Key Points
Question
Was the Life-Sustaining Treatment Decisions Act in South Korea associated with antibiotic utilization patterns among patients with advanced cancer during the end-of-life period?
Findings
In this cohort study including 743 842 decedents from 2012 to 2023, antibiotic prescriptions and days of therapy per 1000 patient-days showed modest increases following the act’s enactment in 2018. Although broad-spectrum antibiotics continued their preexisting upward trajectory, other antibiotic agents exhibited a positive slope shift after the legislation.
Meaning
The findings in this study suggest that the enactment of the Life-Sustaining Treatment Decisions Act was not associated with a reduction in systemic antibiotic use among patients with advanced cancer.
This cohort study evaluates changes in systemic antibiotic use among end-of-life patients with advanced cancer following the 2018 Life-Sustaining Treatment Decisions Act in South Korea.
Abstract
Importance
South Korea’s 2018 Life-Sustaining Treatment Decisions (LSTD) Act was enacted to promote patient autonomy, yet its impact on systemic antibiotic utilization near the end of life in patients with advanced cancer remains unexplored.
Objective
To evaluate shifts in systemic antibiotic use among end-of-life patients with advanced cancer following the LSTD Act.
Design, Setting, and Participants
This cohort study used the National Health Insurance Service database to evaluate decedents with 8 major advanced cancers (lung, liver, stomach, colorectal, pancreatic, prostate, gallbladder or biliary tract, and breast). Inverse probability of treatment weighting was performed to balance cancer types, comorbidities, and demographic characteristics. Data were collected and analyzed between October 2025 and February 2026.
Exposures
Implementation of the LSTD Act, comparing the pre–LSTD Act (2012-2017) and post–LSTD Act (2018-2023) eras.
Main Outcomes and Measures
The primary outcomes were antibiotic prescription proportion and days of therapy (DOT) per 1000 patient-days during the final 6 months of life. Broad-spectrum antibiotics, defined as antipseudomonal β-lactams, carbapenems, or glycopeptides, were separately analyzed. The utilization patterns were evaluated across 5 predefined intervals using odds ratios and risk ratios. A seasonally adjusted monthly interrupted time-series analysis was performed to characterize the longitudinal shifts in DOT during the final months of life.
Results
A total of 743 842 eligible decedents (mean [SD] age, 71.9 [12.2] years; 487 690 male [65.6%]) were evaluated, including 348 813 in the pre–LSTD Act era and 395 029 in the post–LSTD Act era. During the final 6 months of life, the overall prescription proportion increased from 88.3% to 90.8% (odds ratio, 1.34; 95% CI, 1.32-1.37), widening particularly in the final 3 months. Similarly, total antibiotic DOT showed a significant overall increase from 231.4 to 261.6 per 1000 patient-days (risk ratio, 1.14; 95% CI, 1.13-1.14) across all intervals. These 2 measures showed comparable trends across broad-spectrum antibiotics and other classes. Monthly interrupted time-series analysis revealed a significant postenactment slope increase in overall antibiotic DOTs (β = 0.94; 95% CI, 0.67-1.21; P < .001), driven largely by other antibiotics alongside the ongoing increase in broad-spectrum antibiotics.
Conclusions and Relevance
In this retrospective cohort study of patients with advanced cancer, the LSTD Act was not associated with reduced end-of-life antibiotic use, suggesting the need to reshape perceptions of antibiotics as potential life-sustaining therapy.
Introduction
In patients with advanced cancer, achieving a dignified death has become a paramount goal of contemporary end-of-life care. To support this objective, established metrics of life-sustaining treatment (LST), such as cardiopulmonary resuscitation and late-stage chemotherapy, have been closely monitored worldwide to prevent overtreatment.1 However, antibiotic use remains a deeply embedded yet largely unscrutinized component of this trajectory, often escaping regulatory and clinical oversight. Furthermore, rapid advancements in oncologic therapies have paradoxically intensified end-of-life clinical aggressiveness.2,3,4 Consequently, broad-spectrum antibiotic use is highly prevalent among patients with advanced cancer during their final stages of life.5,6,7 This escalation not only exposes vulnerable patients to adverse drug events, but also promotes multidrug-resistant organism colonization.8 Such colonization is a barrier to hospice and palliative care access, depriving patients of comfort-oriented support that is essential to a dignified death.9,10
In 2018, South Korea enacted the Life-Sustaining Treatment Decisions (LSTD) Act to respect patient autonomy and ensure the best interests of terminally ill patients during the legally defined imminent stage of dying. Since its implementation, the act has raised awareness of patient self-determination and timely LST decisions,11 contributing to a decreased reliance on highly invasive measures.12,13,14 However, although primary LST components, including cardiopulmonary resuscitation, mechanical ventilation, hemodialysis, chemotherapy, extracorporeal membrane oxygenation, blood transfusion, and vasopressors, feature as explicit checklist items on official forms, antibiotics are not specifically enumerated and remain ambiguous.
In this context, evaluating post–LSTD Act changes in antibiotic use among patients with advanced cancer during the end-of-life period can help reshape policy and clinical perceptions of antibiotics as a potential LST. We hypothesized that end-of-life antibiotic consumption would decrease or stabilize following the act, resulting from a broader culture shift toward withholding invasive terminal care. Using a nationwide population-based cohort of end-of-life patients with advanced cancer, we analyzed longitudinal prescribing patterns before and after the legislation to illuminate the gap between health policy design and actual clinical practice.
