{"id":180503,"date":"2025-10-29T15:06:10","date_gmt":"2025-10-29T15:06:10","guid":{"rendered":"https:\/\/teknomers.com\/en\/we-spoke-with-spanish-hospitals-that-are-already-using-ai\/"},"modified":"2025-10-29T15:06:12","modified_gmt":"2025-10-29T15:06:12","slug":"we-spoke-with-spanish-hospitals-that-are-already-using-ai","status":"publish","type":"post","link":"https:\/\/teknomers.com\/en\/we-spoke-with-spanish-hospitals-that-are-already-using-ai\/","title":{"rendered":"We spoke with Spanish hospitals that are already using AI."},"content":{"rendered":"\n
\n

“With artificial intelligence in consultation, we can look our patients in the eye again.” This is the opinion of Manel Ramos<\/strong>, an internal medicine doctor at the Hospital Cl\u00ednic in Barcelona and Director of the AI Master’s program in Research and Teaching at the University of Barcelona.<\/p>\n

<\/p>\n

This idea may seem paradoxical: automating processes to focus on the human element, but this technology aims to correct a problem created by an earlier one. With digital records in consultations\u2014positively impacting many aspects\u2014”we went from being listeners to writers; the medical professional only looks at the screen while filling in predefined fields,” explains Ramos. If AI handles the bureaucratic part, doctors can concentrate on attending to patients.<\/p>\n

<\/p>\n

Ramos refers to programs known as medical scribes<\/em><\/strong> that are currently being piloted at Cl\u00ednic and at other public hospitals in Spain, such as Hospital La Paz in Madrid, and private centers like those in the Quir\u00f3n group. The objective is to turn the doctor from a mere data entry clerk into a patient-focused professional: AI captures the information while the provider speaks, looks at, and attends to the patient, without interruptions every few seconds to write.<\/p>\n

<\/p>\n

\n
\n
\n
<\/div>\n<\/p><\/div>\n<\/p><\/div>\n<\/div>\n

“The program sits in the consultation with the patient and serves as an assistant for the professional: it records notes from the conversation and then makes a brief summary that the professional validates,” explains Araceli Camacho<\/strong>, Director of Digital Transformation at Hospital Cl\u00ednic and coordinator of the pilot program. The goal, she states, is for “the professional to have more quality time with the patient.” And Ramos reiterates: “There is always human review; validation by a doctor is necessary.”<\/p>\n

<\/p>\n

A Long History of AI in Medicine<\/strong><\/h2>\n

The Scribe<\/em> programs belong to the realm of generative artificial intelligence, designed for communication in natural language, with ChatGPT being the most popular example. Since its explosion in 2022, this technology has been integrated into multiple everyday tasks. Medical centers are no exception.<\/p>\n

Besides taking notes during consultations, Ramos notes that generative AI can also assist with administrative tasks: “For example, if a patient brings in 50 pages of studies, AI can summarize them in 5 minutes. The goal is to reduce bureaucratic burdens and allow professionals to focus on patient care.”<\/p>\n

<\/p>\n

Although the mass expansion of generative AI is recent, its application in medical consultations is not new. According to Camacho, “it can be applied in many fields within a hospital. It helps improve care quality, seeks efficiency in processes, and aids in research.”<\/p>\n

<\/p>\n

\n
\n
\n
\n \"Stroke\n <\/div>\n<\/p><\/div>\n<\/p><\/div>\n<\/div>\n

AI work began in the 1950s and has been applied to medicine since the 1970s<\/strong> with early models for the early diagnosis of infectious diseases. In the 21st century, increased processing capacity has enabled the training of algorithms for prevention, diagnosis, personalized treatments, and hospital management.<\/p>\n

<\/p>\n

At La Paz University Hospital, for example, AI is used “across 24 clinical services through about 40 projects that cover practically the entire spectrum of AI: from robotics to generative and predictive AI based on texts, structured data, signals, or medical images,” says Fernando Mart\u00edn S\u00e1nchez, Deputy Medical Director for Medical Informatics, Digital Strategy, and Innovation at the Madrid center.<\/p>\n

<\/p>\n

\n
\n
\n \"Clinic\"\/
\n \"Clinic\"
\n The Hospital Cl\u00ednic de Barcelona.<\/span>\n <\/div>\n<\/p><\/div>\n<\/div>\n

Almost all disciplines benefit, from cardiology to neurology, including pathological anatomy. However, where AI is advancing the fastest is in the field of imaging: radiology, CT scans, MRIs, and endoscopies. “Radiology is where AI is being integrated the quickest,” notes Ramos. “It helps confirm normal results, prioritizes urgent cases, and reduces the workload for radiologists.”<\/p>\n

<\/p>\n

Significant progress is also evident in dermatology (where it can screen images of lesions) and oncology (with substantial development in personalized treatment selection). However, “areas closer to direct human contact, like psychology or psychiatry, are advancing more slowly, fortunately,” adds Ramos.<\/p>\n

<\/p>\n

Spanish companies such as Quibim<\/a> have published studies showing significant increases in diagnostic sensitivity for specific tumors. There are also systems in ophthalmology to detect retinopathies or glaucoma.<\/p>\n

