Why companies need a global health policy

Executive summary

81% of work-related deaths are caused by disease, not injury (WHO/ILO, 2021). Yet the health data that would show where that risk sits is most often still kept site by site, in the hands of each local medical team, and is not accessible to HSE, HR and management.

As a result, boards receive little to no reporting on employee health, although 83% of C-suite executives say that workforce well-being should be discussed at board level (Deloitte, 2023).

Since medical costs keep increasing, health data is also the strongest cost argument a medical director can make internally. The medical cost trend facing US employers is 9% in 2027 (PwC, 2026), and insurers project 10.3% worldwide for 2026 (WTW, 2026).

In Great Britain, workplace injury and new cases of work-related ill health cost £22.9 billion a year, and 72% of that comes from illness rather than injury (HSE, 2023/24). A health program built around recordable incidents only addresses the remaining 28%.

Where the £22.9 billion goes

Annual cost of workplace injury and new cases of work-related ill health, Great Britain

72%

ill health, £16.4bn, mostly unmeasured on site

28%

injury, £6.5bn, the part incident programs track

This paper looks into how to make health data usable across a company while keeping the medical details confidential. We’ll cover why health data now matters to companies, why it is still scattered across sites and systems, what makes it compliant and usable, and what each function gains from it.

What health data covers

In the context of employee health, health data is usually understood as the clinical record, which is one reason it has been left to doctors. In practice, it includes every record of a health-related activity for every worker, both employees and contractors.

1. Primary care delivered on site
2. Occupational health follow-up: fitness to work, exposure history, surveillance history, sick leaves, statutory reporting
3. Prevention campaigns and vaccinations
4. Case management: medical evacuation, referral, approval, hospitalization follow-up

Each of these is recorded by different teams or providers, often in different tools, which is how one employee’s health ends up scattered across sites and systems.

That breadth is also what makes the subject sensitive. The individual record is confidential and should stay with doctors. Yet without a shared system, medical information often circulates by email, especially when managing cases with external service providers, which is itself a breach of confidentiality.

What HR, HSE and executives need is the same data, but anonymized, aggregated and presented as shares, trends and comparisons, with no names attached.

Read this way, one dataset serves several departments at once. HR sees sick leave, HSE sees fitness-to-work, incidents and exposure groups, and finance sees the full cost of health, well beyond medical spend.

Chapter 1

Why health data is now key for companies

Chapter 2

Health data is siloed by inheritance

Health data has stayed with medical teams for confidentiality reasons. It is some of the most private information a person shares, and they share it with a doctor in confidence.

Each site has also long worked with its own methods and tools: on paper, in spreadsheets, in software installed locally. Across a company’s sites, there is usually no common standard for how health information is collected. A worker’s record is complete only on the site where it was made.

Local records are not comparable across sites or over time, and local systems were not built to give different people access to different parts of a record. Currently, most of the time, the data they hold cannot be used beyond the clinic without breaching confidentiality.

Part of the picture sits outside the clinic altogether. HR records sick leaves. HSE records incidents. Both are health-related, but they are kept in systems separate from the medical record, so none of it can be read together.

From any single site, fragmentation rarely looks like a problem. Each site can be fully compliant with local rules. The gaps appear at group level. A missing entry at one site and a late one at another look minor on their own, but regulators add them up across sites, and a company holding its records in separate systems cannot.

Chapter 3

Health data governance

Using health data beyond the clinic depends on the system that holds it. Collecting it the same way on every site and storing it centrally is a requirement for both compliance and accessibility.

Three conditions for compliant use of health data

1. 01Country rules metOne record per employee, processed under each country's data protection and security rules.
2. 02Role-based accessPhysicians work in the full record. HR, HSE and executives see anonymized, aggregated data.
3. 03A log of every accessEach view of a record is recorded, so the question of who saw what, and when, has an answer.

What any analysis can find is decided when the data is entered. A record that is incomplete on the day of a consultation stays incomplete, however skilled the team that wrote it: what the system asks for, and how, sets the limit. Four requirements for usability:

1. Structured. Coded vocabularies such as ICD-10, and constrained fields wherever a value should be constrained. Free text can be analyzed with natural language processing, but only if it is written consistently. Structure also means fewer platforms, with health data kept in one system rather than spread across several.
2. Complete, and visibly incomplete when it is not. In primary care, clinically important information was missing in 13.6% of visits, about one in seven, and judged likely to harm the patient in 44% of those cases (Smith et al., JAMA, 2005).
3. Accurate at source. Each result goes through as few steps as possible, for example taken directly from the medical device that performed the test rather than retyped.
4. Consistent over time. Workers are often followed for decades, so testing techniques, equipment and interpretation have to stay the same over that period (ACOEM, JOEM, 2020).