Methods
Data Source
This retrospective cohort study used data from the South Korean National Health Insurance Service (NHIS) database. As a government-managed, single-payer system, the NHIS provides universal health care coverage for approximately 97% of the South Korean population, creating a comprehensive repository of medical activities. The database encompasses longitudinal records of all insured medical services, including demographic characteristics, diagnostic information coded by the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, detailed outpatient and inpatient prescription data, and verified mortality records. Because antibiotics in South Korea are available exclusively through physician prescriptions, the NHIS database identifies data of nearly exhaustive use for antibiotics nationwide. To ensure the inclusion of a well-defined oncology cohort, the patients were identified using the V193 registry code, which is a specific insurance designation for cancer. This registry requires clinical confirmation of malignant neoplasm for activation and remains active for 5 years during active treatment or surveillance, ensuring high diagnostic accuracy for the study population.
Ethical Statement
This study was approved by the institutional review board of Seoul National University Hospital (No. H-2305-088-1431). The requirement for informed consent was waived because of the use of anonymized retrospective data that were routinely collected as part of the health insurance process and linked to mortality data from the Korea National Statistical Office. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.15
Study Population and Definition
We identified a nationwide cohort of adults (aged ≥18 years) with primary malignant neoplasms who used health care services between 2012 and 2023. The final analysis was restricted to decedents whose deaths were confirmed by December 31, 2023, and who maintained a continuous V193 cancer registry code for at least 1 year preceding their death. This threshold served as a pragmatic criterion to ensure a stable baseline of cancer care, minimize missing clinical encounters during end-of-life care, and specifically capture patients with a sufficient clinical time frame to discuss and formulate advance directives or goals of care along a typical advanced cancer trajectory. In this study, we focused on the most prevalent cancer types (lung, liver, stomach, colorectal, pancreatic, prostate, gallbladder or biliary tract, and breast cancers), with patients categorized according to their primary diagnosis to maintain cohort specificity. Patients with hematological malignant neoplasms were excluded because their clinical trajectory and the mandatory nature of high-intensity antibiotic use for febrile neutropenia could greatly differ from those with solid tumors. In addition, decedents with missing data or incomplete clinical records in the NHIS database were excluded. To evaluate the impact of the LSTD Act, the study period was divided into the pre–LSTD Act (2012-2017) and post–LSTD Act (2018-2023) eras.
Outcomes
The primary outcome was the use of systemic antibiotics as defined by the World Health Organization Anatomical Therapeutic Chemical classification system (J01, antibacterials for systemic use). Within this category, broad-spectrum antibiotics were specifically operationalized into 4 classes available in South Korea: antipseudomonal penicillins, antipseudomonal cephalosporins, carbapenems, and glycopeptides (eTable 1 in Supplement 1). To provide a comprehensive assessment of prescription shifts, outcomes were analyzed across 3 distinct categories: (1) all systemic antibiotics (composite all); (2) broad-spectrum antibiotics (composite broad spectrum); and (3) other antibiotics, representing systemic agents not classified as broad-spectrum antibiotics (composite other). The broad-spectrum and other antibiotic categories represent mutually exclusive subsets that collectively comprise all systemic antibiotic use.
Consistent with our prior work on end-of-life antibiotic trajectories,7 the final 6 months of life were stratified into 5 clinical intervals to determine the accelerated functional decline and shifting prescription patterns as death approached: T1, 6 months to 3 months before death; T2, 3 months to 1 month before death; T3, 1 month to 2 weeks before death; T4, 2 weeks to 1 week before death; and T5, the final week of life. This temporal segmentation reflected the rapid functional deterioration typically observed in patients with advanced cancer and aligns with the established evidence of shifting antibiotic trajectories during the end-of-life period.
Antibiotic consumption was quantified by 2 complementary metrics at each interval. First, the prescription proportion was calculated as the percentage of patients who received at least 1 systemic antibiotic dose. Second, the days of therapy (DOT) per 1000 patient-days were used for adjusting for the varying lengths of each interval, which were calculated as the total number of treatment days divided by the aggregate person-days at risk within that period.
Covariates of Interest
We collected a comprehensive set of demographic and clinical variables from the linked NHIS records. Demographic covariates included age, sex, insurance status, household income (stratified by quartiles), and residential location (categorized as the Seoul capital area or provinces). The clinical characteristics focused on the baseline status and disease burden on each patient, including primary cancer type, year of mortality, and Charlson Comorbidity Index score.
Statistical Analysis
Data were collected and analyzed between October 2025 and February 2026. Baseline characteristics were compared using the Student t test for continuous variables and the χ2 test for categorical variables. To minimize the potential confounding factors, we applied inverse probability of treatment weighting (IPTW) using the aforementioned covariates. Covariate balance after weighting was assessed using weighted standardized mean differences, with values less than 0.10 considered acceptable. In addition, covariate balance after propensity score weighting was visually evaluated using Love plots.
