Abstract
Background
Aggressive end-of-life (EOL) care is common in hospitals, yet contemporary evidence from mainland China comparing cancer and non-cancer decedents—particularly regarding timing of interventions—is limited.
Objectives
To quantify Aggressive EOL care intensity and timing during terminal hospitalization among cancer and non-cancer decedents, identify associated factors, and assess temporal trends.
Methods
We conducted a retrospective observational study of adult in-hospital decedents (≥ 18 years) at a tertiary teaching hospital in Nanchang, China (August 1, 2017–December 31, 2024). The primary outcome was a core Aggressive EOL care intensity score (0–5), defined as the unweighted sum of five domains (CPR-related interventions, respiratory support, renal support, circulatory support, and ICU death). Secondary outcomes were high-intensity Aggressive EOL care (score ≥ 2 and ≥ 3). Timing was assessed using population-level occurrence within ≤ 7 days before death and among recipients, the most proximal (final recorded) occurrence relative to death. Multivariable Poisson regression with robust standard errors and logistic regression were used. Temporal trends were assessed using adjusted generalized linear models.
Results
Among 3,205 decedents, 990 (30.9%) had cancer. Non-cancer decedents had higher Aggressive EOL care use across all domains and higher intensity (≥ 3 domains: 22.9% vs. 6.7%). Non-cancer status was associated with a higher core aggressive EOL care intensity score (adjusted IRR 2.22, 95% CI 1.96–2.44), and emergency admission showed the strongest association with higher intensity (IRR 1.75, 1.66–1.85). Among recipients, the most proximal CPR-related intervention occurred within ≤ 7 days in 98.2% and on the day of death in 80.4%. Temporal trends differed by cancer status (P for interaction = 0.047).
Conclusions
Aggressive EOL care is common and occurred predominantly near death, with substantially higher multi-domain intensity among non-cancer decedents. Emergency admission emerges as a key leverage point for timely goals-of-care communication and earlier, diagnosis-agnostic palliative integration.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12904-026-02026-y.
Keywords: Aggressiveness end-of-life care, Palliative care, Intensive care unit, Cancer, Noncancer disease
Introduction
Approximately 23% of deaths in China occur in hospitals, and this proportion is expected to increase significantly due to population aging and dying patients’ growing preference for hospital-based end-of-life (EOL) care [1, 2].
Near the end of life, many patients experience escalation to high-intensity, life-sustaining interventions, including intensive care unit (ICU) admission, endotracheal intubation and mechanical ventilation, cardiopulmonary resuscitation, circulatory support, and renal replacement therapy—often described collectively as aggressive EOL care [3, 4]. While such treatments can be lifesaving for selected patients, when applied in the setting of irreversible advanced illness they may prolong the dying process, compromise the quality of death, and increase caregiver psychological distress and financial burden [5, 6].
A persistent concern in EOL care is the disparity between patients dying from cancer and those dying from non-cancer conditions. Compared with cancer, many non-cancer serious illnesses follow less predictable trajectories with episodic exacerbations and greater prognostic uncertainty, complicating recognition of the terminal phase and timely advance care planning [7]. International evidence suggests that, relative to cancer decedents, non-cancer decedents are more likely to experience ICU admission and life-sustaining treatments near the end of life and are less likely to receive palliative care [8–10].
In mainland China, palliative care capacity and integration remain uneven, and available evidence has historically focused on advanced cancer populations, with comparatively limited attention to non-cancer diseases and hospital integration [11]. Contemporary, hospital-based epidemiological data describing multi-domain aggressive EOL care intensity and its timing among in-hospital decedents—and how these patterns differ between cancer and non-cancer patients—remain scarce.
To address this gap, we conducted a retrospective study of adult in-hospital decedents to compare the intensity and timing of aggressive EOL care during the terminal hospitalization between cancer and non-cancer patients. We also identified factors associated with higher treatment intensity and assessed whether temporal trends differed by cancer status.
Methods
Study design and setting
We conducted a retrospective observational study using electronic medical records from a large tertiary teaching hospital in Nanchang, Jiangxi Province, China, from August 1, 2017, to December 31, 2024. The hospital serves as a regional referral center providing comprehensive medical and surgical care. It operates two campuses with more than 4,000 open beds and, in 2024, delivered approximately 3 million outpatient visits and 195,600 inpatient admissions, reflecting a high-volume, high-acuity case mix. Clinical data were routinely collected and managed in accordance with national healthcare regulations and institutional policies.
