Data analytics in OSH

What is data analytics in OSH

Preventive management generates a vast amount of data: accident reports, incident reports, assessments and their revisions, plans with measures and deadlines, inspections and observations, training and equipment delivery records, collective health surveillance results, noise, chemical agent, and temperature measurements, and an increasing amount of data from sensors, wearable devices, and production systems. Traditionally, this data has been used to fulfill documentation obligations and calculate accident rates; data analytics transforms it into actionable knowledge.

A distinction is usually made between descriptive analytics (what happened: indices, trends, distributions by center, shift, task, or agent), diagnostic analytics (why: correlations between conditions, exposures, and damages; root cause analysis), predictive analytics (what might happen: models that estimate the probability of incidents or non-compliance based on leading indicators), and prescriptive analytics (what to do: prioritization of measures and allocation of resources). The first two are achievable with well-recorded data and simple tools; the latter two require volume, quality, and validated models.

The value of analytics depends on the quality of the source data: consistent definitions, coherent coding (e.g., form and material agent of the accident report), complete recording of incidents, not just accidents resulting in lost time, and exposure data (hours worked, staffing levels, tasks) that allow for the calculation of rates. Preventive management software that centralizes records is the typical foundation for analytics.

Common uses

• Dashboards. Outcome indicators (accident rates, absenteeism due to occupational contingencies ) and proactive indicators (reported incidents, inspections, measures implemented on time, current training, safety observations) by center and period.
• Accident analysis. Breakdown by type of accident, task, time, seniority, contract and shift to identify patterns and risk areas.
• Lead indicators. Use of incidents, unsafe conditions, near misses, and nonconformities as warning signs before accidents.
• Prioritization of planning. Ordering of pending measures according to risk, cost, timeframe and recurrence of deficiencies.
• Hygiene and ergonomics. Time series of measurements and sensor data to detect anomalous exposures and evaluate the effectiveness of the measures.
• Public health surveillance. Epidemiological analysis of anonymized or aggregated results by health personnel to link exposures and harms.
• Predictive models. Estimation of the probability of an incident by position or period, with validation and human supervision, as support and not as a substitute for technical judgment.

Requirements and limits

• Data protection. Health data is a special category according to the General Data Protection Regulation and Organic Law 3/2018; its analytical processing requires a legal basis, minimization, anonymization or aggregation and restricted access to healthcare personnel.
• Confidentiality and non-discrimination. The results cannot be used for adverse individual decisions or to profile workers; Law 31/1995 limits the use of health surveillance data to preventive purposes.
• Quality and biases. Models reproduce data biases: underreporting of incidents or overrepresentation of certain centers distorts the conclusions.
• Technical interpretation. The analytics support the decision of the prevention technicians, management, and representatives; correlations are not causes, and the models must be validated.
• Participation. Worker safety representatives must be informed and consulted about the analysis systems and their purposes.

Organizational application: how to implement data analytics

1. Define the questions that the analytics should answer and the associated indicators, with the participation of management, the prevention service and the representatives.
2. Inventory data sources, unify definitions and codings, and centralize records in a preventive management system.
3. Ensuring quality: complete incident log, exposure data (hours, template), mandatory fields and validations.
4. Build descriptive dashboards by center and period, with outcome and proactive indicators, before tackling advanced models.
5. Establish the data protection framework: risk analysis or impact assessment where appropriate, anonymization or aggregation of health data and access by role.
6. Incorporate the analysis into the management routine: monthly review with managers, health and safety committee and management review.
7. Validate any predictive model with historical data, document its limitations, and keep the final decision in the hands of people.

Preventive management software with analytical capabilities allows you to exploit the recorded data without manual exports, maintain the traceability of each indicator to its origin, and apply differentiated access controls to health data.

Limits and common mistakes

1. Analyzing only accidents with lost work time, ignoring incidents and unsafe conditions, which are the majority of preventive information.
2. Compare centers or periods without normalizing for exposure or registration criteria.
3. Processing identifiable health data outside of healthcare personnel or using it for decisions about people.
4. Relying on predictive models without validation or explanation, or replacing risk assessment with them.
5. Building dashboards that no one reviews or connects to decisions and measures.
6. Forgetting to consult with representatives and inform staff about analysis systems.

Analytics does not replace risk assessment or health surveillance: it complements them with evidence to make better decisions.

Regulatory and reference framework

• Law 31/1995, articles 16, 22 and 23. Assessment and investigation of damages, confidentiality of health surveillance and documentation.
• Regulation (EU) 2016/679, General Data Protection Regulation . Processing of personal data and special categories of data, including health data.
• Organic Law 3/2018, of December 5. Protection of personal data and guarantee of digital rights.
• ISO 45001. Chapter 9: Monitoring, measurement, analysis and evaluation of occupational health and safety performance.
• European Agency for Safety and Health at Work (EU-OSHA) . Digitalization of work: opportunities and risks of data-based systems for safety and health.

The Spanish Strategy for Safety and Health at Work 2023-2027 includes the improvement of knowledge and information systems as a line of action; in Colombia, Resolution 0312 of 2019 requires structure, process and outcome indicators in the SG-SST.