The longitudinal impact of the LSTD Act was evaluated on the basis of 3 statistical approaches. First, we applied weighted generalized linear models incorporating stabilized IPTW to compare antibiotic utilization in each discrete interval preceding death (T1-T5). Logistic regression was used to estimate odds ratios (ORs) for prescription proportions, and Poisson regression models with person-days as an offset variable were used to estimate risk ratios (RRs) for DOT per 1000 patient-days. To complement these relative measures, IPTW-adjusted absolute risk differences (RDs) for proportions (expressed in percentage points [pp]) and for DOT (expressed in days) were calculated as absolute effect sizes. Second, a sensitivity analysis was conducted by restricting the cohort to a 4-year window immediately surrounding policy enactment (2016-2017 vs 2018-2019), with IPTW reestimated within this subcohort to evaluate the primary outcomes (OR, RR, and RD) and to minimize potential confounding from long-term secular trends in antimicrobial stewardship guidelines and oncology care. Finally, a seasonally adjusted, segmented interrupted time-series analysis was performed using monthly data from 2012 to 2023. Separate slope and level-change parameters were modeled for each analytical segment to flexibly evaluate immediate level shifts and postenactment slope changes in DOT per 1000 patient-days during the final month of life without imposing rigid timeline assumptions. All statistical tests were 2-sided, with P < .05 considered significant. Statistical analyses were performed using SAS statistical software version 9.4 (SAS Institute).
Results
Baseline Characteristics of the Study Population
Among the 743 842 eligible decedents included in the final analysis (348 813 decedents [46.9%] in the pre–LSTD Act era and 395 029 decedents [53.1%] in the post–LSTD Act era), the mean (SD) age was 71.9 (12.2) years, and 487 690 decedents (65.6%) were male (Table 1). Lung cancer was the most prevalent primary malignant entity (205 183 decedents [27.6%]), followed by liver cancer (136 433 decedents [18.3%]). Following IPTW, all baseline covariates, including age, sex, primary cancer type, and comorbidity burden, were effectively balanced between the 2 groups, with all weighted standardized mean differences falling below 0.10. Covariate balance before and after propensity score weighting is presented as a Love plot in eFigure 1 in Supplement 1.
Table 1. Baseline Characteristics of Study Cohort.
| Characteristic | Decedents, No. (%) | SMD | |||
|---|---|---|---|---|---|
| Total (N = 743 842) | Pre–LSTD Act (n = 348 813) | Post–LSTD Act (n = 395 029) | Unweighted | Weighted | |
| Age, mean (SD), y | 71.9 (12.2) | 70.5 (12.3) | 73.2 (11.9) | .23 | <.001 |
| Sex | |||||
| Male | 487 690 (65.6) | 229 104 (65.7) | 258 586 (65.5) | .005 | <.001 |
| Female | 256 152 (34.4) | 119 709 (34.3) | 136 443 (34.5) | ||
| Insurance | |||||
| National health insurance | 670 076 (90.1) | 315 732 (90.5) | 354 344 (89.7) | .03 | .001 |
| Medical aid | 73 766 (9.9) | 33 081 (9.5) | 40 685 (10.3) | ||
| Household income | |||||
| First quartile (lowest) | 184 500 (24.8) | 79 776 (24.9) | 104 724 (26.5) | .08 | <.001 |
| Second quartile | 80 827 (10.9) | 40 560 (11.6) | 40 267 (10.2) | ||
| Third quartile | 108 991 (14.6) | 51 766 (14.9) | 57 225 (14.5) | ||
| Fourth quartile | 141 431 (19.0) | 68 828 (19.7) | 72 603 (18.4) | ||
| Fifth quartile (highest) | 228 093 (30.7) | 107 883 (30.9) | 120 210 (30.4) | ||
| Residence | |||||
| Seoul capital area | 311 822 (41.9) | 143 724 (41.2) | 168 098 (42.6) | .03 | <.001 |
| Provinces | 432 020 (58.1) | 205 089 (58.8) | 226 931 (57.4) | ||
| Noncancer Charlson Comorbidity Index score | |||||
| 0 | 38 651 (5.2) | 20 973 (6.0) | 17 678 (4.5) | .16 | <.001 |
| 1-2 | 270 540 (36.4) | 138 269 (39.6) | 132 271 (33.5) | ||
| ≥3 | 434 651 (58.4) | 189 571 (54.4) | 245 080 (62.0) | ||
| Clinical diagnosis | |||||
| Lung cancer | 205 183 (27.6) | 96 744 (27.7) | 108 439 (27.4) | .09 | .07 |
| Liver cancer | 136 433 (18.3) | 68 371 (19.6) | 68 062 (17.2) | ||
| Colorectal cancer | 108 510 (14.6) | 50 101 (14.4) | 58 409 (14.8) | ||
| Stomach cancer | 104 974 (14.1) | 54 048 (15.5) | 50 926 (12.9) | ||
| Pancreatic cancer | 69 823 (9.4) | 29 731 (8.5) | 40 092 (10.2) | ||
| Gallbladder and biliary tract cancer | 51 120 (6.9) | 22 363 (6.4) | 28 757 (7.3) | ||
| Prostate cancer | 36 635 (4.9) | 13 704 (3.9) | 22 931 (5.8) | ||
| Breast cancer | 31 164 (4.2) | 13 751 (4.0) | 17, 413 (4.4) | ||
Abbreviations: LSTD, Life-Sustaining Treatment Decisions; SMD, standardized mean difference.
Antibiotic Use Patterns Based on Prescription Proportion
Across the study cohort, the proportion of patients receiving systemic antibiotics exhibited a comparable temporal pattern in both eras, marked by an initial increase from T1 to T2, with a subsequent gradual decline as death approached (Figure 1). Regarding the entire 6-month period, the overall prescription proportion for composite all antibiotics showed a minor increase from 88.3% in the pre–LSTD Act group to 90.8% in the post–LSTD Act group (RD, 2.8 pp; OR, 1.34; 95% CI, 1.32-1.37) (Table 2), and a similarly modest change was observed for composite other antibiotic classes (RD, 2.0 pp; OR, 1.19; 95% CI, 1.17-1.20). However, the prescription proportion for composite broad-spectrum antibiotics showed a more substantial increase (RD, 10.1 pp; OR, 1.52; 95% CI, 1.51-1.54).
