Bridging technology and empathy through artificial intelligence: Enhancing subthreshold mental disorder diagnosis in preventive occupational medicine

This case report highlights the experience of a 21-year-old apprentice in a carpentry workshop who developed sub-threshold anxiety and obsessive-compulsive disorder (OCD) following a minor workplace injury. After an incident where a wood splinter entered his eye, the patient’s psychological condition worsened, manifesting heightened anxiety, somatizations, and hypochondriacal thoughts, significantly impairing his ability to work. Despite an initial face-to-face medical evaluation, subclinical symptoms went unnoticed, delaying the diagnosis. The introduction of artificial intelligence (AI) into the diagnostic process, via a customized digital system, facilitated the administration of neuropsychological assessments and provided real-time clarification of diagnostic questionnaire items. This AI-assisted approach identified previously unreported symptoms, such as intrusive thoughts, avoidance behaviors, and insomnia, leading to a timely diagnosis of anxiety disorder comorbid with OCD. Early intervention, combining pharmacological treatment (sertraline, clomipramine, and valproic acid) and cognitive-behavioral therapy, rapidly alleviated the patient’s symptoms. This case demonstrates the potential of AI in enhancing diagnostic precision for mental health disorders in occupational settings, reducing stigma, and expediting treatment, thus improving patient outcomes and supporting workplace mental health initiatives.

INTRODUCTION

There is a growing interest in exploring the association between sub-threshold anxiety disorders and other mental disorders, particularly in understanding how work-related stress contributes to their development in both susceptible and non-susceptible individuals ¹. It is well-established that workplace stress might act as a primary catalyst for the onset of psychological distress, from mild anxiety and adjustment disorders to more severe issues like depression, burnout, and post-traumatic stress disorder (PTSD) as well as that such stress may worsen pre-existing psychopathological conditions, including mood disorders. Consequently, mental health challenges in the workplace are of significant concern due to their impact on individual wellbeing and broader socio-economic factors, including increased absenteeism, reduced productivity, and job quitting ². Occupational physicians – pivotal professionals in both preventing and detecting psychiatric disorders that are often already present yet sub-threshold among workers – encounter a significant challenge in effectively and timely monitoring mental health in the workplace both for the limitation of available adequate tools and for their rather infrequent interactions with psychiatrists in managing diagnosed mental illnesses ^1,2. Historically, the investigation of work-related stress has relied on risk assessment tools like questionnaires to identify potential stressors in the workplace. Despite this, occupational physicians primarily depend on medical histories collected during preventive medicine visits and the ability to request psychiatric consultations as their main tools. However, this approach may present significant limitations, including dependence on the operator’s experience, interpretation of sub-threshold symptoms, effective communication with workers — complicated by the stigma surrounding mental illness and workers’ fear of job loss — and the formulation of potential intervention strategies. In such a complex scenario, a prompt recognition and characterization of psychiatric disorders (including anxiety spectrum disorders and PTSD) in everyday clinical practice might still remain challenging ¹.

Given the complexity of early diagnoses, the integration of artificial intelligence (AI) into diagnostic guidance might perhaps hold promising potential, as AI’s effectiveness might extend to early detection of mental disorders through meticulous monitoring of symptom progression, thereby enhancing diagnostic precision and opportunities for early intervention ³.

CASE DESCRIPTION

We present herein the case of a 21-year-old apprentice in a carpentry workshop, tasked with cutting, turning, planning, coloring, gluing, and preparing wooden components. One day, while observing the turning process of a piece of wood, he felt a splinter enter his right eye. Fortunately, the incident did not cause significant injuries: the eye was slightly reddened (likely exacerbated by the patient’s own rubbing), but remained pain-free, allowing him to continue working. In the weeks following the incident, the young man’s psychological condition significantly worsened, while manifesting heightened anxiety, constant self-monitoring of his eye, and repeated requests for reassurance from different doctors, particularly ophthalmologists. He experienced somatic symptoms including tachycardia and sensations of a foreign body in the eye, alongside nightmares and hypochondriac ideation, primarily characterized by fears of contamination. Such intense discomfort made him increasingly restless at work, particularly fearful of approaching active machinery. Previously, during his hiring process, an occupational physician deemed him fit for the job with no limitations or prescriptions. However, due to the persistence and increasing severity and impairment of his symptoms, he was referred to a psychiatrist by his physician.

During the psychiatric evaluation, a maternal family history was reported positive for anxiety disorder and obsessive-compulsive disorder. His personality profile showed good personal resources, satisfactory social adaptation, and balanced emotional life, although with vulnerabilities in judgment, impulsivity, a tendency towards precision and order.

His overall functional setup was solid, with adequate awareness of his own intrapsychic experiences and those of others. His past medical history did not indicate severe psychiatric disorders, except for brief episodes during adolescence of hypochondriacal ideation with intrusive thoughts and rituals, particularly characterized by contamination fears, accompanied by state anxiety, somatic symptoms, and catastrophic ideation that also temporarily affected his school performance. In any case he had obtained a high-school degree, These episodes occurred seasonally in spring and resolved spontaneously within a few weeks, so he had never sought psychiatric help before the incident at work. During the consultation, the patient showed significant difficulty in recounting his clinical history, overwhelmed by feelings of shame (fear of losing his job as well as of judgment by the clinician besides a significant subjective sensation of stigma). He was involved in a love relationship since one year and referred to have good friends and some hobbies during the week ends (runnig, hìtrekking).