Study population
We included adult (aged ≥ 18 years) in-hospital decedents with complete diagnostic records during the study period. Patients admitted for acute conditions unrelated to underlying medical illness (e.g., traffic accidents or pesticide poisoning) were excluded (Fig. 1). The study was approved by the Medical Ethics Committee of Nanchang University (NCUREC202504004), with a waiver of informed consent because only anonymized routinely collected data were used. This study is reported in accordance with the STROBE statement.
Fig. 1.
Flow chart of patient selection
Outcomes
The primary outcome was the core aggressive EOL care (range 0–5), defined as the unweighted sum of five prespecified aggressive EOL care domains during the terminal hospitalization: CPR-related interventions, respiratory support, renal support, circulatory support, and ICU care. Each domain was coded as present (1) if any corresponding intervention occurred at least once and absent (0) otherwise, and the score was calculated as the total number of domains received.
Secondary outcomes were high-intensity care defined a priori as a core score ≥ 2 and ≥ 3, representing increasing levels of aggressive EOL care treatment.
Variables and definitions
Cancer status was defined as the presence of a malignant neoplasm diagnosis (ICD-10: C00–C97) recorded in any diagnosis field during the terminal hospitalization, reflecting documented cancer diagnosis during the index admission rather than an adjudicated cause of death. Comorbidity burden was assessed using the Charlson comorbidity index (CCI) derived from ICD-10 codes using a validated algorithm. Emergency admission was defined as admission through the emergency department or urgent transfer at presentation. Calendar year of death was extracted from the date of death and modeled as a categorical variable. Palliative care was identified using ICD-10 code Z51.5 recorded in any diagnosis field during the terminal hospitalization.
Five aggressive EOL care domains were prespecified and coded as present if any corresponding intervention occurred at least once during the terminal hospitalization: CPR-related interventions (cardiopulmonary resuscitation and/or defibrillation); respiratory support (tracheostomy, endotracheal intubation, invasive mechanical ventilation [IMV], and/or non-invasive ventilation [NIV]); renal support (continuous renal replacement therapy [CRRT], intermittent renal replacement therapy [RRT], and/or venous catheterization for dialysis); circulatory support (extracorporeal membrane oxygenation [ECMO] and/or intra-aortic balloon pump [IABP]); and ICU care (death occurring in the ICU).
Among patients receiving a given intervention, we calculated the interval between the last recorded occurrence of that intervention and death and summarized whether the last occurrence was on the day of death and within 7 days before death.
Statistics analysis
Baseline characteristics were summarized overall and by cancer status. Categorical variables are presented as counts (percentages), and continuous variables as medians with interquartile ranges (IQR).
Because data were available only from August to December 2017, we conducted a prespecified sensitivity analysis excluding 2017 admissions and refitted all models to assess robustness to potentially incomplete capture in 2017.
For the primary outcome (core score), we fitted multivariable Poisson regression models with robust (sandwich) standard errors to provide valid inference under potential overdispersion and mild model misspecification. Effect estimates are reported as incidence rate ratios (IRRs) with 95% confidence intervals (CIs). We conducted two sensitivity analyses: (1) excluding decedents with terminal LOS ≤ 1 day to assess the influence of extremely short terminal hospitalizations; and (2) recalculating the core score excluding the death in ICU domain (range 0–4) and refitting the primary Poisson model.
For secondary binary outcomes (core score ≥ 2 and ≥ 3), we fitted multivariable logistic regression models adjusting for cancer diagnosis, age group, sex, CCI, emergency admission, and calendar year of death (categorical; 2017 as the reference).
Because palliative care was rare and demonstrated complete separation (no high-intensity events among patients coded as receiving palliative care), its association with high-intensity care was estimated using Firth’s penalized logistic regression with profile penalized-likelihood 95% confidence intervals.
To evaluate temporal trends and whether they differed by cancer status, we fitted generalized linear models with calendar year of death treated as a categorical variable, adjusted for age group, sex, CCI, emergency admission, and palliative care. We included cancer status and a cancer-by-year interaction term; effect modification was assessed using Wald tests comparing models with and without the interaction term. Adjusted annual estimates and 95% CIs were derived using marginal standardization. Statistical analyses were performed in R version 4.3.3. All tests were two-sided with a significance level of 0.05.
