establish which historical series changes, who validates it and how readers are informed.
- Confirm that exported figures retain units, formulas and periods without disclosing diagnoses or unnecessary information about identifiable individuals.
Frequently asked questions
Can a rate based on hours be compared directly with Resolution 0312?
They are different measures when their denominators differ. Keep each indicator’s name, formula and unit visible, and explain its purpose. A corporate rate based on hours may coexist with the Colombian indicator, but should not replace it or be combined with it in a single series.
What should I show if the monthly data is not yet final?
Label the result provisional, identify the cut-off date and state which source still requires validation. Assign an owner and a date for completing the close. Converting missing information to zero or hiding the period would make the underlying trend harder for readers to interpret.
Why can better recording make the chart look worse?
Adding previously omitted cases can increase an indicator without an equivalent deterioration in working conditions. Explain the change in coverage and retain a consistent comparison. Analysis should distinguish changes in preventive performance, data corrections and modifications to the way events or people are recorded.
Should every indicator have a target of zero?
No. A target’s meaning depends on whether it measures harm, delivery, recording quality or verification of controls. Avoid objectives that discourage accident reporting. Pair outcome figures with preventive measures and explain the decisions made in response, rather than treating every downward trend as an improvement.
Legislation and consulted sources
- Resolution 0312 of 2019, article 30. Colombia: reference for minimum indicators, definitions and reporting frequencies. Spanish text.
- Decree 1072 of 2015, articles 2.2.4.6.19–2.2.4.6.22. Colombia: framework for structure, process and outcome indicators within SG-SST. Spanish text.
Sources consulted on 6 October 2026.