Figure 1. Line Graphs of the Proportion of Patients Receiving Antibiotics Across End-Of-Life Intervals.
This figure illustrates the proportion of patients who received broad-spectrum antibiotics during 5 end-of-life time intervals: T1 (6 to 3 months before death), T2 (3 months to 1 month before death), T3 (1 month to 2 weeks before death), T4 (2 weeks to 1 week before death), and T5 (last week of life). Circles represent the observed proportions of patients who received at least 1 dose of each antibiotic class at each interval. The solid line represents the era before the Life-Sustaining Treatment Decisions Act, and the dashed line represents the era after the act.
Table 2. Comparison of Antibiotic Use Before and After Enactment of the LSTD Act.
| Antibiotic type and time period | Prescription proportions, % | Days of therapy per 1000 patient-days | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Pre–LSTD Act | Post–LSTD Act | Weighted RD, pp (95% CI) | Weighted OR (95% CI) | P value | Pre–LSTD Act | Post–LSTD Act | Weighted RD, pp (95% CI) | Weighted RR (95% CI) | P value | |
| All | ||||||||||
| Whole period | 88.3 | 90.8 | 2.8 (2.6 to 2.9) | 1.34 (1.32 to 1.37) | <.001 | 231.4 | 261.6 | 31.1 (29.9 to 32.4) | 1.14 (1.13 to 1.14) | <.001 |
| T1 | 50.2 | 50.3 | 0.2 (−0.1 to 0.4) | 1.01 (0.99 to 1.02) | .09 | 146.7 | 159.5 | 12.7 (11.4 to 14.0) | 1.09 (1.08 to 1.10) | <.001 |
| T2 | 57.4 | 60.6 | 3.5 (3.3 to 3.8) | 1.16 (1.15 to 1.17) | <.001 | 298.4 | 334.7 | 38.8 (36.6 to 40.9) | 1.13 (1.12 to 1.14) | <.001 |
| T3 | 34.5 | 39.6 | 5.5 (5.2 to 5.7) | 1.26 (1.25 to 1.28) | <.001 | 435.4 | 514.1 | 81.7 (77.9 to 85.6) | 1.19 (1.18 to 1.20) | <.001 |
| T4 | 18.7 | 22.7 | 4.2 (4.0 to 4.4) | 1.29 (1.28 to 1.31) | <.001 | 345.4 | 426.2 | 81.4 (77.1 to 85.8) | 1.24 (1.22 to 1.25) | <.001 |
| T5 | 22.5 | 27.1 | 4.4 (4.2 to 4.6) | 1.27 (1.25 to 1.28) | <.001 | 174.6 | 213.0 | 36.1 (34.0 to 38.2) | 1.21 (1.19 to 1.22) | <.001 |
| Broad spectrum | ||||||||||
| Whole period | 54.4 | 63.9 | 10.1 (9.9 to 10.3) | 1.52 (1.51 to 1.54) | <.001 | 63.8 | 83.7 | 21.0 (20.6 to 21.5) | 1.33 (1.32 to 1.34) | <.001 |
| T1 | 14.4 | 17.5 | 3.3 (3.1 to 3.5) | 1.28 (1.26 to 1.30) | <.001 | 28.0 | 36.7 | 9.0 (8.7 to 9.4) | 1.33 (1.30 to 1.35) | <.001 |
| T2 | 26.1 | 31.0 | 5.4 (5.2 to 5.6) | 1.30 (1.29 to 1.32) | <.001 | 87.6 | 110.9 | 25.5 (24.6 to 26.3) | 1.29 (1.28 to 1.31) | <.001 |
| T3 | 17.9 | 22.7 | 5.1 (5.0 to 5.3) | 1.38 (1.36 to 1.39) | <.001 | 152.4 | 203.6 | 53.9 (52.2 to 55.6) | 1.36 (1.34 to 1.37) | <.001 |
| T4 | 9.9 | 13.5 | 3.8 (3.6 to 3.9) | 1.45 (1.42 to 1.47) | <.001 | 120.9 | 173.4 | 54.1 (52.2 to 56.0) | 1.45 (1.43 to 1.47) | <.001 |
| T5 | 12.4 | 17.3 | 4.9 (4.7 to 5.0) | 1.47 (1.45 to 1.49) | <.001 | 61.8 | 89.8 | 28.0 (27.1 to 28.9) | 1.45 (1.43 to 1.48) | <.001 |
| Other | ||||||||||
| Whole period | 85.9 | 87.7 | 2.0 (1.8 to 2.1) | 1.19 (1.17 to 1.20) | <.001 | 167.6 | 177.9 | 10.1 (9.2 to 11.0) | 1.06 (1.05 to 1.07) | <.001 |
| T1 | 49.2 | 49.0 | −0.1 (−0.4 to 0.1) | 0.99 (0.98 to 1.00) | .30 | 118.6 | 122.8 | 3.7 (2.7 to 4.7) | 1.03 (1.02 to 1.04) | <.001 |
| T2 | 55.0 | 57.6 | 2.9 (2.6 to 3.1) | 1.12 (1.11 to 1.13) | <.001 | 210.9 | 223.8 | 13.3 (11.8 to 14.8) | 1.06 (1.06 to 1.07) | <.001 |
| T3 | 30.8 | 34.6 | 3.9 (3.7 to 4.1) | 1.20 (1.18 to 1.21) | <.001 | 283.1 | 310.5 | 27.9 (25.2 to 30.6) | 1.10 (1.09 to 1.11) | <.001 |
| T4 | 15.8 | 18.4 | 2.7 (2.5 to 2.9) | 1.21 (1.20 to 1.23) | <.001 | 224.5 | 252.7 | 27.3 (24.2 to 30.5) | 1.12 (1.11 to 1.14) | <.001 |
| T5 | 17.8 | 19.8 | 1.8 (1.6 to 2.0) | 1.12 (1.11 to 1.14) | <.001 | 112.8 | 123.2 | 8.1 (6.7 to 9.7) | 1.07 (1.06 to 1.09) | <.001 |
Abbreviations: LSTD, Life-Sustaining Treatment Decisions; OR, odds ratio; pp, percentage points; RD, absolute risk difference; RR, risk ratio; T1, 6 months to 3 months before death; T2, 3 months to 1 month before death; T3, 1 month to 2 weeks before death; T4, 2 weeks to 1 week before death; T5, last week of life.