Results
Among 3,205 in-hospital decedents, 990 (30.9%) had a cancer diagnosis. Compared with non-cancer decedents, cancer decedents were younger, with a lower proportion aged ≥ 80 years, but had a higher CCI. Length of stay was longer among cancer decedents, whereas total hospitalization costs were higher among non-cancer decedents. Palliative care was substantially more common in cancer than non-cancer decedents (9.0% vs. 0.3%; Table 1).
Table 1.
Baseline characteristics of in-hospital decedents, overall and by cancer status a
| Variables | Overall (n = 3205) | Cancer(n = 990) | Non-Cancer (n = 2215) | P value |
|---|---|---|---|---|
| Age, year (IQR) | 70.0(57.0–81.0) | 69.0(59.0–77.0) | 72.0(56.0–83.0) | < 0.001 |
| < 60 | 909(28.4) | 261(26.4) | 648(29.3) | |
| 60–69 | 620(19.3) | 266(26.9) | 354(16.0) | |
| 70–79 | 725(22.6) | 257(26.0) | 468(21.1) | |
| ≥ 80 | 951(29.7) | 206(20.8) | 745(33.6) | |
| Male | 2173(67.8) | 672(67.9) | 1501(67.8) | 0.95 |
| Charlson comorbidity index (IQR) | 6.0(4.0–7.0) | 7.0(5.0–9.0) | 5.0(3.0–6.0) | < 0.001 |
| Total Cost |
5296.5 (2251.1–11576.3) |
4632.5 (2165.8–8546.8) |
5732.0 (2292.4–13409.4) |
< 0.001 |
| LOS, days (IQR) | 8.0(3.018.0) | 12.0(6.0–22.0) | 6.0(2.0–15.0) | < 0.001 |
| Palliative | 96(3.0) | 89(9.0) | 7(0.3) | < 0.001 |
| Insurance | < 0.001 | |||
| UEBMI | 1555(48.5) | 563(56.9) | 992(44.8) | |
| URBMI | 479(14.9) | 145(14.6) | 334(15.1) | |
| Other issuance | 402(12.5) | 144(14.5) | 258(11.6) | |
| Full self-pay | 769(24.0) | 138(13.9) | 631(28.5) | |
| Year of death | < 0.001 | |||
| 2017 | 122(3.8) | 39(3.9) | 83(3.7) | |
| 2018 | 329(10.3) | 112(11.3) | 217(9.8) | |
| 2019 | 295(9.2) | 99(10.0) | 196(8.8) | |
| 2020 | 433(13.5) | 142(14.3) | 291(13.1) | |
| 2021 | 515(16.1) | 191(19.3) | 324(14.6) | |
| 2022 | 685(21.4) | 206(20.8) | 479(21.6) | |
| 2023 | 381(11.9) | 72(7.3) | 309(14.0) | |
| 2024 | 445(13.9) | 129(13.0) | 316(14.3) |
LOS Length of Stay, UEBMI Urban Employee Basic Medical Insurance, URBMI Urban Resident Basic Medical Insurance, IQR Interquartile Range
a Data are presented as number (percentage) of patients unless otherwise indicated
Aggressive EOL care during the terminal hospitalization were common and differed markedly by cancer status (Table 2). Non-cancer decedents consistently had higher use of CPR-related interventions and respiratory, renal, and circulatory support than cancer decedents, and most events clustered within the last 7 days of life. For example, respiratory support occurred in 65.8% of non-cancer versus 25.2% of cancer decedents (within ≤ 7 days: 50.7% vs. 19.7%), and CPR-related interventions occurred in 25.2% versus 12.1% (within ≤ 7 days: 24.7% vs. 11.9%). Death in ICU was nearly threefold more frequent among non-cancer than cancer decedents (46.5% vs. 15.6%). Correspondingly, high core aggressive EOL care intensity scores was concentrated among non-cancer decedents, with 22.9% receiving ≥ 3 core domains compared with 6.7% of cancer decedents.
Table 2.