<\/p>\n

\n
\n
\n
\n \"AI\n <\/div>\n<\/p><\/div>\n<\/p><\/div>\n<\/div>\n

In these cases, we are dealing with predictive or discriminative artificial intelligence<\/strong>, based on neural networks and deep learning. “Such systems are trained with vast amounts of data. For example, they are provided with many images of patients with pneumonia and others of healthy patients. The model learns to recognize patterns, and when presented with a new image, it can determine whether it is pneumonia,” explains Mart\u00edn S\u00e1nchez.<\/p>\n

<\/p>\n

Furthermore, robotic tools for surgical interventions have been successfully implemented in around 80 Spanish hospitals, where the Da Vinci surgical robot<\/a> has performed thousands of procedures in oncology and even for organ transplants in recent years.<\/p>\n

<\/p>\n

Slow Implementation in Spain<\/strong><\/h2>\n

The pace of implementing these tools remains “very low,” in the eyes of Manel Ramos. “Each area integrates these technologies at different rates: those closest to imaging are faster, while those involving personal interaction progress more slowly.”<\/p>\n

<\/p>\n

The crux of this slowness<\/em> lies in the need to comply with strict protocols, some of which are still under development. “We need to ensure patient data protection. Voice records are delicate; patients must provide informed consent and be aware that recordings are being made,” warns Ramos.<\/p>\n

<\/p>\n

\n
\n
\n
\n \"hospital\"\/
\n \"hospital\"
\n <\/a>
\n An imaging diagnostic machine at La Paz Hospital in Madrid. (Comunidad de Madrid)<\/span>\n <\/div>\n<\/p><\/div>\n<\/div>\n

In addition to the Data Protection Law<\/a>, at the European level, a crucial regulation for the entry of such technologies into various fields, including medicine, will be established with the European AI Act<\/a>, approved in June 2024. This regulation introduces a common legal framework<\/strong> to evaluate the risk level of systems before use. High-risk systems (like those in healthcare) must meet requirements for transparency, human oversight, and data quality.<\/p>\n

<\/p>\n

The regulation also prohibits unacceptable practices, such as cognitive manipulation or real-time biometric identification in public spaces, and imposes transparency on generative tools, which must indicate when content was created artificially.<\/p>\n

<\/p>\n

Meanwhile, the Ministry of Health<\/strong> expects to approve the “AI Strategy” in the National Health System (SNS) soon. At the autonomous level, communities are also investing in plans that incorporate AI to streamline screenings and reduce diagnosis times for diseases like cancer.<\/p>\n

<\/p>\n

“All these initiatives arise from the bottom up; the services themselves identify a need and, either through their means or through collaborations, have developed and tested these pilots,” explains Mart\u00edn S\u00e1nchez. Therefore, the implementation of AI in healthcare does not stem from a coordinated strategy at the national or regional levels, nor are there official lists. However, hospitals “must adapt to the existing legislation, such as the EU AI Regulations or the Data Protection Law.” <\/p>\n

<\/p>\n

\n
\n
\n
\n \"Quibim\n <\/div>\n<\/p><\/div>\n<\/p><\/div>\n<\/div>\n

And once AI is finally integrated into a hospital, the process does not end: it must be evaluated whether this technology truly delivers the promised advantages<\/strong> or if it becomes a waste of time and money, harming the patient in the process. “Developing a system may take a year, but clinically evaluating it takes three to four years. Additionally, it’s unclear who should oversee this: the hospital, the provider, or the administration; European law does not address clinical effectiveness, only requiring demonstration of accuracy and reliability,” explains Mart\u00edn S\u00e1nchez.<\/p>\n

<\/p>\n

For Ramos, the key lies in scientific methodology: “This is essentially a scientific study, with hypotheses, objectives, methodology, and data protection. Everything must pass through an ethics committee to verify compliance with laws and ensure patient safety. This significantly slows down implementation.”<\/p>\n

<\/p>\n

Is AI Reliable in Medical Contexts?<\/strong><\/h2>\n

S\u00e1nchez points out that “predictive AIs are easier to evaluate than generative ones. You can test them with a new set of images and measure how many times they are accurate. If they are accurate 99% of the time, you can reasonably trust them. This is not the case with generative AI, which produces new content. Its evaluation is far more complex.”<\/p>\n

<\/p>\n

“Proving that technology works is one thing, and demonstrating that it improves health outcomes is another. There is still no universal methodology for evaluating clinical effectiveness. Neither doctors nor patients will trust a system that does not demonstrate real benefits,” he adds.<\/p>\n

<\/p>\n

\n
\n
\n
\n \"Healthcare\n <\/div>\n<\/p><\/div>\n<\/p><\/div>\n<\/div>\n

AI continues to face risks. “Generative intelligence can exhibit what we know as hallucinations\u2014that is, it can invent information,” warns Mart\u00edn S\u00e1nchez, referring to false or fabricated responses from models like ChatGPT, stemming from insufficient data or biases in training.<\/p>\n