Analysis of the 5 intervals revealed distinct prescription patterns between antibiotic classes (Table 2 and Figure 1). Regarding the composite all antibiotics, use showed a minimal change during T1 (RD, 0.2 pp; OR, 1.01; 95% CI, 0.99-1.02) and modest increases from T2 (RD, 3.5 pp; OR, 1.16; 95% CI, 1.15-1.17) through T5 (RD, 4.4 pp; OR, 1.27; 95% CI, 1.25-1.28). In contrast, the composite broad-spectrum category exhibited consistently higher prescription rates across all intervals, starting from T1 (RD, 3.3 pp; OR, 1.28; 95% CI, 1.26-1.30) through T5 (RD, 4.9 pp; OR, 1.47; 95% CI, 1.45-1.49). Sensitivity analyses replicated these trajectories, albeit with slightly smaller rate differences (eTable 2 in Supplement 1).
Antibiotic Use Patterns Based on DOT
DOT per 1000 patient-days showed an overall increase across the 6-month end-of-life trajectory in the post–LSTD Act group (Table 2 and Figure 2). In regard to the entire period, all antibiotics use increased from 231.4 to 261.6 DOT per 1000 patient-days (RD, 31.1 pp; RR, 1.14; 95% CI, 1.13-1.14), a trend that was more pronounced for the composite broad-spectrum category (RD, 21.0 pp; RR, 1.33; 95% CI, 1.32-1.34) than in the composite other category (RD, 10.1 pp; RR, 1.06; 95% CI, 1.05-1.07). Although differences in prescription proportions were minimal at T1 and widened only starting from T2, all antibiotic use was consistently higher in the post–LSTD Act group starting from the earliest interval, T1 (RD, 12.7 pp; RR, 1.09; 95% CI, 1.08-1.10), a disparity that persisted through T5 (RD, 36.1 pp; RR, 1.21; 95% CI, 1.19-1.22). Sensitivity analyses showed similar patterns across intervals, although the rate differences were smaller in magnitude than those in the overall cohort (eTable 2 in Supplement 1).
Figure 2. Line Graphs of Days of Therapy (DOT) for Antibiotics Across End-of-Life Intervals.
This figure shows DOT per 1000 patient-days for each antibiotic class across the 5 end-of-life intervals: T1 (6 to 3 months before death), T2 (3 months to 1 month before death), T3 (1 month to 2 weeks before death), T4 (2 weeks to 1 week before death), and T5 (last week of life). Circles represent the observed DOT per 1000 patient-days for each antibiotic class at each time interval. The solid line represents the era before the Life-Sustaining Treatment Decisions Act, and the dashed line represents the era after the act.
Quarterly Trends in Antibiotic Days of Therapy
Overall antibiotic use showed a significant upward trend following the act, driven by a positive slope change (β = 0.94; 95% CI, 0.67 to 1.21; P < .001) (Figure 3A). In subanalysis, composite broad-spectrum antibiotics exhibited a continuous upward trend throughout the entire study period, with no significant change in slope after the act (β = 0.09; 95% CI, −0.01 to 0.18; P = .06) (eFigure 2 in Supplement 1). In contrast, the composite other category, which remained relatively stable before 2018, shifted to a significant upward trajectory only after the enactment of the act (β = 0.85; 95% CI, 0.60 to 1.11; P < .001) (eFigure 3 in Supplement 1). Testing for interaction revealed a significant divergence in slope changes between the broad-spectrum and other antibiotic categories (slope-change difference, −0.76; 95% CI, −1.03 to −0.50; P < .001) (Figure 3B).
Figure 3. Line Graphs of Interrupted Time Series Analysis of Antibiotic Days of Therapy (DOT) Before and After the Life-Sustaining Treatment Decisions Act.
Graphs show interrupted time series analysis results with seasonal adjustment for monthly antibiotic utilization during the 1-month period preceding death are presented. The vertical dashed line indicates the implementation of the act in January 2018. Blue markers and solid lines represent observed monthly DOT per 1000 patient-days and seasonally adjusted fitted trends during the era before the act (2012-2017), whereas orange markers and solid lines represent those during the era after the act (2018-2023). Dotted lines depict the counterfactual projected trends, assuming trajectories before or after the act had continued without intervention. Triangular makers denote composite broad-spectrum antibiotics, and circular markers denote composite other antibiotics.