Aggressive end-of-life care interventions during hospitalization and within ≤ 7 days before death, overall and by cancer status a
| Indicator | Overall occurrence (n = 3205) (%) | Cancer (n = 990) (%) | Non-cancer (n = 2215) (%) | |||
|---|---|---|---|---|---|---|
| Any during hospitalization | Within ≤ 7 days before death | Any during hospitalization | Within ≤ 7 days before death | Any during hospitalization | Within ≤ 7 days before death | |
| CPR-related interventions (any) | 678 (21.2) | 666 (20.8) | 120 (12.1) | 118 (11.9) | 558 (25.2) | 548 (24.7) |
| CPR | 650(20.3) | 639 (19.9) | 116(11.7) | 115 (11.6) | 534(24.1) | 524 (23.7) |
| Defibrillation | 137(4.3) | 135 (4.2) | 13(1.3) | 12 (1.2) | 124(5.6) | 123 (5.6) |
| Respiratory support (any) | 1707 (53.3) | 1318 (41.1) | 249 (25.2) | 195 (19.7) | 1458 (65.8) | 1123 (50.7) |
| Tracheostomy | 119(3.7) | 50 (1.6) | 16(1.6) | 9 (0.9) | 103(4.7) | 41 (1.9) |
| Intubation | 1054(32.9) | 848 (26.5) | 164(16.6) | 134 (13.5) | 890(40.2) | 714 (32.2) |
| IMV | 1309(40.8) | 958 (29.9) | 166(16.8) | 125 (12.6) | 1143(51.6) | 833 (37.6) |
| NIV | 429 (13.4) | 278 (8.7) | 77 (7.8) | 45 (4.5) | 352 (15.9) | 233 (10.5) |
| Renal support (any) | 425 (13.3) | 309 (9.6) | 59 (6.0) | 43 (4.3) | 366 (16.5) | 266 (12.0) |
| CRRT | 312 (9.7) | 226 (7.1) | 42 (4.2) | 31 (3.1) | 270 (12.2) | 195 (8.8) |
| RRT | 116 (3.6) | 65 (2.0) | 14 (1.4) | 7 (0.7) | 102 (4.6) | 58 (2.6) |
| Venous Catheterization for Dialysis | 189 (5.9) | 128 (4.0) | 31 (3.1) | 20 (2.0) | 158 (7.1) | 108 (4.9) |
| Circulatory support (any) | 134 (4.2) | 106 (3.3) | 4 (0.4) | 3 (0.3) | 130 (5.9) | 103 (4.7) |
| ECMO | 84 (2.6) | 67 (2.1) | 3 (0.3) | 2 (0.2) | 81 (3.7) | 65 (2.9) |
| IABP | 66 (2.1) | 47 (1.5) | 2 (0.2) | 2 (0.2) | 64 (2.9) | 45 (2.0) |
| Care setting | ||||||
| Death in ICU | 1185 (37.0) | 154 (15.6) | 1031 (46.5) | |||
| Core aggressive EOL intensity | ||||||
| 0 | 1120 (34.9) | 656 (66.3) | 464 (20.9) | |||
| 1 | 732 (22.8) | 156 (15.8) | 576 (26.0) | |||
| 2 | 778 (24.3) | 111 (11.2) | 667 (30.1) | |||
| ≥ 3 | 575 (17.9) | 67 (6.7) | 508 (22.9) | |||
CPR Cardiopulmonary Resuscitation, IMV Invasive Mechanical Ventilation, NIV Non-invasive Mechanical Ventilation, ECMO Extracorporeal Membrane Oxygenation, IABP Intra-Aortic Balloon Pump, CRRT Continuous Renal Replacement Therapy, RRT Renal Replacement Therapy, ICU Intensive Care Unit
a Data are presented as number (percentage) of patients
Among recipients, the most proximal occurrence of each intervention was typically close to death (Table 3). The most proximal CPR-related intervention occurred almost entirely within the last week (≤ 7 days: 98.2%), with most occurring on the day of death (80.4%). Respiratory support was similarly end-of-life proximal (≤ 7 days: 77.2%; intubation 80.5%; IMV 73.2%), whereas renal and circulatory supports were less often last administered on the day of death (day 0: 10.8% and 20.9%). Timing patterns were broadly similar between cancer and non-cancer recipients, with imprecise estimates for rare interventions in the cancer group.
