Mine workers are examined and tested throughout their careers, yet the data from that surveillance rarely holds together. Here is where it breaks down, and what it takes to fix it.
Few workforces are monitored as closely for their health as mine workers. Before someone starts work underground or in a pit, they are examined, and throughout their career they are tested again: lung function, chest X-rays, hearing and, in some operations, blood or urine monitoring for metals. The purpose is important. Health surveillance should pick up early signs of occupational disease while something can still be done, and show employers where exposure controls need to improve.
The pattern that follows will be familiar to anyone who has worked with clinical data. The surveillance itself happens, but the data it produces is hard to bring together and use. Results arrive from many providers as PDFs, scans and paper forms. Workers move between sites, contractors and employers. And the questions that matter most, such as whether a worker's lung function has been declining over ten years, can only be answered if the record holds together over time.
This guide sets out nine gaps we see in health surveillance data and practical ways to close each one. It is written for occupational health providers and software vendors, mine medical services, health and safety leaders in mining and other high-risk industries, and the teams running occupational disease and compensation programmes.
A single worker's surveillance file can draw on an in-house mine clinic, a contracted occupational health provider, an imaging practice for chest X-rays, a spirometry (lung function) provider, an audiologist and a pathology laboratory. In Queensland, for example, the doctors, X-ray imaging practices and spirometry practices serving coal mine workers must each be approved by the regulator, Resources Safety & Health Queensland (RSHQ).3 That separation exists for good reasons, because it protects the quality of each test.
The difficulty is what happens next. Each provider reports in its own format, with its own test names, units and layout, and much of it still arrives as PDFs, scanned forms or paper. Someone then has to read those reports and type the values into the system that holds the worker's record, or the reports are filed as attachments and the values inside them are never used at all.
Figure 1. One worker, many sources. Each provider reports in its own format, and someone has to turn those reports into usable data.
How to close it: Agree a minimum structured dataset for each surveillance test (for example, spirometry values, hearing thresholds at each frequency and the chest X-ray classification) and ask every provider to supply it. Where providers cannot, convert incoming reports into that structure as they arrive, with clinical review for anything uncertain, so every result lands in the record as data rather than as an attachment.
Mining depends heavily on contractors and on people who move between operations, employers and countries over their working lives. The rules recognise that movement, but they do not make the record travel with the worker. In Queensland, a new health assessment is required when a coal mine worker changes employer.4 In South Africa, mine employers must keep a prescribed record of employees doing work that is under medical surveillance, and must send the relevant part of it to the Medical Inspector when that employee leaves the mine.1
In practice, the new employer or provider often starts from a fresh baseline because earlier results are unavailable, unreadable or in a format their system cannot take in. That means tests are repeated, and a gradual decline that would be obvious across three employers is invisible within any one of them.
Figure 2. A gradual decline that spans three employers can be invisible inside any one of them.
How to close it: Build the record around the worker rather than the employer or provider, with a stable identifier and results held as structured data. With the worker's consent, prior baselines can then move with them, and a new provider can read earlier results in a consistent format instead of starting again.
Many occupational diseases take years to appear. Lung diseases caused by dust, such as coal workers' pneumoconiosis and silicosis, develop after long periods of exposure, and noise-induced hearing loss builds up gradually over a working life. The law reflects this. In South Africa, an employee's record of medical surveillance must be stored safely and kept for 40 years from the last date of surveillance.2 Queensland's major reforms to its coal mine worker health scheme followed the re-identification of coal workers' pneumoconiosis in 2015,7 and the scheme now also allows retired and former workers to access health assessments on a voluntary basis.3
All of this depends on old records being found, read and compared with new ones. Paper files from fifteen years ago, scanned reports with handwritten values, and results stored in systems that have since been replaced are exactly the records that are hardest to use when a worker's health, or a compensation claim, depends on them.
How to close it: Treat historical surveillance records as data to migrate, not just documents to archive. Convert the values that matter (lung function, hearing thresholds, X-ray findings, exposure-related results) into a structured, consistent form while keeping the original document linked for audit, so trends across a whole career can be seen and checked.
Surveillance tests produce some of the hardest documents in healthcare to work with. A spirometry printout combines a table of values, such as forced expiratory volume in one second (FEV1) and forced vital capacity (FVC), with flow curves and the reference values used. An audiogram is usually a chart, with the hearing threshold at each frequency plotted as a symbol rather than written as a number. Chest X-ray reports for dust-related disease are often read against the International Labour Organization (ILO) classification of radiographs of pneumoconioses,5 and the result may sit in free text or on a scanned form. On site, readings from portable devices are often photographed or written down by hand.
The information is all there, but it is held as pictures of data rather than data. Until someone extracts it, a system cannot compare this year's result with last year's, flag a worker for follow-up or count how many workers have shown a decline.
Figure 3. The values exist, but as pictures. Illustrative values only.
How to close it: Extract the values that drive decisions from every report, whatever its format, and map them to a defined field in the worker's record. Use automation for scale, but route anything uncertain, such as a faint audiogram symbol or a handwritten value, to a clinically trained reviewer, and keep the source document attached so every number can be traced back.
Even when results are already digital, they rarely line up. One provider writes "FEV1", another "FEV 1.0", and a third reports only a percentage of the predicted value. Spirometry results depend on which reference equations were used, audiometry may be tested at different sets of frequencies, and some laboratory tests used in biological monitoring, such as blood lead, are reported in different units in different countries. Regulators also set their own forms: Queensland coal mine worker assessments, for instance, must be completed on an approved form through the RSHQ ResHealth system.4 A mining group operating in several countries, or a provider serving several clients, ends up with records that look similar but cannot be safely compared.
