CEO navigates strict EU regulations to launch AI in healthcare organization

Contents

1 INTRODUCTION.. 3

2 AI APPROACH TO ENABLE HEALTHCARE TRANSFORMATION – SCENARIO & GOVERNANCE. 4

3 AI TOOL SELECTION.. 5

4 USE CASE – MEDICAL TEXT-TO-DIAGNOSIS CODING TO HELP DATA QUALITY. 6

5 USE CASE – CUSTOMER SUPORT CHAT BOT. 8

6 USE CASE – MEDICAL TELE HEALTH.. 9

7 LESSONS LEARNED.. 10

8 CONCLUSION.. 11

9 LINKS OF INTEREST. 12

 1 INTRODUCTION

🌟 Discovering the Power of AI. As the CEO of MAZ, a leading healthcare organization in Spain, I'm at the helm of providing essential healthcare assistance to 750,000 active workers, managing work-related injuries, and ensuring economic benefits during their sick leave. In a realm where personal and medical information is sacred, navigating the tight ship of European data protection and AI regulations is part of our daily voyage.

🚀 A Spark of Innovation. My journey with AI began in early 2023 with ChatGPT. Like many, I was captivated by its potential to revolutionize how we approach problems—unlocking doors to possibilities once deemed unimaginable. Yet, the stringent European landscape governing sensitive data demanded caution. I issued a directive within MAZ, restricting ChatGPT's use with our data, while encouraging my team to explore its capabilities personally, ensuring compliance without stifling innovation.

📚 The Mindvalley AI Mastery Epiphany. Enrolling in the "Mindvalley AI Mastery" course was a turning point. Immersed for six months, I became enthralled by AI's boundless potential, envisioning a future where AI could transform MAZ's operations. My excitement was tempered by the challenge of navigating our compliance landscape, unable to use real MAZ data in platforms like ChatGPT, Airtable, and others. This course illuminated the vast capabilities of AI and its imminent impact on our professional and personal lives.

🤔 Confronting Dual Challenges. Understanding AI's scope sparked twin challenges: accelerating our process reengineering and AI technology incorporation to boost efficiency and service quality, and ensuring our 1,200-strong workforce remains valuable in an AI-augmented future.

🚧 Navigating Compliance and Innovation. The journey has been a delicate dance of pushing for AI implementation while heeding the cautious chords played by our Data Protection and Compliance Officers. Their refrain—"No way... need to evaluate risks..."—became a familiar melody in our innovation meetings. Yet, the drive for progress persisted. "Come on!" I'd argue, only to be met with "Sorry, no way."

🔍 A Path Forward. Waiting for a secure AI tools environment was our reality, balancing the promise of AI with the imperatives of EU regulations and our IT and staff capabilities.

🌈 Embarking on Our AI Journey. Today, we stand on the precipice of change, having taken our first steps into the AI realm. This tale is not just mine but a beacon for fellow AI Mastery enthusiasts navigating similar waters.

2 AI APPROACH TO ENABLE HEALTHCARE TRANSFORMATION – SCENARIO & GOVERNANCE

🔒Navigating Dual Regulations for Medical Data Use in Europe. The deployment of medical data in Europe is governed by two critical frameworks: the European Data Protection Regulation and the specific regulation for the Use of Artificial Intelligence. This necessitates a robust approach to ensure the security and confidentiality of data and processes, introducing new prerequisites such as obtaining patient consent for AI data processing and mandating cloud services to host servers within European territory. 🇪🇺

💡 Establishing AI Governance at MAZ. To navigate these complexities, MAZ has instituted a tailored AI Governance system, selected AI tools compliant with European standards, and launched three Proofs of Concept (PoCs) to pioneer AI integration.

These efforts include two commercial solutions that adhere to European regulations, two cloud-based PoCs developed by our internal ITC team, and a third PoC leveraging commercial medical solutions for comparative analysis, aiming for eventual implementation.

🗂️Core Components of MAZ's AI Governance. Our governance framework encompasses:

- AI Use Policy: endorsed by our Board of Directors, this policy outlines the ethical and regulated use of AI within our operations.

- Workshops for Use Case Identification: these sessions foster a technology-forward mindset, enabling brainstorming and case evaluations to pinpoint viable AI applications.

- Project Management Office for Process Automation Requests: this office orchestrates project selection, prioritization, resource allocation, and the integration of AI initiatives with ongoing ICT projects, ensuring a strategic and cohesive approach.

🚫European Regulatory Landscape's Impact on AI Adoption: the stringent regulatory environment in Europe, particularly within healthcare and finance sectors, has led many companies to halt or strictly limit AI tool adoption. Some organizations have gone as far as banning generative AI tools like ChatGPT, with company firewalls blocking access to these resources.

👨‍⚖️Setting a Precedent with MAZ's Approach: MAZ's strategy not only aligns with the existing legal framework but also pioneers a secure and ethical paradigm for AI utilization in healthcare. The Board of Directors' ratification of our AI Use Policy signifies our commitment to adhering to all European AI regulations, championing both ethical and transparent AI practices. 

