The Association of Occupational Health and Wellbeing Professionals​

Supporting you through your Occupational Health career

Over 1000 Members
Registered Charity No. 1205635

Artificial Intelligence (AI) and the digitisation of clinical decision-making: What’s in it for the OH professional?

By Lucy Kenyon

Published 12 September 2023

Document,Management,System,Concept.,Digital,Asset,Management,,Document,Imaging,,Workflow,

Technology that might move our industry forward is subject to the mathematical modelling behind policies. In this article, I am exploring the factors that influence Insurance Risk Assessment. Self-insuring organisations such as the NHS, primary care, prevention and treatment services are already using the latest wearable technology, the use of which is subject to consent to share data with AI. OH services are required to hold insurance policies and their records are ‘Special data’ with additional safeguards over and above patient records.

Clinical decision-making pathways have been featured in NICE guidelines since 2002 when the first clinical guideline was published on schizophrenia. They are designed to improve consistency of treatment across the UK and this makes the model ideal to address the public concerns around variability in OH advice. Pre-placement recommendations are probably the most automated OH process at present. Applicants and new starters answer yes or no to a series of questions, often resulting in the need for a follow-up call from a health professional to clarify whether there is an impact on work and/or adjustments required.

A McKinsey study reported the lowest automation potential for professionals managing people (9%) of work time and those applying expertise (18%). Other factors were costs to automate, availability and cost of skilled workers, quality gains), and regulatory and social-acceptance considerations. Professions applying medical expertise were rated low-risk overall.

AI and OH

We cannot discuss the digitisation of OH and the use of AI without considering policy underwriters for OH businesses. My own experience is that most underwriters still have little understanding of the scope and purpose of OH beyond listed chargeable activity titles with no requirement to clarify the decision-making processes involved. Actuaries and underwriters working with experienced OH brokers such as BHIB report being wary of services using triage and automated decision-making.

OH needs technology that facilitates efficiencies, learning and growth at a similar rate to other areas of healthcare. Would training help the financial and insurance sectors to safely remove or reduce the barriers to technological advancement whilst maintaining data protection?
The recent conviction of a health professional is a powerful example of the consequences of not understanding or learning from incidents. It also generated earnest discussions amongst fellow OH professionals regarding the consequences of and liability for ‘fitness for work’ recommendations, as we contemplate the outcome and impact of a new Clothier style enquiry into how fitness to practice might best be assessed.

AI has the potential to prevent incidents from becoming tragedies and tragedies themselves being repeated.

Custom and practice continue to see the term ‘Fitness Certificate’ issued by OH services. Most of us have experienced the frustrations of our service users when a relatively new starter experiences difficulties before the end of their probationary period despite having been issued with a Fit certificate at the time of the job offer. It is clear from the number of employers expressing at best disappointment or referring individuals within the first 6 months of employment that their expectations and interpretation of these certificates are a guarantee of fitness.

One positive legacy of COVID is how it changed managers’ perceptions of OH advice. Our reports were suddenly welcomed enthusiastically as sources of evidence for protective arrangements or reassurance that the available evidence suggested that they were unlikely to experience severe illness.  Although R Kelly v Royal Mail (UKEAT/0262/18/RN) is not a landmark judgement, Justice Choudhury was eloquently clear in his recommendations about OH advice; “That assistance and guidance may be to the effect that the employee is a disabled person: and, unless the employer has good reason to disagree with the basis of such advice, he will ordinarily respect it in his dealings with the employee. In other cases, the guidance may be that the opinion of the adviser is that the employee is not a disabled person. In such cases the employer must not forget that it is still he, the employer, who has to make the factual judgement as to whether the employee is or is not disabled: he cannot simply rubber-stamp the (OH) adviser’s opinion that he is not.”

This raises fundamental questions about the nature, reach and outcomes of OH advice in a digital age. As soon as AI apps and systems appeared, my business partner asked me whether it was safe to use one of the familiar AI systems to generate an OH report.

Pioneering data-driven technologies in OH

In 1948 the first radiology images were sent 24 miles over a telephone line. Since then technological developments have paved the way for the transmission of psychiatric consultations via closed-circuit TV and in-flight monitoring and self-reporting of heart rate, temperature and respiration via microphone in the ‘60s. 1982 saw the first digital health assessment using a wearable biometric sports watch.