Table 3.
Timing of aggressive end-of-life care interventions relative to death among recipients, overall and by cancer status a
| Indicator (denominator) | Overall (%) | Cancer (%) | Non-cancer (%) | ||||||
|---|---|---|---|---|---|---|---|---|---|
| day 0 | ≤ 3 days | ≤ 7 days | day 0 | ≤ 3 days | ≤ 7 days | day 0 | ≤ 3 days | ≤ 7 days | |
| CPR events domain (n = 678; cancer n = 120; non-cancer n = 558) | 80.4 | 95.6 | 98.2 | 77.5 | 97.5 | 98.3 | 81 | 95.2 | 98.2 |
| CPR (n = 650; cancer n = 116; non-cancer n = 534) | 80.6 | 95.5 | 98.3 | 77.6 | 98.3 | 99.1 | 81.3 | 94.9 | 98.1 |
| Defibrillation (n = 137; cancer n = 13; non-cancer n = 124) | 75.9 | 92.7 | 98.5 | 84.6 | 84.6 | 92.3 | 75.0 | 93.5 | 99.2 |
| Respiratory support domain (n = 1707; cancer n = 249; non-cancer n = 1458) | 25.5 | 54 | 77.2 | 26.1 | 55 | 78.3 | 25.4 | 53.8 | 77 |
| Intubation (n = 1054; cancer n = 164; non-cancer n = 890) | 32.6 | 61.1 | 80.5 | 29.3 | 59.8 | 81.7 | 33.3 | 61.3 | 80.2 |
| Tracheostomy (n = 119; cancer n = 16; non-cancer n = 103) | 5.0 | 16.0 | 42 | 18.8 | 37.5 | 56.2 | 2.9 | 12.6 | 39.8 |
| IMV (n = 1309; cancer n = 166; non-cancer n = 1143) | 21.5 | 49.1 | 73.2 | 21.1 | 48.2 | 75.3 | 21.5 | 49.3 | 72.9 |
| NIV (n = 429; cancer n = 77; non-cancer n = 352) | 11.2 | 36.1 | 64.8 | 11.7 | 36.4 | 58.4 | 11.1 | 36.1 | 66.2 |
| Renal support domain (n = 425; cancer n = 59; non-cancer n = 366) | 10.8 | 49.4 | 72.7 | 13.6 | 44.1 | 72.9 | 10.4 | 50.3 | 72.7 |
| RRT (n = 116; cancer n = 14; non-cancer n = 102) | 7.8 | 37.9 | 56 | 7.1 | 7.1 | 50 | 7.8 | 42.2 | 56.9 |
| CRRT (n = 312; cancer n = 42; non-cancer n = 270) | 8.7 | 45.8 | 72.4 | 7.1 | 40.5 | 73.8 | 8.9 | 46.7 | 72.2 |
| Venous Catheterization for Dialysis (n = 189; cancer n = 31; non-cancer n = 158) | 11.1 | 47.1 | 67.7 | 16.1 | 48.4 | 64.5 | 10.1 | 46.8 | 68.4 |
| Circulatory support domain (n = 134; cancer n = 4; non-cancer n = 130) | 20.9 | 61.9 | 79.1 | 0 | 50 | 75 | 21.5 | 62.3 | 79.2 |
| ECMO (n = 84; cancer n = 3; non-cancer n = 81) | 17.90 | 64.30 | 79.8 | 0.0 | 66.7 | 66.7 | 18.5 | 64.2 | 80.2 |
| IABP (n = 66; cancer n = 2; non-cancer n = 64) | 22.7 | 48.5 | 71.2 | 0.0 | 0.0 | 100 | 23.4 | 50.0 | 70.3 |
| Death in ICU (n = 1185; cancer n = 154; non-cancer n = 1031) | 100 | 100 | 100 | ||||||
a Data are cumulative percentages (%) among recipients (denominators shown in the indicator column)
Day 0 denotes the day of death. The ≤ 3 day and ≤ 7 day columns indicate the cumulative proportion whose most proximal occurrence was on Day 0 or within the preceding 3 or 7 days, respectively. Timing was defined using the most proximal occurrence (closest to death); for domains, timing was the minimum time-to-death across component procedures
In multivariable Poisson regression with robust standard errors, non-cancer status was associated with a higher expected core aggressive EOL care intensity score (adjusted IRR 2.22, 95% CI 1.96–2.44; p < 0.001). Emergency admission was strongly associated with higher intensity (IRR 1.75, 95% CI 1.66–1.85; p < 0.001). Male sex was associated with modestly higher intensity (IRR 1.11, 95% CI 1.05–1.19; p < 0.001), whereas higher CCI was associated with slightly lower intensity (IRR 0.97 per point, 95% CI 0.95–0.99; p = 0.012). Palliative care was associated with markedly lower intensity (IRR 0.14, 95% CI 0.08–0.26; p < 0.001), while age group was not independently associated with the score. Calendar year was significantly associated with intensity; compared with 2017, years 2019–2024 were consistently associated with higher intensity. Findings were similar after excluding 2017 admissions (Table 4). Sensitivity analyses excluding decedents with terminal LOS ≤ 1 day (Supplementary Table S2) and excluding the ICU-death domain from the composite score (Supplementary Table S3) yielded materially similar estimates.