The same rules have to apply on every site. With one record per employee, and medical activities recorded the same way everywhere, results can be compared across sites and over time. A baseline is only useful if everything measured after it was measured the same way.

Data that meets these conditions can be analyzed at a scale no single site could reach. A common reason AI projects fail is poor data quality: garbage in, garbage out. Consistent standards for data collection, across sites and over time, are a prerequisite for confidence in the reports.

AI research shows what’s possible once data is clean and structured across many workplaces:

• 83,294 OSHA injury narratives were used to train a deep learning model. The model predicted injury severity with 0.95 accuracy for hospitalization and 0.98 for amputation, and surfaced the causes described in the free text (Khairuddin et al., PeerJ Computer Science, 2024).
• 1.2 million Ohio workers’ compensation claims were coded by machine learning. Industries were ranked by the claims that ergonomic or slip, trip and fall prevention could have avoided (JOEM, 2018).

Chapter 4

Benefits by role

Medical teams: better care through a longitudinal view

A result is read against the employee’s own history since hire, across transfers and providers.

In spirometry, comparing a worker to their own baseline detects excessive loss of lung function earlier than comparing a single test to population norms, and a decline of 15% or more from that baseline is a reason to refer (ACOEM, JOEM, 2020). Hearing works the same way: a standard threshold shift, an average change of 10 dB or more at 2,000, 3,000 and 4,000 Hz against the worker’s baseline audiogram, can only be detected when that baseline follows the worker (OSHA, 29 CFR 1910.95).

Reading a worker against their own baseline

Hire10 years20 years30 years

Lung functionOwn baseline15% below: refer

Continuity also changes outcomes: in a nationwide Danish cohort, patients who had fewer than 25% of their contacts with their usual provider had a mortality rate 2.6 times higher, and 49% more potentially inappropriate medication, than patients with full continuity (Prior et al., BMC Medicine, 2023).

Chief medical officers: global health policy, steered from evidence

Coverage and testing status are current on any given day, on every site, so it is clear where a program is working and where it is not. Over time, policy can be set from evidence about the workforce, and care can move from reacting to cases to preventing them. Clinical detail stays with the physicians, and the access log shows who saw what, and when.

Who saw what, and when

Role	View	Logged
Site physician	Full record	Yes
Chief medical officer	Coverage by site	Yes
HSE lead	Anonymized trends	Yes

Employees: care is safer

Every physician who sees them decides on their full history rather than part of it. They no longer have to repeat the same information at each consultation. Their confidentiality is protected by how the system is built, and they can access part of their own record directly.

One record that follows the worker

1. HireSite A
2. TransferSite B
3. TodaySite C

HSE: fewer incidents, with prevention aimed at the gaps

Many incidents start with a health condition that a physician knew about but the people organizing the work could not see. Aggregated across sites, the same data shows HSE where the workforce stands against its own policies and procedures: cardiovascular risk factors in roles with heavy physical load, reduced lung function among workers exposed to dust or fumes, hearing loss in high-noise areas.

Where the workforce stands against policy

Indicator	Site A	Site B	Site C
Cardiovascular risk factors roles with heavy physical load	Within policy	Gap	Within policy
Reduced lung function dust and fume exposure groups	Gap	Within policy	Within policy
Hearing loss high-noise areas	Within policy	Within policy	Gap

Individual decisions can also be shared without the diagnosis behind them. One approach is to separate the fitness-to-work outcome from its medical basis: HSE sees that a worker is fit with a restriction, such as no work at height, while the reason stays in the physician’s record.

HR: higher workforce availability

Anonymized sick leave trends by site and group of employees show where absence is building, so it can be addressed before it becomes long term. Early patterns matter: in a Danish register study of young employees, those with frequent short absences in their first year of full-time work had a higher risk of long-term sickness absence over the following two years (Sørensen et al., BMJ Public Health, 2025).

First year of full-time work

Frequent short absences

→

Following two years

Higher risk of long-term sickness absence

Sick leave is also one of the hardest figures to compare. Medical teams, HR, contractors’ employers and insurers each record it with their own definitions, and sick pay rules differ from one country to the next. Comparing one site with another can mislead. Following each site over time is more reliable, because its definitions stay the same.

