Misdiagnosis because AI cannot examine patients
Patient N.Y. K (35 years old, residing in Ho Chi Minh City) was hospitalized after many weeks of continuous dry cough, cough with phlegm but did not go for examination. Initially, the patient thought that this was just a symptom of a common cold or respiratory infection, so he searched for information on AI platforms for reference. Based on the received suggestions, the patient bought medicine and treated it at home.
However, after a period of using the medicine, the symptoms did not subside. The prolonged cough accompanied by fatigue and rapid weight loss made the patient worried and decided to go to the hospital for examination.
At Military Hospital 175, doctors determined that the patient had pulmonary tuberculosis. Because the disease had been going on for a long time, the patient was indicated for hospitalization for inpatient treatment before continuing the tuberculosis treatment regimen according to professional instructions.
Dr. Nguyen Trung Nguyen - Director of the Poison Control Center, Bach Mai Hospital - said that the unit once received a 63-year-old patient in Ninh Binh who was in critical condition after following sinusitis treatment instructions from an "AI doctor".
The patient was instructed to burn a mixture of coconut skull, đinh lăng and buffalo horn in an avocado tube and then cover himself with a blanket to inhale smoke. After a few hours, the patient fell into a coma, suffered burns to the face and was diagnosed with severe CO poisoning, heart damage, rhabdomyolysis, and risk of brain damage, requiring endotracheal intubation for treatment.
Another case is Mr. N.V.H (34 years old, Hanoi) suffering from chest tightness after exercise. After entering the symptoms into the AI application, this tool leaned towards diagnosing muscle pain or costochondritis and recommended rest. Believing that the disease was not serious, he bought pain relievers himself. The next day, the pain increased, spreading to his left shoulder, accompanied by sweating. Upon admission, the doctor determined that he had acute coronary syndrome, requiring emergency coronary revascularization intervention.
AI only analyzes based on user data provided, cannot perform clinical examinations, blood pressure measurements, listening to the heart, reading electrocardiograms or prescribing tests. As long as one important piece of information is missing, suggested results may be far from reality.
According to Dr. Nguyen Trung Nguyen, in medicine, many diseases have similar manifestations but completely different levels of danger. Abdominal pain may just be a digestive disorder but it can also be appendicitis or acute pancreatitis. Headache may be due to stress but it can also be cerebral hemorrhage. Cough and fever may be flu but it can also be severe pneumonia or dengue fever. Distinguishing requires examination, testing and clinical experience - things that AI cannot replace.
Missing the "golden time" for treatment
The biggest risk when overly trusting AI is losing the "golden time" for treatment. Stroke, myocardial infarction, anaphylactic shock, meningitis and many other emergency diseases all need to be treated in a very short time. If AI assesses symptoms as "not worrying", or patients understand themselves subjectively, they may delay going to the hospital and lose the opportunity for effective treatment.
AI also cannot observe facial expressions, skin color, breathing rate, alertness or abnormal signs that doctors can recognize immediately upon direct contact with the patient.
Another risk is that many people use AI to ask what medicine to take. AI can synthesize commonly used medications but does not know if the patient has allergies, is pregnant, has liver, kidney, cardiovascular disease, diabetes or is using other medications. Just one common medication can cause serious consequences if used by the wrong person or at the wrong dose, even causing drug interactions, liver damage, kidney failure or severe allergic reactions.
In addition, AI can create a "safe" feeling when using phrases such as "common", "not serious", or vice versa, causing users to worry too much when listing many potential dangerous medical conditions.
AI only supports, not replaces doctors
Assoc. Prof. Dr. Le Tuan Linh - Deputy Director of Hanoi Medical University Hospital said that AI is being widely applied in medicine, from image analysis, disease screening to supporting doctors to make decisions. However, AI is only a supporting tool, and cannot replace examination, diagnosis and prescription.
According to him, the worrying thing is not AI but the fact that many people see AI as an "online doctor". AI only analyzes the data provided by users, cannot conduct clinical examinations or comprehensively evaluate patients, so it may give incorrect suggestions or recommendations for inappropriate medications, increasing the risk of slow treatment and complications.
Assoc. Prof. Dr. Le Tuan Linh emphasized that AI helps optimize the medical examination and treatment process and supports doctors in making decisions, but the final decision must still be made by the doctor. "Medicine is not just data but also clinical experience, responsibility and empathy. In the near future, AI will be a "powerful support arm", and cannot replace the doctor.
Lieutenant Colonel, Dr. Nguyen Hai Cong - Head of the Department of Respiratory Medicine, Military Hospital 175 - said that many people buying medicine on the suggestion of AI without examination can lead to many consequences such as using medicine for the wrong disease, in the wrong dosage, increasing the risk of side effects, drug toxicity and slowing down the treatment process.
Especially, for infectious diseases, delaying going to the hospital can cause serious consequences. Dr. Cong cited that if a patient has pulmonary tuberculosis but assumes that they only have a common respiratory infection and is treated at home, the disease may progress more severely, while increasing the risk of infection for relatives and the community.
According to Dr. Tran Van Phuc (Xanh Pon General Hospital), AI is creating a major step forward in medicine. This tool can give appropriate recommendations in some situations, such as advising patients to go to the hospital instead of self-medicating. However, AI is only a tool for processing data and language, and cannot self-examine patients or make accurate diagnoses based only on the symptoms provided by patients. In the future, AI can take on procedural tasks such as reading films, analyzing data, helping to reduce manual labor time so that doctors can focus on clinical thinking and treatment. AI should be seen as an assistant to support doctors, not a replacement.