How to close it: Map every incoming result to one consistent data model, resolving variant test names, converting units and recording which method or reference values were used. Align that model to medical coding standards where they exist, while still catering to the organisation's own code sets and regulator forms, so results can be compared across providers, sites and countries.
When data is typed in by hand from many sources, gaps and errors creep in. A study of occupational health records at a South African platinum mine found that thousands of audiometry screening and occupational hygiene records (4,675 in total) were incomplete or inaccurate, and noted that the use of different databases across mines leads to missing records and inconsistent recording of noise measurements.6 The authors describe this not only as a data problem but as an ethical one, because poor quality surveillance data can lead to unequal care for individual workers.
This is the quiet cost of fragmented data. The testing still happens and is still paid for, but the surveillance it is meant to support becomes less reliable.
How to close it: Check data as it enters the record, not years later. Validate that required fields are present and that values are plausible, flag missing results, and send uncertain items to clinical review. Make sure reviewers' corrections feed back into how future reports are handled, so quality improves over time rather than being fixed one record at a time.
Surveillance is most useful when medical results can be read alongside exposure: which site, which role, which dust, noise or chemical levels, and for how long. In many organisations, occupational hygiene measurements sit in one system, medical surveillance in another, and treatment records somewhere else again. In the same South African study, researchers drew on separate audiometry, hearing screening and occupational hygiene databases, and the database holding medical disease and treatment records could not be accessed at all.6
Without that link, it is hard to see which controls are working, which groups of workers are most at risk, or whether a cluster of results reflects a real exposure problem.
How to close it: Link medical and exposure data through shared, consistent identifiers for the worker, role, site and time period. For workforce-level analysis, use pseudonymised data so patterns can be studied without exposing individual medical detail.
Surveillance data does not only serve the worker and employer. It has to be reported. In South Africa, employers must keep surveillance records until the mine closes and then deliver them to the Medical Inspector,8 and must provide the relevant part of a worker's record when that worker leaves.1 In Queensland, completed coal mine worker health assessment records are uploaded to a dedicated health assessment portal.4 Annual reports, audits, inspections and compensation claims add further requests.
When the record is a collection of PDFs and paper files, every one of these requests becomes a manual project. When it is structured data, most of them become a query.
How to close it: Structure the record so that the information regulators, auditors and compensation bodies ask for can be produced directly, with the source documents attached. Plan ahead for events such as a worker leaving, a contract ending or a mine closing, so records can be handed over complete and in a usable form.
Occupational health sits in a delicate position. Employers need to know whether a worker is fit for a role and what restrictions apply. The clinical detail behind that decision belongs to the worker. South African law is explicit: a worker's record of medical surveillance must be kept confidential and may be released only in line with medical ethics, when required by law or a court order, or with the worker's written consent.2 In Queensland, the health assessment report sets out fitness for duty and any work restrictions,4 which is the outcome an employer needs.
When fitness outcomes and clinical findings sit together in the same scanned document, sharing one without the other is difficult, and people end up either over-sharing or locking everything away.
How to close it: Separate the fitness outcome from the clinical detail in the data itself, and control access by role. Record consent clearly, confirm where data may be processed under local law, and choose partners who can process data within the country where that is required.
Put together, closing these gaps means turning surveillance from a collection of documents into a worker-centred record that holds together across providers, employers and years. That record can then do far more than satisfy a compliance requirement.
Figure 4. From scattered documents to a record that supports decisions, reporting and prevention.
Health surveillance exists to protect people who do some of the most demanding work there is. Making the data behind it usable is how that protection holds up over a whole career. There are two ways we can support you.
Health surveillance is the planned, repeated testing of workers exposed to health hazards such as dust, noise and chemicals. In mining it typically includes an entry medical, then periodic lung function tests, chest X-rays, hearing tests and, where relevant, biological monitoring, so that early signs of occupational disease can be found and exposure controls improved.
It depends on the jurisdiction. In South Africa, the Mine Health and Safety Act requires an employee's record of medical surveillance to be stored safely and not destroyed for 40 years from the last date of medical surveillance. Other countries set their own periods, so organisations should check the rules that apply to them.
Some schemes require a new assessment with each employer, as in the Queensland Coal Mine Workers' Health Scheme. In other cases tests are repeated because earlier results are unavailable, unreadable or in a format the new provider's system cannot take in.
Yes. Values in PDFs, scans and photographs can be extracted and mapped into a consistent structure, as long as the process includes clinical review for uncertain results and keeps the original document linked for audit.
By separating the fitness-for-work outcome, which employers need, from the clinical detail, which stays confidential, and by controlling access by role. For analysis across a workforce, data can be pseudonymised. Organisations should also confirm where data may be processed under local law.
Yes. Any workforce under health surveillance, including construction, oil and gas, manufacturing, chemicals and agriculture, faces the same challenges of many providers, mobile workers, long time horizons and confidential records.
This article is for general information and is not legal or medical advice. Organisations should confirm the requirements that apply in their own jurisdictions.
Suhina Singh is a physician, the Founder and CEO of Jonda Health, and a health data strategist working at the intersection of artificial intelligence and clinical information systems.
