Occupational health

GDPR and security in occupational health

GDPR for occupational health in practice: controller or processor, what the employer may see, data processing agreements, EU/EEA storage and erasure.

By PortwayPublished Updated 10 min read

GDPR, security, occupational health, patient data, encryption

In this article
  1. The two roles, and why they get mixed up
  2. What the employer gets to see
  3. Where the data is stored
  4. The security requirements that should be a given
  5. Data minimization and weeding: where most fall short
  6. Checklist for procuring a system
  7. How we do it in Portway
  8. Sources
  9. Frequently asked questions

Short answer: The question that decides everything else is not encryption but roles. In the same collaboration you are a data processor for the client company for orders and billing basis, and yourself the data controller as a care provider for the care documentation. That means you need a data processing agreement for the one part and none at all for the other, and the employer should never reach the second. Encryption, logging and storage within the EU/EEA are requirements you should set, but a system that meets them and still shows the wrong information to the client company is not compatible with the rules.

Occupational health handles health data, sick leave and rehabilitation cases. These are sensitive personal data under Article 9 of the GDPR, and processing is prohibited as a main rule, allowed only on the basis of an exception. For you the exception is normally health care purposes; for the employer it is instead a matter of obligations under employment law. These are two different legs, and they carry different weight.

The two roles, and why they get mixed up

This is the most common shortcoming in the sector, and it shows most clearly when someone asks the question in a procurement.

You as processor. When a client company orders a health examination for a named employee, you process the information on its behalf. The client company decides the purpose; you carry it out. That needs a data processing agreement under Article 28, and your system supplier becomes in turn a sub-processor.

You as controller. Once the employee is a patient with you, you are a care provider. Record-keeping follows the Patient Data Act, the purpose is determined by the care provider, and the client company has no say over that processing. Here you are the controller, and a data processing agreement with the client company would be plainly wrong.

Practical consequences of keeping them apart:

  • The client company can request its order data. It cannot request medical records for its employees.
  • An erasure request from an employee cannot reach the record, which has statutory retention periods.
  • A processing agreement worded to cover "all personal data in the service" is wrong, and a well-read procurement officer notices.

What the employer gets to see

The simplest rule, and the one that holds: the employer gets to know that something has been ordered and carried out, never what emerged.

Concretely, the client company sees that an employee has a booked time for a service covered by the agreement, and that it has been delivered. The client company does not see the reason, earlier complaints or the assessment.

Two places where it tends to leak:

  • Free-text fields. A comment field on the order that is visible to the customer sooner or later fills with something that should not be there. The solution is two separate fields, internal notes and a customer-visible thread, not an instruction to write carefully.
  • Client users without scoping. If everyone who orders at the client company sees the whole company's orders, they also see which colleagues are going to counseling. Department scoping is a privacy issue, not a convenience. We go through the roles in the requirements list for ordering portals.

In rehabilitation the boundary becomes sharper, not blurrier. The employer needs to know which adjustments are required for the employee to be able to work, not the diagnosis behind it. A medical statement that goes to the employer should be worded in terms of function and limitations. We go through the flow day by day in the sick and recovery process.

Where the data is stored

The requirement in practice is storage within the EU/EEA. What makes third-country transfers difficult is not that they are forbidden but that they need a legal basis that can be produced on request, plus an assessment of the destination country's legislation, work most providers do not want to do and customers do not want to review.

So be concrete and do not promise more than you deliver. A claim that all data sits in one particular country is a strong one in a procurement, and a bid that says it without it being true is a problem later. If your data sits with a cloud provider in the Netherlands or Ireland, that is the EU/EEA and perfectly workable: say so.

The question to put to a system supplier is not "do you have Swedish hosting" but:

  • In which country are the production database and the backups?
  • Which sub-processors exist, and where do they process data? Support and troubleshooting count.
  • Does any processing take place outside the EU/EEA, and if so on what basis?

The sub-processor question is the one that most often reveals something. A supplier with production in the EU can still have a support function that reaches data from a third country.

