Symptom Documentation Differences in Acute Cancer Care Suggest Sociodemographic Disparities

Unplanned acute care is common among patients with cancer, but there are differences in symptom documentation across racial, sex, and socioeconomic subgroups, according to a study published in JAMA Network Open. The findings highlight possible disparities and suggest a need for interventions that would promote equity and improve symptom management for patients with cancer.

The study highlighted symptom patterns in the short-term preceding ED visits and unplanned hospitalizations using a previously validated natural language processing a pipeline. | Image credit: Tyler Olson - stock.adobe.com

“Despite well-documented data on disparities in cancer care management and unmet symptom management needs, there remains a knowledge gap in understanding the sociodemographic differences in symptom identification and how they may potentially impact cancer care management,” the authors noted. In the new analysis, they aimed to illuminate symptom patterns in the short term preceding emergency department (ED) visits and unplanned hospitalizations using a previously validated natural language processing (NLP) pipeline.

In the US, 1,777,566 new cancer cases were reported in 2021, and 608,366 people died of cancer in 2022, according to the CDC data.2 “Many of these patients will require unplanned acute care with an ED visit or hospital admission due to acute complications from the disease, symptoms from treatment, or comorbidities,” the authors wrote.1 “Such unplanned acute care can greatly affect patient outcomes, quality of life, and health care system spending.”

The researchers used NLP, which uses machine learning to process and evaluate language and identify patterns, to assess routine clinical documentation to characterize cancer symptoms documented in the 30 days prior to acute care. A total of 28,709 patients treated at a single tertiary-care institution between January 1, 2013, and December 31, 2023, had 70,606 acute care visits and 854,830 preceding symptoms. The median age was 61 (IQR, 48-70) years, 53.62% of patients who had acute care encounters were men, and 56.64% were White.

The most common symptoms were pain, nausea, and vomiting, which accounted for 7.54%, 6.74%, and 5.79% of documented symptoms, respectively. Other common symptoms were fatigue (5.26%), constipation (3.93%), fever (3.39%), generalized muscle weakness (3.32%), extremity edema (3.28%), dyspnea (3.12%), and headache (2.92%).

Overall, 35,998 encounters had high symptom burdens, defined as 10 or more symptoms reported, and patients who were female, Asian, Black, American Indian or Alaska Native, or covered by Medicaid were more likely to have high symptom burdens prior to acute care visits. The adjusted odds ratio (aOR) for a high symptom burden prior to acute care was 1.14 (95% CI, 1.10-1.18; P < .001) among women, 1.22 (1.17-1.28; P < .001) for Asian patients, 1.17 (95% CI, 1.10-1.25; P < .001) for Black patients, 1.21 (95% CI, 1.01-1.44; P = .04) for American Indian or Alaska Native patients, and 1.10 (95% CI, 1.05-1.14; P < .001) for Medicaid-insured patients.

On the other hand, those who were 65 years or older or uninsured were less likely to have a high symptom burden prior to acute care events ([aOR, 0.96; 95% CI, 0.92-1.00; P = .04] and [aOR, 0.58; 95% CI, 0.45-0.76; P < .001], respectively).

Women had higher rates of symptom documentation vs men across the 10 most common symptoms, but encounters with patients aged 65 or older were less likely to document 9 of the 10 most common symptoms. Generalized muscle weakness was more commonly reported in patients who were of races or ethnicities other than White, with Asian and Black or African American individuals having the greatest likelihood of muscle weakness ([OR, 1.40; 95% CI, 1.34-1.46; P < .001] and [OR, 1.36; 95% CI, 1.28-1.44; P < .001], respectively).

The study was limited by its retrospective nature, inclusion of patients from a single institution, and the fact that most patients were White and had insurance, which means the findings may not be generalizable to all populations. However, the identification of sociodemographic differences in symptoms warrants further investigation to explore potential disparities, the authors noted.

“To our knowledge, this is the first analysis to characterize NLP-extracted documented symptoms preceding acute care visits and their association with specific sociodemographic and clinical variables with a large dataset of patient records that span over a decade,” the authors wrote. “This analysis highlights differences in cancer symptom documentation across racial, sex, and socioeconomic subgroups, suggesting potential areas of disparities. This raises attention to the potential need to develop targeted interventions to ensure equitable access to health care for improved symptom management.”

References

1. Chang C, Chen JJ, Feng J, et al. Patterns in symptoms preceding acute care in patients with cancer. JAMA Netw Open. Published online April 22, 2025. doi:10.1001/jamanetworkopen.2025.6366
2. Cancer data and statistics. CDC. Updated February 13, 2025. Accessed April 22, 2025. https://www.cdc.gov/cancer/data/index