Table 4.
Factors associated with the core aggressive EOL care intensity score (0–5): multivariable Poisson regression with robust standard errors
| Variable | Adjusted IRR (95% CI) | P value | Sensitivity analysis (excluding 2017) | |
|---|---|---|---|---|
| Adjusted IRR (95% CI) | P value | |||
| Non-Cancer (reference: Cancer) | 2.22 (1.96–2.44) | < 0.001 | 2.17 (1.92–2.44) | < 0.001 |
| Age group (reference: age < 60) | ||||
| 60–69 | 1.00 (0.92–1.10) | 0.93 | 0.99 (0.90–1.09) | 0.86 |
| 70–79 | 1.01 (0.92–1.12) | 0.81 | 1.01 (0.91–1.11) | 0.89 |
| ≥ 80 | 0.96 (0.86–1.07) | 0.47 | 0.95 (0.85–1.06) | 0.36 |
| Male | 1.11 (1.05–1.19) | < 0.001 | 1.12 (1.05–1.19) | < 0.001 |
| Charlson comorbidity index | 0.97 (0.95–0.99) | 0.012 | 0.97 (0.95–0.99) | 0.009 |
| Palliative care | 0.14 (0.08–0.26) | < 0.001 | 0.14 (0.08–0.26) | < 0.001 |
| Emergency admission | 1.75 (1.66–1.85) | < 0.001 | 1.73 (1.64–1.83) | < 0.001 |
| Year of death (reference:2017) | Year of death (reference:2018) | |||
| 2018 | 1.21 (0.96–1.52) | 0.11 | - | |
| 2019 | 1.65 (1.32–2.07) | < 0.001 | 1.36 (1.18–1.58) | < 0.001 |
| 2020 | 1.72 (1.39–2.13) | < 0.001 | 1.42 (1.24–1.62) | < 0.001 |
| 2021 | 2.03 (1.64–2.51) | < 0.001 | 1.68 (1.48–1.90) | < 0.001 |
| 2022 | 1.67 (1.36–2.07) | < 0.001 | 1.39 (1.22–1.57) | < 0.001 |
| 2023 | 1.60 (1.29–1.99) | < 0.001 | 1.33 (1.16–1.52) | < 0.001 |
| 2024 | 1.60 (1.29–1.98) | < 0.001 | 1.32 (1.16–1.51) | < 0.001 |
IRR incidence rate ratio
Estimates are adjusted IRRs with 95% CIs from multivariable Poisson regression with robust (sandwich) standard errors. Models included cancer status, age group, sex, Charlson comorbidity index, palliative care indicator, emergency admission, and calendar year (categorical; reference: 2017 in the primary analysis). In the sensitivity analysis excluding 2017, calendar year was re-referenced to 2018
In multivariable logistics analyses, non-cancer decedents had substantially higher odds of high-intensity care at both thresholds (≥ 2: aOR 4.55, 95% CI 3.70–5.56; p < 0.001; ≥3: aOR 3.03, 95% CI 2.27–4.17; p < 0.001). Emergency admission was strongly associated with high-intensity care (≥ 2: aOR 10.43, 95% CI 7.64–14.23; p < 0.001; ≥3: aOR 4.35, 95% CI 3.44–5.50; p < 0.001). Palliative care was inversely associated with high-intensity care (≥ 2: aOR 0.02, 95% CI < 0.01–0.14; p < 0.001; ≥3: aOR 0.06, 95% CI < 0.01–0.43). Results were materially unchanged in sensitivity analyses excluding 2017 (Supplementary Tables S1A–S1B).