Discussion
This cohort study found that, contrary to its intended goal of promoting a dignified death by reducing invasive end-of-life care, enactment of the 2018 LSTD Act in South Korea was not associated with a mitigation of antibiotic use; rather, it accompanied a continued increase in antibiotic consumption during the final months of life. To our knowledge, this is the first nationwide study to examine how national LST legislation influences antibiotic utilization patterns among patients with advanced cancer near the end of life. Our findings suggest that legislative constraints on primary invasive interventions do not automatically curtail antibiotic prescribing, indicating that unless antibiotics are recognized alongside other life-sustaining measures, they may continue to be utilized as routine practice in terminal care.
Several structural and clinical factors contribute to an inevitable baseline increase in antibiotic use during the final months of life in patients with cancer. First, advances in oncology, specifically the shift toward targeted and less toxic therapies, have extended survival. This progress often results in a prolonged, medically fragile state during the final months of life,16 thereby expanding the window of vulnerability to opportunistic infections and necessitating intensive antimicrobial support. Second, the widespread use of novel agents, such as immunotherapy, introduces unique treatment-related complications, including immune-related adverse events that often mimic systemic infections.17,18 Because these events are often clinically indistinguishable from bacterial infections, they frequently prompt the empirical use of antibiotics. Third, the increasing prevalence of multidrug-resistant organisms in tertiary medical centers inherently leads to higher consumption of broad-spectrum antibiotics.19,20
However, given that other major LST interventions have reportedly been curtailed or stabilized after the legislation under similar clinical scenarios, the post–LSTD Act increase in antibiotic use may appear to be associated with factors unique to antimicrobial decision-making. Specifically, because antibiotics are not explicitly specified as LST withholding items in statutory forms, antibiotics may serve as a clinical compromise during end-of-life discussions fostered by the act. When invasive measures are withheld, these discussions can heighten families’ fears of abandonment while simultaneously imposing substantial ethical burden on clinician.21,22 In response, clinicians may utilize antibiotics as a noninvasive therapeutic bridge to reassure families of ongoing care and to alleviate their own moral guilt, a phenomenon consistent with observed prescribing increases during transitions to palliative care.5 Thus, by regulating only overt LSTs, the act may inadvertently sustain antibiotics in terminal care.
Another possible factor relates to the restrictive legal definition of imminent stage of dying under the current LSTD Act.23 The legislation primarily focuses on patients expected to die within days, leaving a clinical gray zone during the preceding weeks or months of gradual decline. This narrow framework may act as a potential barrier to the timely initiation of palliative care consultations or hospice transition interventions, which are known to curtail broad-spectrum antibiotic prescribing near the end of life.24,25,26 Within this legal constraint, clinicians may find it difficult to shift the primary focus toward comfort-oriented care until a patient is officially certified as being in this imminent stage.27,28 Consequently, by deferring formal legal recognition to the final days of life while omitting antibiotics from explicit decision checklists, the current framework allows routine antibiotic prescribing to persist unexamined during the preceding weeks of clinical decline.
To address these limitations, several institutional strategies may warrant consideration. First, modifying official decision forms to explicitly incorporate antibiotics as a component of LST could serve as a pivotal step to reshape clinical perceptions and discourage unexamined routine prescribing. Second, reexamining the legal framework regarding imminent stage of dying under the LSTD Act could be valuable, as the current focus on the final days of life may narrow the window for earlier goals-of-care discussions and palliative care transitions.29 Finally, a collaborative framework between palliative care teams and antimicrobial stewardship programs may warrant consideration as a supportive strategy.30,31 Such an interdisciplinary model could help clinicians weigh the clinical necessity of antibiotics alongside the overall goals of palliative care, ensuring that antimicrobial practices align with patient-centered comfort and the reduction of nonbeneficial interventions near death.
Limitations
This study has some limitations. First, since this study was based on retrospective administrative claims data, precisely ascertaining specific clinical indications for each antibiotic prescription was not possible. The lack of detailed clinical information, such as microbiological results, radiographic findings, or the severity of suspected infections, limited our ability to distinguish between appropriate therapeutic use and potentially nonbeneficial administration. Second, the administrative nature of the dataset did not determine the evolving perceptions and preferences of patients and their families of end-of-life care. Shared decision-making processes, which are crucial for determining the intensity of antimicrobial therapy, involve complex psychosocial factors that are not fully reflected in billing records. Third, requiring the V193 cancer code for at least 1 year prior to death ensured stable longitudinal tracking but may introduce survival bias. This operational definition could exclude patients with hyperacute deterioration or aggressive malignant neoplasms while potentially incorporating deaths driven by other acute medical events. Fourth, although IPTW was applied to balance baseline characteristics between the eras, there may still be unadjusted variables, such as hospital-specific protocols or individual physician attitudes toward the LSTD Act, which could influence prescription patterns. Fifth, our study period (2012-2023) spanned the COVID-19 pandemic.32 The accelerated surge in antibiotic consumption observed from 2020 onwards makes it difficult to disentangle the cumulative, long-term impact of the LSTD Act from pandemic-related confounders, such as modified infection control protocols or heightened empirical prescribing for febrile oncology patients.33
Conclusions
In this retrospective cohort study of patients with advanced cancer, enactment of the LSTD Act was not associated with reduced end-of-life antibiotic consumption, which instead showed a modest increasing trend. These findings suggest that unless antibiotics are conceptualized as a potential form of LST rather than mere supportive care, antibiotic therapy may continue to serve as a routine component of terminal care. Future strategies should refine legislative and clinical frameworks to explicitly account for antibiotics, supported by collaborative antimicrobial stewardship in palliative settings, to promote dignified end-of-life care.
