Stressful life events recorded in 39.7 million clinical notes: associations with mental disorders, comorbidity, and frailty
By applying natural language processing to 39.7 million clinical notes, researchers successfully identified stressful life events and demonstrated their strong association with increased risks of mental disorders, comorbidity, and frailty, while also revealing significant underrecording of specific events like violence and financial distress in clinical settings.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer
Human health is often measured by what can be counted: blood pressure, cholesterol levels, or the presence of a specific virus. Yet, a person's life is also shaped by events that are harder to quantify but just as powerful: the death of a loved one, the end of a marriage, or the crushing weight of debt. These stressful moments are known to wear down both the mind and the body, yet they rarely appear in the organized, structured lists that doctors use to track patient care. Instead, these crucial details hide within the free-flowing, unstructured notes that clinicians write during visits. Because these notes are written in natural language rather than coded data, they have historically been difficult to study on a large scale, leaving a gap in our understanding of how life's hardest moments translate into long-term health risks.
A team of researchers in Finland has now bridged that gap by turning to a technology called natural language processing. This method allows computers to read and understand human language, much like a skilled reader scanning a book for specific themes. The researchers applied this tool to a massive archive of nearly forty million clinical notes from over two hundred thousand patients. Their goal was to find mentions of five specific types of difficult life events: grief, divorce, financial trouble, violence, and family crises. By teaching the computer to recognize these patterns in Finnish text, they could extract these hidden stories from the medical records and see how they connected to the patients' future health.
The study revealed that while these events were often overlooked in the structured parts of medical records, they were indeed present in the narrative notes, though not as frequently as they likely occur in real life. The computer found that about twenty-two percent of the patients had a note mentioning grief, while ten percent had notes about financial distress. Smaller percentages were recorded for divorce, violence, and family problems. The researchers found that the computer was highly accurate in its reading, correctly identifying these events in the vast majority of cases. However, when they compared their findings to national statistics, it became clear that many events were simply not being written down. For instance, while national surveys suggest that violence affects a significant portion of the population, it appeared in the clinical notes of less than one percent of patients in any given year. This suggests that many people either do not disclose these traumatic experiences to their doctors, or that doctors do not always record them, perhaps due to time pressures or discomfort.
Despite this under-recording, the events that did appear in the notes told a clear and consistent story about health. The researchers found that patients with documented stressful life events were significantly more likely to develop mental health conditions later on. The risk was not limited to just one type of disorder; it extended to stress-related conditions, anxiety, and depression. The more different types of life events a person had documented, the higher their risk became. For example, a patient with multiple recorded life events was nearly twice as likely to develop anxiety or depression compared to someone with no such records. This pattern held true even when the researchers accounted for other factors like age and sex.
The impact of these life events was not limited to mental health; they were also linked to physical decline. The study measured comorbidity, which is the number of different long-term health conditions a person has, and frailty, a measure of how weak or vulnerable a person's body has become. Patients with documented stressful events showed higher rates of both. The connection was particularly strong for financial distress, which was associated with the highest increase in physical health risks. Interestingly, the study also found that the link between these life events and physical decline was weaker in patients who already had a diagnosed mental health condition. This suggests that once a mental health diagnosis is present, the additional physical toll of a life event might be less distinct, or that the mental health diagnosis itself already accounts for much of the risk.
The authors conclude that while stressful life events are currently scattered and sparse in medical records, the information they contain is vital. The fact that these events are linked to such serious health outcomes means that they are not just background noise but are central to a patient's well-being. By using technology to make these hidden notes visible, doctors could potentially gain a more complete picture of their patients' lives. This would allow them to weigh the medical risks of a disease alongside the psychosocial risks of a difficult life, leading to care that addresses the whole person rather than just their symptoms. The study does not claim to have solved the problem of recording these events, nor does it suggest that the computer can predict the future with certainty, but it clearly demonstrates that the stories patients tell in their notes are powerful indicators of their health trajectory.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.