INTEGRATION OF ARTIFICIAL INTELLIGENCE IN THE DIAGNOSTIC PROCESS

The application of AI in this clinical context involved a customized digital system that included a tablet for administering diagnostic questionnaires. This specific platform was integrated with the application programming interface (API) of a sophisticated language model designed to interact in a user-friendly manner while administering the questionnaires. The AI was trained to respond to user queries, clarifying any questionnaire item the patient might not understand. This ensured that the system was not only efficient but also engaging and supportive for the patient.

The neuropsychological tests administered through this platform included several well-established instruments, such as the Pittsburgh Sleep Quality Index (PSQI) ⁴, Zung Self-Rating Anxiety Scale (SAS) ⁵, Hamilton Depression Rating Scale (HDRS) ⁶, Trauma and Loss Spectrum Self Report (TALS) ⁷, and Yale-Brown Obsessive-Compulsive Scale (Y-BOCS) ⁸. The integration of AI in the diagnostic process significantly expedited the interpretation of questionnaire results, uncovering a range of symptoms that had gone unnoticed during the initial face-to-face assessment. These included intrusive thoughts, ideational ruminations about the consequences of the event, somatic symptoms, acute anxious reactions with avoidance of trauma-related situations (such as the use of electromechanical tools), phonophobia with alarm reactions, impaired concentration, irritability, emotional lability, feelings of shame, and central insomnia. Following AI-assisted assessment, a trained clinician ultimately administered the Structured Clinical Interview for DSM-5 (SCID-5) ⁹.

To uphold the highest standards of data privacy, the entire procedure was developed in compliance with European data protection regulations, ensuring that all patient interactions were securely managed and confidential. The case presented has been slightly modified to protect the patient’s identity.

DISCUSSION

Early diagnosis of sub-threshold anxiety disorders in occupational medicine continues to be a complex endeavor, complicated by symptoms that are often mistaken for individual personality traits or, in more pronounced cases, closely mimic those of burnout syndrome, depression, and other acute disorders such as adjustment disorder or acute stress disorder. Furthermore, the diagnostic process is significantly influenced by ‘operator-dependent’ factors, including variability in clinician experience and their specific intervention strategies ¹.

In our opinion, an early introduction of AI in the diagnostic process was pivotal in reducing feelings of shame and stigma that often accompany traditional assessments. The precise explanations provided by AI empowered the patient to fully understand and respond to each part of the questionnaire, thereby revealing subclinical symptoms of obsessive-compulsive disorder (OCD) that had previously gone unnoticed during the initial vis-à-vis interview. Such a more refined understanding led to an early diagnosis of comorbid anxiety disorder with OCD, enabling the immediate implementation of a specific treatment plan consisting of a combination of sertraline (up to 200 mg/day), clomipramine (up to 25 mg/day) and valproic acid (up to 300 mg/day) that rapidly mitigated his symptoms. At the same time, given the elevated insight and resources of the patient, cognitive-behavioral therapy was also introduced. In this specific case, the timely identification and accurate characterization of these subthreshold symptoms, which were likely unexpressed during the initial interview due to fear of clinician judgment, underscore the significant potential of AI to enhance diagnostic precision and treatment efficacy in clinical psychiatry.

Indeed, the integration of AI in psychiatry and occupational medicine holds potential for refining diagnostic and treatment methodologies. By employing therapeutic algorithms and monitoring clinical and biochemical parameters in real time, AI may enable the development of increasingly precise and targeted intervention strategies that could not only enhance the likelihood of improving overall outcomes but also accelerate the diagnostic process, providing insights and predictive models for complex conditions and identifying key indicators of various psychiatric and occupational disorders ³. Furthermore, it should be noted that the effectiveness of such AI systems hinges critically on the quality rather than on the mere quantity of data: in other words, high-quality data boosts the model’s accuracy and reliability. In such a view, well-curated data, representing diverse patient interactions, should allow AI to handle a wide array of queries more effectively, fine-tuning its responses to capture the nuances of human communication and better address the specific needs of its users. This approach ensures that AI applications in clinical settings are not only functional, but truly intuitive and beneficial. Therefore, aided by these tools, early diagnosis might also assist the occupational physician in conducting preventive assessments to identify subclinical psychological distress, thereby preserving the worker’s psychophysical integrity and directing them to a psychologist to develop effective coping strategies, and promoting workplace mental health ¹⁰.

Conflicts of interest statement

The authors declare no conflict of interest.

Funding

None.

Ethical consideration

Not applicable.

Authors’ contributions

J.M., F.M: conceptualization; F.M., J.M.: writing - original draft preparation; G.V., F.M., J.M.: development of methods (AI model); F.M., D.M., S.B.: writing - review and editing; D.M. and S.B.: supervision.