In 1983 the American Industrial Hygiene Association reported that “The use of the microcomputer offers a simple, inexpensive solution to the numerous problems associated with data management in an OH program”. I first used an OH records management system when helping Peter Mattison set up health surveillance for Body Shop manufacturing in 1995. In the last 30 years, implementation on a large scale has remained challenging on technical, logistical, and financial levels. Comparative studies of digital diagnostics have published reassuring data on safety and feasibility, but the implementation challenges referred to above highlight the need for training and the knowledge of pitfalls.

The DHSC code of conduct lists the following behaviours expected from those developing, deploying and using data-driven technologies:

Respect for persons

Respect for human rights

Participation

Accounting for decision

Many clinicians remain reluctant to sign off reports on self-assessment questionnaires alone and are concerned about reporting without the clinical consultation that has historically represented their profession.

Cost-efficiency analyses and OH issues need to be addressed comprehensively. Health risk assessment is being blended into the traditional OH workflow, and the approval of AI for routine assessments is beginning to challenge human evaluation as the gold standard.

Past digital assessment implementations and future possibilities

The addition of artificial intelligence coupled with technical challenges to increase availability and efficiency in OH is already driving changes to the profession, and we must retain ownership if we are to maintain the protection of the ‘special data’.

Historically, completed questionnaires were digitized for detailed evaluation and record keeping, audit and teaching. The introduction of “tele-assessment”, pioneered in OH in the early 2000s, saw consultations carried out by telephone rather than in person, challenging the idea that visual assessment of the individual would provide greater diagnostic accuracy than competent clinical assessment.  The intention was to reduce time away from the workplace for the individual and time lost by clinicians and service users in case of late arrival and unplanned or complex documentation.

Faxed or scanned images of forms and letters within OH workflows have propelled us into digital medicine (DM). Digital images and video streams can now be shared in real-time (telemedicine), bridging the physical distance between clinicians and service users. For me, this has led to real-time case conferences at the end of consultations to ensure that the manager has immediate recommendations to consider pending the final report.

Digital health surveillance images can now be evaluated by OH physicians for early signs of occupational disease with digital record keeping developing the capability for new trend recognition and analysis that outstrips the traditional assessment of symptoms and links images to clinical data. For me, the most exciting aspect is the opportunity to record and monitor decision-making using AI and make occupational disease truly preventable and ultimately a thing of the past.

Quality

Digital OH provides enables in-house validation of representative cases and techniques. Should OH work to the Digital Imaging and Communications in Medicine (DICOM), standard to ensure that diagnostically used data storage is adequate?

In conclusion, the digitisation of routine diagnostics is complex and requires expertise, human resources, and a continuous focus on technical advances. Clinicians may also need external professionals to define specifications for major hardware procurements and compare vendors’ offers in-depth meaningfully.

A quick search for AI clinical workflow produced several applications providing analysis of reports before actual sign-out, configured to analyse the assessment report as well as images of screening to automatically alert the clinician to potential diagnostic discrepancies before final sign-out. I understand that AI that collects and analyses human input has been proposed to triage cancer pre-screening cases for reporting and to balance workload distribution. This has huge potential for mesothelioma and other occupational cancers.

AI applications have been shown to benefit from human input for increased classification performance and reduced training. This concept is called “Human-in-the-loop”.

With the majority of OH professionals operating independently, cost-effectiveness can only realistically be achieved with a digital assessment workflow that does not cost more than the human processing time and also enables human audit on request.

Digital DSE risk evaluation

We cannot talk about the use of technology within OH without considering the foreseeable risks associated with continuous DSE use. Time-efficient hardware and software must consider digital eye and upper limb strain. Between 64% and 90% of computer users experience visual symptoms such as dry eyes, headaches, eyestrain, burning eyes, diplopia, and blurred vision when using (computer) screens (M Rosenfield, 2011). Musculoskeletal injury from computer use has been estimated to account for at least half of work-related injuries (PC Bohr, 2000).

Keywords:

digital occupational healthmachine learningartificial intelligenceoccupational healthcomputer vision syndromeautomation; clinical guidelines, clinical protocols, underwriting, medical malpractice, medical negligence, public liability

 

Lucy Kenyon | Linkedin

Lucy Kenyon is a Specialist Independent OH Nurse Consultant who has been teaching and mentoring Occupational Health since 1996 for the RCN, Universities of Birmingham, Coventry, Derby and West of Scotland. She delivers professional development courses for the National Performance Advisory Group, Cordell Health and employers.  Lucy was President of iOH from 2017 until 2020. She has been a Non-Executive Director since 2021. She has a Master of Medical Science and International Certificate in Occupational Hygiene from Birmingham, where she has taught and worked on research projects.

OH Today Spring 2023
Read original article here

Share

Skip to content