 3 AI TOOL SELECTION

🔄 Diverse Process Integration. As providers of occupational health services, we navigate through two distinct types of processes: core and support. Core processes, such as healthcare assistance for workers injured on the job and managing sick leave compensations, leverage complex ERP solutions. Support processes, including purchasing, contracting, and financial management, utilize a variety of business management applications.

🛠️ Strategic AI Tool Selection. Within our Microsoft Office 365 ecosystem, the selection of initial AI tools was straightforward: Microsoft Copilot for administrative tasks and Microsoft Power Automate for enhancing workflow automation. Our team, comprising 65% healthcare and 35% administrative staff, benefits from diverse expertise in Office 365 and ERP applications. The phased deployment—starting with Copilot in February 2024, followed by Power Automate—ensures a structured integration.

💼 Key Selection Criteria. The selection criteria were meticulously chosen to align with our technical environment and EU regulatory compliance, particularly concerning data privacy. For healthcare-specific solutions, factors such as immediate availability, ease of implementation, and compliance with European regulations were paramount.

🎓 Comprehensive Training Strategy. Our 2024 roadmap includes crucial steps like awareness, training, use case identification, and risk evaluation, culminating in the initial AI implementations. A gradual upskilling approach will ensure that by the second quarter of 2024, up to 10% of our 1,200 staff will proficiently use the selected AI tools. By 2025, we aim for all employees to be adept in AI, enhancing their contribution to the organization without reducing headcount. Instead, we'll focus on boosting process efficiency and developing new services to better serve our customers.

🔮 Anticipating Future Innovations. Accurate predictions of occupational pathologies will enable us to implement targeted labor risk prevention measures, elevating our clients' competitiveness. Within the realm of work-related healthcare, we anticipate introducing innovative features to address the rising incidence of sick leaves in Europe, showcasing our commitment to advancing healthcare outcomes and organizational efficiency through AI.

 4 USE CASE – MEDICAL TEXT-TO-DIAGNOSIS CODING TO HELP DATA QUALITY

🏥 Healthcare Service Overview. From the 1,000 hospitals and clinics across Spain providing our healthcare services, 84 are staffed by our own professionals, including a significant facility with 500 specialists in Zaragoza.

📊 2023 Service Snapshot. Our dedicated team delivered care to 340,000 patients through 678,000 medical procedures, including 23,000 physiotherapy sessions, 180,000 consultations, 120,000 diagnostic imaging tests, 57,000 emergency room visits, 3,000 surgeries, and 900 intensive care unit stays, among others.

🔍 Critical Need for Accurate Medical Coding. Ensuring precise coding of medical procedures is vital for delivering appropriate emergency services, diagnosis, treatment, and rehabilitation. With a 15% annual turnover among our 276 doctors, replacing departing staff with equally experienced professionals poses a significant challenge. This turnover underscores the necessity of our Medical-Text-To-Diagnosis Coding solution to minimize the risk of incorrect coding by accurately processing data from unstructured texts such as doctor's notes and discharge summaries.

🚀 Innovative AI Solution. Our solution aims to automatically extract and label crucial medical information from unstructured data, enhancing decision support across various clinical scenarios. This innovative Proof of Concept, developed by the MAZ ITC team using Azure APIs, is a testament to our internal capabilities to integrate advanced technologies into our healthcare processes. While designed to be multilingual, it primarily serves our Spanish-speaking patients and professionals.

💡 Proof of Concept and Skills Development. Before entering full operation, this Proof of Concept will undergo thorough testing. The successful integration of Azure APIs into our multi-application environment marks a significant achievement, showcasing our ITC department's proficiency in leveraging cutting-edge AI tools for healthcare excellence.

5 USE CASE – CUSTOMER SUPORT CHAT BOT

🤖 Introducing Our Customer Support ChatBot. This initiative represents another innovative Proof of Concept within our organization, set to begin testing in March 2024. Our ambition is to transition this project into full operational use by June 2024.

💬 Enhancing User Experience. Through this ChatBot, user inquiries will receive instant responses from a cloud-based system, meticulously trained with data encompassing all online-processable procedures. This advancement allows users to bypass traditional online help systems, posing their questions in natural language for a more intuitive interaction.

🔧 Developed by Our In-House Experts. The MAZ ITC development team has crafted this solution utilizing Azure APIs, demonstrating our commitment to leveraging advanced technology to improve customer support.

📈 Skill Advancement and Integration. The development and integration of the Azure APIs into our multi-application customer support framework underscore the growing expertise of our internal ITC department. This project not only enhances our customer service capabilities but also enriches our team's skill set in integrating complex cloud-based solutions.

6 USE CASE – MEDICAL TELE HEALTH

🏥 Expanding Radiology Services Across Our Network. Among the 1,000 clinics we operate in Spain, 84 are directly staffed by our team, including a notable 150-bed hospital in Zaragoza equipped with a dedicated Radio-Image Service and a team of five Radiologists.