<\/p>\n

Ramos believes these risks are diminishing with more robust models: “Errors often stem from not asking the right question or failing to validate the answer. Sometimes, we expect a 0% error rate from machines<\/strong> when humans are not perfect either.” He adds: “As an internist specializing in autoimmune diseases, I can ask for an opinion on lupus or Crohn\u2019s disease to then validate the answer, but I cannot ask about a complex cancer case because I won’t know if it is wrong.”<\/p>\n

<\/p>\n

\n
\n
\n \"AI
\n \"AI
\n <\/a>\n <\/div>\n<\/div>\n

Additionally, the algorithms governing these AIs generally reflect human biases, which can be dangerous when making decisions. In June, Civio<\/a> revealed that Quantus Skin, a skin cancer detection system slated for implementation in the Basque Country, missed one in three melanomas, showing a sensitivity of 69% and neglecting patients with darker skin because it was trained only on white patients. Specialists labeled its results as “poor” and “dangerous.”<\/p>\n

<\/p>\n

Another study by the same outlet warns about an anomaly already occurring in Spanish hospitals: AI is being integrated<\/a> into the Spanish public health system through systems like Eyelib<\/strong>, a robot performing ophthalmological diagnoses, which is presented as a solution to waiting lists but is utilized without prior notification to patients about their data being used in an AI system.<\/p>\n

<\/p>\n

Towards the End of “Data Fillers”<\/strong><\/h2>\n

To ensure healthcare can leverage AI without becoming a burden, training staff is essential. “We have challenges related to integration and adaptation of teams. Staff must be educated. AI represents a change not just in healthcare but in all fields,” says Araceli Camacho. The La Paz expert agrees: “We are making significant efforts to train clinicians and researchers. We have several courses annually, but we also need permanent experts<\/strong> in AI in healthcare, which is quite challenging to find.”<\/p>\n

<\/p>\n

\n
\n
\n
\n \"Doctors\n <\/div>\n<\/p><\/div>\n<\/p><\/div>\n<\/div>\n

Moreover, the environmental consequences of the excessive use of AI concern experts. At Complutense University of Madrid, there are warnings about the environmental impact: data centers required to train and run AI models consume vast amounts of water and energy. It’s estimated that for every twenty responses generated by tools like ChatGPT, one liter of water is used. To mitigate these effects, they propose improving water efficiency, situating data centers appropriately, and promoting environmental labeling.<\/p>\n

<\/p>\n

\n
\n
\n \"Medical
\n \"Medical
\n <\/a>\n <\/div>\n<\/div>\n

According to Mart\u00edn S\u00e1nchez, “training models requires electricity, water for cooling data centers, and produces waste. Our hospital has a dedicated unit to study these issues and sensitize staff to avoid excessive use. It’s like leaving the lights on all night: we must use technology rationally.”<\/p>\n

<\/p>\n

“This pattern of being merely data fillers<\/strong> has to change. By next year, 2026, we will likely begin seeing significant shifts in many hospitals,” notes Ramos, for whom it is crucial to respect the concept of human in the loop<\/em>, meaning humans must always be involved in generative processes, both at the start and the end, with validation.<\/p>\n

<\/p>\n

“The impact of this technology is even greater than that of electricity or the Internet. Its influence will be absolute and cross-sectional across all areas of knowledge and everyday life,” concludes the internist.<\/p>\n

<\/p>\n

In Xataka | We didn’t know why some superbugs were resistant to antibiotics. This AI figured it out in two days.<\/p>\n

In Xataka | Chinese company Alibaba has an AI to detect pancreatic cancer. It’s so effective that the US has expedited its approval.<\/p>\n

Image Credit | Comunidad de Madrid<\/a><\/p>\n<\/div>\n


\n
General News – 2<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"

“With artificial intelligence in consultation, we can look our patients in the eye again.” This is the opinion of Manel Ramos, an internal medicine doctor at the Hospital Cl\u00ednic in Barcelona and Director of the AI Master’s program in Research and Teaching at the University of Barcelona. This idea may seem paradoxical: automating processes to […]<\/p>\n","protected":false},"author":1,"featured_media":180504,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[36399],"tags":[330,11993,408],"class_list":["post-180503","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-technology","tag-hospitals","tag-spanish","tag-spoke"],"_links":{"self":[{"href":"https:\/\/teknomers.com\/en\/wp-json\/wp\/v2\/posts\/180503","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/teknomers.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/teknomers.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/teknomers.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/teknomers.com\/en\/wp-json\/wp\/v2\/comments?post=180503"}],"version-history":[{"count":0,"href":"https:\/\/teknomers.com\/en\/wp-json\/wp\/v2\/posts\/180503\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/teknomers.com\/en\/wp-json\/wp\/v2\/media\/180504"}],"wp:attachment":[{"href":"https:\/\/teknomers.com\/en\/wp-json\/wp\/v2\/media?parent=180503"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/teknomers.com\/en\/wp-json\/wp\/v2\/categories?post=180503"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/teknomers.com\/en\/wp-json\/wp\/v2\/tags?post=180503"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}