Executives: health becomes a measure of performance

Medical costs stop being a line that only goes up, because the company can see what drives them: incidents, sick leave, exposure. Health indicators can be managed centrally, the way financial data already is. Better employee health could generate $3.7 trillion to $11.7 trillion in global economic value, 4% to 12% of global GDP, with 54% to 77% of it coming from productivity (World Economic Forum and McKinsey Health Institute, 2025).

Global economic value of better employee health

$3.7 to $11.7 trillion

$0$3T$6T$9T$12T

4% to 12%

of global GDP

54% to 77%

of that value coming from productivity

In practice, that means a short set of indicators reviewed alongside the rest of the business. The key results in this chapter are a starting point:

• Program coverage by site: the share of workers whose fitness-to-work, surveillance and vaccinations are up to date
• Health-related incidents: incidents with a health cause, and fitness-to-work restrictions by job position
• Workforce availability: days lost to work-related ill health, the figure large EU companies already publish under the CSRD
• Cost of ill health: medical spend, sick leave and replacement, in one figure

Each indicator is only as reliable as the records behind it, which is why the conditions in chapter 3 come first.

Chapter 5

Conclusion

Health data becomes usable when there is one record per employee, not one per site: collected in the doctor's hands and read across the company. That is the principle behind MyC, and it comes from practice.

Dr. Laurent Bonnardot spent more than twenty years practicing medicine in some of the world’s most isolated places, and served as medical director for companies on remote industrial sites, before founding MyC.

MyC holds one health record per employee, covering occupational health and primary care: fitness to work, sick leave, consultations, case management and pharmacy. Medical teams work in the full record. HR, HSE and executives work from anonymized trends. The platform is certified to ISO 27001 and NEN 7510, and complies with HIPAA and the GDPR.

Because every activity is recorded the same way on every site, the data can be analyzed beyond reporting to identify early risks, such as:

• Sick leaves recurring in the same crew, an early sign of burnout
• Repeated consultations for the same symptoms on one site
• Fitness-to-work restrictions piling up in one job position

The conditions described in this paper are rarely met all at once. Instead, it’s a gradual evolution. The three levels below map where companies stand and what to work on next.

Three levels of maturity

1. Fragmented. Health information and practices depend on individuals, local routines or scattered records. Start with a shared process and a consistent way to record information.
2. Structured. Repeatable processes and maintained records support your health program, but coordination or oversight still depends on manual work. Focus on shared visibility and consistent follow-through.
3. Governed. Shared information, clear responsibilities and regular reviews support decisions across sites. Sustain these practices and check that improvements lead to better outcomes.

The first level matches the situation described in chapter 2: records that depend on each doctor and each site. The second level is where data quality at entry starts to hold. The third is where health data is used across sites, and where prevention can be aimed at the risks that data shows.

At that level, health data stops being a record of what has already happened to employees. It becomes a way to change what happens next.

References

1. WHO/ILO, joint estimates of the work-related burden of disease and injury, 2021.
2. Deloitte, Workplace well-being research, 2023.
3. HSE, Costs to Great Britain of workplace injuries and new cases of work-related ill health, 2023/24.
4. PwC, Medical cost trend: Behind the numbers 2027, 2026.
5. WTW, 2026 Global Medical Trends Survey.
6. GDPR, Regulation (EU) 2016/679, Art. 9.
7. LGPD, Law 13.709/2018, Art. 11 (Brazil).
8. HHS, Employers and health information in the workplace.
9. OSHA, 29 CFR 1910.1020, Access to employee exposure and medical records.
10. Decree on electronic data processing by care providers, Stb. 2017, 446 (Netherlands).
11. European Commission, Certification of EHR systems under the European Health Data Space Regulation (EU) 2025/327.
12. OSHA, national news release, 2022.
13. Johnson, American Economic Review, 2020.
14. CSRD, ESRS S1, Commission Delegated Regulation (EU) 2023/2772.
15. EU-OSHA and ILO, 2017.
16. ISSA and DGUV, Return on prevention, 2011.
17. Directive (EU) 2022/431, amending Directive 2004/37/EC, Art. 15.
18. Smith et al., JAMA, 2005.
19. ACOEM, Spirometry in occupational health, JOEM, 2020.
20. OSHA, 29 CFR 1910.95, Occupational noise exposure.
21. EU-OSHA, Cancer and work.
22. EUROGIP, occupational cancers.
23. Khairuddin et al., PeerJ Computer Science, 2024.
24. JOEM, 2018.
25. Prior et al., BMC Medicine, 2023.
26. Sørensen et al., BMJ Public Health, 2025.
27. World Economic Forum and McKinsey Health Institute, 2025.