Broad-spectrum antibiotic use at the end of life in advanced cancer clusters in discrete time windows rather than occurring evenly across the final months. In a nationwide South Korean cohort, prescribing peaked between three months and one month before death, while total antibiotic days peaked between one month and two weeks before death. Patterns were consistent across antibiotic classes and cancer types, with greater exposure in hematologic malignancies than in solid tumors. These findings identify a time-bound opportunity to align antimicrobial decisions with patient goals and to target stewardship where it is most likely to change practice.
Among 515,366 decedents with advanced cancer, 55.9% received at least one broad-spectrum agent during the last six months of life. The prescription proportion was highest during three months to one month before death at 28.1%, then declined to 13.3% in the final week. Total consumption, expressed as days of therapy per 1000 patient-days, peaked at 190.0 during one month to two weeks before death. Exposure remained higher in hematologic cancers across all intervals. In the final week, leukemia showed greater odds of receipt compared with lung cancer (odds ratio, 1.50; 95% CI, 1.43 to 1.58) and higher consumption (adjusted relative risk, 1.21; 95% CI, 1.19 to 1.23).
To place the cohort findings into clinical context, Contagion spoke with study investigator Shin Hye Yoo, MD, PhD. In the following email Q&A, he discusses the timing drivers behind antibiotic use, implications for hematologic malignancies, and how to align late-life prescribing with patient goals.
Contagion: Your study showed broad-spectrum antibiotic prescriptions peaked 3 months to 1 month before death, with consumption highest at 1 month to 2 weeks before death. What clinical or systemic factors do you believe drive this timing pattern?
Yoo: “This pattern likely reflects the natural course of advanced cancer. About three months before death, many patients begin to experience a noticeable decline in their physical function, which often leads to more frequent hospital visits and hospitalizations. At that point, doctors tend to prescribe broad-spectrum antibiotics when patients develop fever or other signs that could suggest infection, even if the exact cause is not clear. Closer to the end of life, the overall number of patients receiving antibiotics becomes smaller, but those who do are usually very ill and remain on stronger treatments for longer. Clinically, this suggests that the last few months of life represent a critical window where doctors, patients, and families need to discuss whether antibiotics are likely to improve comfort or quality of life, or whether they may add to the burden of care.”
Contagion: Hematologic malignancy patients consistently had higher exposure to broad-spectrum antibiotics than those with solid tumors. What implications does this have for stewardship and end-of-life care planning in this subgroup?
Yoo: “In patients with blood cancers, the first line of defense against infection—the neutrophil—often does not function properly, and absolute neutrophil counts can remain very low. This results in much deeper immune suppression than is typically seen in solid cancers, leaving these patients extremely vulnerable to infections. Because of this, caring for them outside of the hospital environment is particularly difficult, and transitions to home care or hospice are often far more challenging. Antibiotic use tends to be prolonged and intensive, which can make it harder to provide end-of-life care that fully reflects patient values and goals. These findings suggest that this subgroup faces unique barriers to goal-concordant care, underscoring the need for more research and innovative models that can support home-based or hospice care even in the setting of severe immune compromise.”
Contagion: How can the findings of this nationwide cohort inform strategies to better align antibiotic use with patient goals, particularly in the last weeks of life?
Yoo: “Our study shows that broad-spectrum antibiotic use is not random but clustered in certain periods near the end of life, which gives us an opportunity to be more intentional. However, One of the biggest challenges in real-world practice is that it is very difficult to know exactly when a patient is entering the final phase of life. Because of this uncertainty, doctors often start broad-spectrum antibiotics when a patient deteriorates, but the key is to reassess—whether this is a temporary, reversible setback or whether the patient is moving into the dying process. Our study highlights that there is a window where more deliberate prescribing decisions can be made, and that stewardship efforts need to be flexible rather than rigidly time-based. A practical way forward is to use time-limited antibiotic trials, where treatment is begun but continued only if reassessment shows it is truly helping. This approach allows clinicians to honor patient goals and priorities while also reducing unnecessary burdens. Infectious disease physicians and palliative care teams can work together to support patients and families in these discussions, ensuring that antibiotic use at the end of life is both clinically appropriate and aligned with what matters most to the patient.”
What You Need To Know
Broad-spectrum antibiotic use clusters late in life, with prescriptions peaking 3 to 1 months before death and total days peaking 1 month to 2 weeks before death.
More than half of patients received a broad-spectrum agent in the last six months, and exposure was consistently higher in hematologic cancers, including greater final-week use in leukemia.Time-limited antibiotic trials with early reassessment, supported by infectious disease and palliative care teams, can align treatment with patient goals and reduce unnecessary exposure.
This retrospective, population-based cohort study used the South Korean National Health Insurance Service database to identify decedents with advanced cancer who died between July 1, 2002, and December 31, 2021. Broad-spectrum agents included antipseudomonal beta-lactams, carbapenems, and glycopeptides. Antibiotic exposure was assessed across five end-of-life intervals: six to three months, three months to one month, one month to two weeks, two weeks to one week, and the final week before death. Outcomes were prescription proportion and consumption in days of therapy per 1000 patient-days. Logistic regression estimated odds ratios for prescription proportions, and Poisson regression estimated adjusted relative risks for consumption.
Administrative claims may not capture clinical indications or illness severity, which can confound associations between infection risk, treatment intensity, and timing near death. Findings reflect a single national health system and may not generalize to other settings. Logistic models did not adjust for multiple comparisons. Residual confounding by unmeasured factors, including goals-of-care discussions and palliative intent, is possible.
Broad-spectrum antibiotic use concentrates between three months and two weeks before death in advanced cancer, marking a practical checkpoint for antimicrobial stewardship and shared decision-making. Targeting this interval may reduce unnecessary exposure while supporting goal-concordant care, with particular attention to patients with hematologic malignancies.
Reference
Kim JH, Yu J, Yoo SH, et al. Broad-Spectrum Antibiotic Use at the End of Life in Patients With Advanced Cancer. JAMA Network Open. 2025;8(9):e2530980. doi:10.1001/jamanetworkopen.2025.30980