eTable 1. The broad-spectrum antibiotics lists
eTable 2. Comparison of antibiotic use pre- and post-enactment of the LSTD Act (2016–2019)
eFigure 1. Love plot of absolute standardized mean differences (SMDs) before and after inverse probability of treatment weighting (IPTW)
eFigure 2. Interrupted time-series analysis (ITSA) of monthly weighted days of therapy (DOT) per 1000 patient-days for composite broad-spectrum antibiotics (2012–2023)
eFigure 3. Interrupted time-series analysis (ITSA) of monthly weighted days of therapy (DOT) per 1000 patient-days for composite other (2012–2023)
Data Sharing Statement
References
- 1.Abedini NC, Hechtman RK, Singh AD, et al. Interventions to reduce aggressive care at end of life among patients with cancer: a systematic review. Lancet Oncol. 2019;20(11):e627-e636. doi: 10.1016/S1470-2045(19)30496-6 [DOI] [PubMed] [Google Scholar]
- 2.Earle CC, Neville BA, Landrum MB, Ayanian JZ, Block SD, Weeks JC. Trends in the aggressiveness of cancer care near the end of life. J Clin Oncol. 2004;22(2):315-321. doi: 10.1200/JCO.2004.08.136 [DOI] [PubMed] [Google Scholar]
- 3.Teno JM, Gozalo PL, Bynum JP, et al. Change in end-of-life care for Medicare beneficiaries: site of death, place of care, and health care transitions in 2000, 2005, and 2009. JAMA. 2013;309(5):470-477. doi: 10.1001/jama.2012.207624 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Ma Z, Li H, Zhang Y, et al. Prevalence of aggressive care among patients with cancer near the end of life: a systematic review and meta-analysis. EClinicalMedicine. 2024;71:102561. doi: 10.1016/j.eclinm.2024.102561 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Marra AR, Clore GS, Balkenende E, et al. Association of entry into hospice or palliative care consultation during acute care hospitalization with subsequent antibiotic utilization. Clin Microbiol Infect. 2023;29(1):107.e1-107.e7. doi: 10.1016/j.cmi.2022.07.018 [DOI] [PubMed] [Google Scholar]
- 6.Kim NH, Han K, Ji E, et al. Trends of antibiotic use at the end-of-life of cancer and non-cancer decedents: a nationwide population-based longitudinal study (2006-2018). Antimicrob Steward Healthc Epidemiol. 2024;4(1):e83. doi: 10.1017/ash.2024.75 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Kim JH, Yu J, Yoo SH, Sim JA, Keam B, Heo DS. Broad-spectrum antibiotic use at the end of life in patients with advanced cancer. JAMA Netw Open. 2025;8(9):e2530980. doi: 10.1001/jamanetworkopen.2025.30980 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Tamma PD, Avdic E, Li DX, Dzintars K, Cosgrove SE. Association of adverse events with antibiotic use in hospitalized patients. JAMA Intern Med. 2017;177(9):1308-1315. doi: 10.1001/jamainternmed.2017.1938 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Kim JH, Yu J, Lee W, et al. Association between multidrug-resistant organism status and quality of end-of-life care in patients with advanced cancer referred to palliative care: a retrospective cohort study with nationwide data linkage. Clin Microbiol Infect. 2026;32(5):822-828. doi: 10.1016/j.cmi.2025.11.032 [DOI] [PubMed] [Google Scholar]
- 10.Dagli O, Tasdemir E, Ulutasdemir N. Palliative care infections and antibiotic cost: a vicious circle. Aging Male. 2020;23(2):98-105. doi: 10.1080/13685538.2019.1575353 [DOI] [PubMed] [Google Scholar]
- 11.Kim H, Im HS, Lee KO, et al. Changes in decision-making process for life-sustaining treatment in patients with advanced cancer after the life-sustaining treatment decisions-making act. BMC Palliat Care. 2021;20(1):63. doi: 10.1186/s12904-021-00759-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Lee YJ, Ahn S, Cho JY, et al. Change in perception of the quality of death in the intensive care unit by healthcare workers associated with the implementation of the “well-dying law”. Intensive Care Med. 2022;48(3):281-289. doi: 10.1007/s00134-021-06597-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Lee SH, Kim JM, Yeo Y, Kim J. The impact of the Well-Dying Law in Korea: comparing clinical characteristics and ICU admissions. Ann Palliat Med. 2022;11(10):3135-3146. doi: 10.21037/apm-22-509 [DOI] [PubMed] [Google Scholar]
- 14.Oh TK, Song IA. Impact of the Life-Sustaining Treatment Decision Act on the incidence and outcomes of in-hospital cardiopulmonary resuscitation. Crit Care Med. 2026;54(7):1691-1698. doi: 10.1097/CCM.0000000000007136 [DOI] [PubMed] [Google Scholar]
- 15.von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP; STROBE Initiative . Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. BMJ. 2007;335(7624):806-808. doi: 10.1136/bmj.39335.541782.AD [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Teno JM, Gozalo P, Trivedi AN, et al. Site of death, place of care, and health care transitions among US Medicare beneficiaries, 2000-2015. JAMA. 2018;320(3):264-271. doi: 10.1001/jama.2018.8981 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Atchley WT, Alvarez C, Saxena-Beem S, et al. Immune checkpoint inhibitor-related pneumonitis in lung cancer: real-world incidence, risk factors, and management practices across six health care centers in North Carolina. Chest. 2021;160(2):731-742. doi: 10.1016/j.chest.2021.02.032 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Michot JM, Bigenwald C, Champiat S, et al. Immune-related adverse events with immune checkpoint blockade: a comprehensive review. Eur J Cancer. 2016;54:139-148. doi: 10.1016/j.ejca.2015.11.016 [DOI] [PubMed] [Google Scholar]
