Observability of Fifth Universal Definition of Myocardial Infarction components in linked electronic health records: a retrospective MIMIC-IV study
This retrospective MIMIC-IV study demonstrates that while linked electronic health records can partially operationalize components of the Fifth Universal Definition of Myocardial Infarction, significant gaps in data availability—particularly regarding ECG reports, imaging, and adjudication—limit the feasibility of fully retrospective evaluation without local prospective verification.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
When a person arrives at an emergency room with chest pain, doctors face a critical puzzle: is the heart muscle dying because of a blocked artery, or is the damage caused by something else entirely, like a severe infection or low oxygen? To solve this, modern medicine relies on a specific set of rules called the Universal Definition of Myocardial Infarction. These rules require doctors to look for a specific pattern in blood tests that measure heart stress, combined with evidence of acute ischemia, which is a sudden lack of blood flow. This evidence usually comes from a combination of the patient's description of their pain, an electrocardiogram that records the heart's electrical activity, and sometimes imaging scans. However, in the real world, this information is scattered across different computer systems, written in different formats, and recorded at different times. The challenge for researchers is not just to find these data points, but to determine if they are actually linked together in a way that allows a computer to understand the full clinical picture without a human doctor sitting there to connect the dots.
A team of researchers set out to test whether existing hospital computer records could reliably reconstruct this complex definition of a heart attack. They did not try to diagnose new patients or change how doctors treat them. Instead, they performed a careful audit of historical records from a large database of intensive care patients, looking specifically at adults who had shown signs of heart stress in their blood tests. Their goal was to see if the digital files contained all the necessary pieces of the puzzle: the blood test results, the patient's reported symptoms, the machine-generated heart rhythm readings, and the specific timing required to make a diagnosis. They treated the hospital records like a set of building blocks, checking if every single block needed to build the definition was actually present, properly labeled, and connected to the right person at the right time.
The researchers found that while the computer records contained some of the necessary information, they were missing many of the critical pieces required to fully apply the modern definition of a heart attack. The blood test data was available, and the text describing a patient's chief complaint was present for everyone, but these were only partial clues. The computer records lacked the specific, detailed information needed to distinguish between different types of heart attacks or to confirm if a procedure had caused the injury. Most notably, the records did not contain systematic, time-stamped imaging results or the specific evidence of blood clots that the definition requires. In many cases, the computer simply did not have the data, rather than having data that said the condition was absent. This distinction is vital: a missing file is not the same as a negative test result.
When the team looked at the electrocardiogram reports, which are the machine readings of the heart's electrical signals, they found that these were available for about four out of every five patients within the strict time window they set. For the remaining patients, the report was either missing or not linked to the correct visit. Among those who had a report, the machine flagged a positive sign of a heart attack in a small number of cases. The researchers also noticed that the way symptoms were recorded in the text of the medical notes was often the only clue available, and this text was not a complete medical history but rather a snapshot of what the patient said upon arrival. Because the data was incomplete and scattered, the researchers concluded that the computer records could not be used to automatically diagnose a heart attack or to judge the quality of care provided.
The study also examined whether the presence of these scattered clues could predict who would survive the hospital stay. They found that patients who had both a reported symptom and a positive machine reading had a different outcome profile than those who had neither, but the data was too messy to draw firm conclusions about cause and effect. The researchers emphasized that the combination of symptoms and machine readings they found was driven more by how doctors wrote their notes than by a clear biological pattern. In fact, the presence of a symptom alone was associated with a lower risk of death in their analysis, while a positive machine reading was associated with a higher risk, showing that these two pieces of information were telling different stories that could not be simply merged.
Ultimately, the researchers determined that the current state of electronic health records is not ready to automatically apply these complex medical definitions. The data exists, but it is not organized in a way that allows a computer to see the whole picture. The study showed that without a human expert to review the original records, verify the timing, and fill in the gaps, any attempt to use these records for diagnosis would be flawed. The authors suggest that before any new computer system is used to alert doctors or make diagnoses, hospitals must first perform their own local checks to ensure that the necessary data is actually there, properly linked, and accurate. The work serves as a reminder that while technology can store vast amounts of information, it cannot yet replace the careful, human judgment required to piece together a medical diagnosis from a fragmented digital history.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.