📡 Bridging the Gap with Tele-Health. Given the absence of on-site Radiologists at the 84 clinics, our hospital frequently fields radiology inquiries from across our network. Musculoskeletal disorders, the most common conditions diagnosed at our clinics, accounted for 68,000 X-ray examinations last year.

🔍 Innovative Proof of Concept for Bone RX Diagnosis. To streamline and enhance diagnostic processes, particularly for doctors outside the hospital, we're launching a proof of concept to evaluate two market-leading commercial solutions. These AI-driven tools are designed to analyze X-ray images, providing comprehensive insights beyond the symptoms highlighted by patients, thus facilitating a more accurate and expedited diagnosis.

📊 Comparative Analysis for Optimal Selection. The effectiveness of these solutions will be measured against three key criteria: diagnostic accuracy (validated by our musculoskeletal specialists), speed of analysis and data transmission, and compliance with legal standards for handling sensitive information. Following a two-month evaluation period, the best-performing solution will be adopted for widespread use.

⚖️ Navigating Legal Requirements. The challenge here extends beyond technical integration to encompass legal compliance with European regulations. Our organization is committed to developing an AI Usage Policy, conducting risk assessments, evaluating the impact of these diagnostic tools, and securing patient consent for AI data processing.

7 LESSONS LEARNED

🚀 Embracing the AI Revolution: Feeling the urgent need to accelerate and leverage the myriad of AI tools available in the market, we aimed to enhance our processes and expand the quality and scope of services for our customers. Yet, we encountered the formidable challenge of navigating through the strict European regulatory landscape.

⏳ Navigating Compliance and Compatibility: Our journey required patience as we sought tools compatible with MAZ's IT environment. This diligence ensured our alignment with regulations but introduced a delay in AI solution implementation.

🌍 Uniting a Diverse Workforce: In our quest to adopt AI, we recognized the importance of enlightening our extensive team, spread across numerous centers nationwide, about AI's rapidly growing role in our personal and professional lives.

📚 Governance and Organized Implementation: Emphasizing an organized governance of AI, we focused on coherent, compliant, and efficient deployment of selected solutions. The project management office plays a pivotal role in evaluating and selecting use cases, guided by management-defined criteria.

🎓 Strategic Training Approach: A phased training plan is crucial. Initial learning by selected advanced users will lay the groundwork for broader staff training, ensuring proficiency in creating well-prioritized, impactful use cases.

🔮 Anticipating AI's Impact: Despite not yet deploying any AI solution in production, we're aware of the unknowns regarding AI's real impact on our organization. Our goal is to safeguard our employees' value within the organization over the medium term, aspiring to create new, significantly value-added services for our customers—a considerable challenge amid the backdrop of downsizing trends observed in firms that have integrated AI solutions.

📈 Continuous Learning and Adaptation: Our journey is one of perpetual learning, as we remain committed to navigating the complexities of AI implementation while upholding our dedication to our employees and customers.

 8 CONCLUSION

🔄 Embracing Individual Contexts: As always, context is king. We encountered delays, but ultimately, we've identified AI applications that look suitable for MAZ. Now, we're in a thrilling race of innovation, speeding ahead while skillfully navigating the twists and turns to stay on track.

🔭 Anticipating Transformation: Looking to the future, I envision significant transformations across the board in our organization and in healthcare at large, both administratively and clinically. We're on the brink of introducing new processes, advanced medical techniques, and a predictive prowess for occupational diseases that will redefine our clients' organizational structures.

🚀 Evolving with AI: MAZ is committed to evolving in tandem with AI advancements, ensuring that we not only keep pace but also capitalize on the opportunities that this technological progress presents.

🎯 Valuing Our Team in the AI Era: The paramount challenge lies in preserving the invaluable contributions of our employees amidst the expansive potential of AI in the coming years. Our focus is on harnessing AI to enhance, not replace, the unique strengths of our team.

🙏 A Heartfelt Thanks to Mindvalley AI Mastery: A huge thank you to the "Mindvalley AI Mastery" program. This journey has been enlightening, enriching, and instrumental in honing my professional acumen. It's been a period of significant growth, knowledge, and transformation.

9 LINKS OF INTEREST

📜Text Analytics for health (TAfH)

TAfH - General

What is the Text Analytics for health in Azure AI Language? - Azure AI services | Microsoft Learn

Azure cognitive services  - TAfH

Azure Cognitive Services basics - Training | Microsoft Learn

TAfH accepted categories

Categorías de entidad reconocidas por Text Analytics for Health - Azure AI services | Microsoft Learn

Service limits

Data limits for Language service features - Azure AI services | Microsoft Learn

Custom Text Analytics for Health

Using Azure resources in custom Text Analytics for health - Azure AI services | Microsoft Learn

Training of a personalised model for TAfH

Using Azure resources in custom Text Analytics for health - Azure AI services | Microsoft Learn

📜European Union AI regulations

EU AI Act: first regulation on artificial intelligence | Topics | European Parliament (europa.eu)