The security requirements that should be a given

These are not what separates suppliers, but their absence disqualifies:

  • Encryption in transit and at rest. TLS 1.2 or later for transport, AES-256 or equivalent for storage.
  • Role-based permissions, granular enough to tell a case handler, a practitioner and the client company's users apart.
  • Logging of access to sensitive information, searchable afterwards. In a supervisory review by the Swedish Authority for Privacy Protection (IMY) it is the logs that show what actually happened.
  • Backups that have been tested. A copy nobody has tried to restore is a hope.
  • A routine for personal data incidents. Notification to IMY must be made without undue delay and, where feasible, within 72 hours (IMY, in Swedish, retrieved 20 October 2026). Whoever is going to call you at 11 pm on a Friday needs to know that this is the routine that applies.

Data minimization and weeding: where most fall short

The requirements above are easy to buy. What takes work is not accumulating information, and getting rid of it.

Do not collect more than the purpose requires. An order needs to know who, which service and when. It does not need to know why. A form with a "background" field turns an administrative task into health data.

National identity numbers only where needed. In the record they are needed. On a participant list for an ergonomics course they are not.

Different retention periods for different things. The record must be kept for at least ten years under the Patient Data Act, chapter 3, section 17 (in Swedish). Orders and billing basis fall under the bookkeeping rules. A health survey answered for a one-off examination has no reason to stay for years. A system with only "keep everything" makes weeding a manual project that never gets done.

Anonymize statistics for real. Aggregated absence statistics for a safety committee are good and useful. Aggregated statistics on a department of four people are personal data with an extra step. Set a lower limit for group size before figures are shown, and hold it.

Checklist for procuring a system

  • Where are production data and backups, country by country?
  • Which sub-processors exist, and does any of them reach data outside the EU/EEA?
  • Is there a processing agreement that separates order data from care documentation?
  • Can permissions be scoped per client company and per department?
  • Is access to sensitive information logged, and is the log searchable?
  • Is it possible to set different retention times for different data types?
  • Is there a documented incident routine with the 72-hour requirement built in?
  • Can the supplier show what the client company sees, not just describe it?

The last question is the most useful. Ask for a screen, not an answer.

How we do it in Portway

Portway is not a record system. Orders, bookings, surveys and sick and recovery reporting live in Portway; the record stays in your record system.

  • Scoping per role. The client company's administrator sees the whole client company, a department administrator only their department, and an employee only their own.
  • Two threads per order. Your staff have an internal thread the client company never sees, and a thread for messages to the client.
  • Health data behind permission. An administrator can switch off a colleague's right to read health data, and every read of health data is logged.
  • National identity numbers are encrypted in the database and shown only when someone clicks Show. That is logged too.
  • Group results from surveys are never shown for groups with fewer than four answers.
  • Requests for access or erasure are received through a support ticket, and Portway extracts or erases the data.

In Portway, data (database, files and backend) is stored in the EU, with Railway in the Netherlands. Questions about sub-processors and the processing agreement are something we are glad to take in a walkthrough.

Sources

Frequently asked questions

Are we the data controller or the data processor?

Both, in different parts of the same collaboration. For the client company's orders and billing basis you are a processor for the employer. For the care documentation you are yourselves responsible as care provider. The processing agreement should cover only the first part.

Can the employer see the result of a health examination?

No. The employer gets to know that the examination has been carried out and, for medical examinations, the statement the legislation specifies: in practice whether the person is fit for the task, not the medical findings behind it.

Must data be stored in Sweden?

No. The requirement is the EU/EEA, and a transfer to a third country needs a legal basis that can be produced. What is often meant by the question is "does our data end up in the US", and that is the question you should be able to answer no to.

How long must a medical record be kept?

Record documents are kept for at least ten years after the last entry under the Patient Data Act. Orders and billing basis fall under other rules. That they have different retention times is the very reason to keep them in separate systems.

What counts as a personal data incident?

Any unauthorized access, loss or alteration of personal data, including an e-mail with health information to the wrong recipient. Notification to IMY is made without undue delay and, where feasible, within 72 hours, unless the incident is unlikely to result in a risk to the data subjects.

Do we need to carry out a data protection impact assessment (DPIA)?

For processing of sensitive personal data on a larger scale, the starting point is yes. It does not have to be extensive, but it must exist, be dated and describe the measures you have taken. That is what a procurement officer asks for, not the format.

See Portway in your everyday work.

We show Portway with what you do today, so you can see what would take care of itself at your place.