- 19.Yoo JH. Antimicrobial resistance—the ‘real’ pandemic we are unaware of, yet nearby. J Korean Med Sci. 2025;40(19):e161. doi: 10.3346/jkms.2025.40.e161 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Wang L, Chen H, Zhang Y, et al. Global antibiotic consumption and regional antimicrobial resistance, 2010-21: an analysis of pharmaceutical sales and antimicrobial resistance surveillance data. Lancet Glob Health. 2025;13(11):e1880-e1891. doi: 10.1016/S2214-109X(25)00308-0 [DOI] [PubMed] [Google Scholar]
- 21.Lee S, Datta R. Frontiers in antimicrobial stewardship: antimicrobial use during end-of-life care. Antimicrob Steward Healthc Epidemiol. 2023;3(1):e164. doi: 10.1017/ash.2023.207 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Phua J, Joynt GM, Nishimura M, et al. ; ACME Study Investigators and the Asian Critical Care Clinical Trials Group . Withholding and withdrawal of life-sustaining treatments in intensive care units in Asia. JAMA Intern Med. 2015;175(3):363-371. doi: 10.1001/jamainternmed.2014.7386 [DOI] [PubMed] [Google Scholar]
- 23.Heo DS, Yoo SH, Keam B, Yoo SH, Koh Y. Problems related to the Act on Decisions on Life-Sustaining Treatment and directions for improvement. J Hosp Palliat Care. 2022;25(1):1-11. doi: 10.14475/jhpc.2022.25.1.1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Kim JH, Yoo SH, Keam B, Heo DS. Antibiotic prescription patterns during last days of hospitalized patients with advanced cancer: the role of palliative care consultation. J Antimicrob Chemother. 2023;78(7):1694-1700. doi: 10.1093/jac/dkad156 [DOI] [PubMed] [Google Scholar]
- 25.Howley F, Jones E, Anderson C, et al. Factors influencing antimicrobial choice and duration during the last month of life in hospitalized patients. Open Forum Infect Dis. 2025;12(11):ofaf670. doi: 10.1093/ofid/ofaf670 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Sim JA, Yu J, Bae J, et al. Temporal association of palliative care consultation with the trajectory of broad-spectrum antibiotic use at the end of life in advanced cancer: a nationwide linked cohort study. Int J Infect Dis. 2026;169:108779. doi: 10.1016/j.ijid.2026.108779 [DOI] [PubMed] [Google Scholar]
- 27.Park HY, Kim MS, Yoo SH, et al. For the universal right to access quality end-of-life care in Korea: broadening our perspective after the 2018 Life-Sustaining Treatment Decisions Act. J Korean Med Sci. 2024;39(12):e123. doi: 10.3346/jkms.2024.39.e123 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Baek AR, Hong SB, Bae S, et al. ; Korean HAP/VAP Study Group . Comparison of the end-of-life decisions of patients with hospital-acquired pneumonia after the enforcement of the life-sustaining treatment decision act in Korea. BMC Med Ethics. 2023;24(1):52. doi: 10.1186/s12910-023-00931-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Goedken CC, Balkenende E, Livorsi D, et al. Improving shared decision-making around antimicrobial-prescribing during the end-of-life period: a qualitative study of Veterans, their support caregivers and their providers. Antimicrob Steward Healthc Epidemiol. 2024;4(1):e89. doi: 10.1017/ash.2024.61 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Karlin D, Pham C, Furukawa D, et al. State-of-the-art review: use of antimicrobials at the end of life. Clin Infect Dis. 2024;78(3):e27-e36. doi: 10.1093/cid/ciad735 [DOI] [PubMed] [Google Scholar]
- 31.Karlin D, Vijayan T. The path ahead: building community between palliative care and infectious disease teams. Clin Microbiol Infect. 2026;32(5):702-703. doi: 10.1016/j.cmi.2026.01.029 [DOI] [PubMed] [Google Scholar]
- 32.Shim E, Choi W, Kwon D, Kim T, Song Y. Transmission potential of the omicron variant of severe acute respiratory syndrome coronavirus 2 in South Korea, 25 November 2021-8 January 2022. Open Forum Infect Dis. 2022;9(7):ofac248. doi: 10.1093/ofid/ofac248 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Kim T. Lessons from the coronavirus disease 2019 pandemic: implications for antimicrobial stewardship for COVID-19 management. J Korean Med Sci. 2026;41(11):e72. doi: 10.3346/jkms.2026.41.e72 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eTable 1. The broad-spectrum antibiotics lists
eTable 2. Comparison of antibiotic use pre- and post-enactment of the LSTD Act (2016–2019)
eFigure 1. Love plot of absolute standardized mean differences (SMDs) before and after inverse probability of treatment weighting (IPTW)
eFigure 2. Interrupted time-series analysis (ITSA) of monthly weighted days of therapy (DOT) per 1000 patient-days for composite broad-spectrum antibiotics (2012–2023)
eFigure 3. Interrupted time-series analysis (ITSA) of monthly weighted days of therapy (DOT) per 1000 patient-days for composite other (2012–2023)
Data Sharing Statement