Figure 2 shows adjusted temporal trends in the core aggressive EOL care intensity score by cancer status. Across the study period, non-cancer decedents consistently exhibited higher intensity than cancer decedents. Intensity increased over time in both groups, with evidence of differential patterns: non-cancer decedents increased more steeply from 2017 to 2021, whereas cancer decedents showed a more gradual increase followed by plateauing or decline. The cancer-by-year interaction was statistically significant (P for interaction = 0.047), supporting differential temporal trends between cancer and non-cancer decedents.
Fig. 2.
Adjusted temporal trends in core aggressive end-of-life care intensity by cancer status, 2017–2024
Discussion
This study provides contemporary hospital-based evidence on aggressive EOL care among adult in-hospital decedents in mainland China. We observed three central findings. First, multi-domain life-sustaining treatment use was common and differed substantially by diagnosis, with non-cancer decedents consistently receiving more intensive aggressive EOL care than cancer decedents, even after adjustment. Second, aggressive EOL care frequently occurred near death: a substantial proportion of patients received key interventions within the final week of life, and among recipients the most proximal (final recorded) occurrence typically fell close to death. Third, treatment intensity increased over time, and the temporal patterns differed by cancer status, suggesting heterogeneous trajectories across diagnostic groups.
Prior work has largely conceptualized aggressive EOL care using claims-based or registry indicators measured within fixed windows [12]. Comparative studies have also consistently shown that non-cancer decedents are more likely than cancer decedents to receive aggressive care near the end of life [8, 13, 14]. However, fewer studies describe when, during the terminal hospitalization, key life-sustaining interventions occur relative to death—particularly using procedure-level timing and recipient-based metrics that characterize the final recorded episode closest to death [5]. Against this backdrop, we extend the literature by characterizing intervention timing from two complementary perspectives: population-level occurrence within ≤ 7 days before death and, among recipients, the timing of the most proximal (final recorded) occurrence of each intervention.
Using this framework, we observed that a substantial share of patients received aggressive EOL care within the final week of life (Table 2), indicating that late-stage use is common. Among recipients (Table 3), the most proximal (final recorded) occurrence typically fell close to death—most prominently for CPR-related interventions—suggesting that many interventions are last recorded during a brief period of rapid deterioration immediately preceding death [15]. Because timing was defined by the most proximal occurrence, these estimates characterize the timing of the final documented episode and should not be interpreted as total exposure duration or frequency of earlier episodes.
A plausible explanation for the diagnosis-group difference in aggressive EOL care is the well-described contrast in EOL illness trajectories and prognostic uncertainty between advanced cancer and end-stage organ failure [7]. Compared with end-stage organ failure, advanced cancer more often follows a recognizable terminal phase, and functional decline frequently accelerates in a way that is easier to interpret as irreversible [16]. In contrast, organ-failure trajectories are typically prolonged and fluctuating with episodic exacerbations and prognostic uncertainty, which can delay recognition of the terminal phase and compress opportunities for advance care planning [17]. Consequently, patients with cancer and their families may be more prepared for goals-of-care discussions and more willing to transition toward comfort-focused priorities when time-limited life-sustaining trials fail to achieve recovery [16, 18].
Diagnosis-based disparities may be further reinforced by differences in clinician training and service integration, as well as broader sociocultural and health-system contexts in mainland China. Palliative care delivery and referral pathways have historically been more tightly linked to oncology, while patients with advanced non-cancer conditions often receive less palliative care in the last year of life [19]. At the same time, advance care planning remains uncommon and end-of-life decisions are frequently family-centered; death-related discussions may be avoided or delayed, limiting timely clarification of patient goals and preferences [20]. In acute care settings—particularly during emergency presentations—time pressure, communication barriers, and variable institutional practices may constrain limitation-of-treatment decisions and favor escalation during clinical crises [21]. Surveys also suggest that preferences for life-prolonging treatment are common in Chinese populations, which may further sustain rescue-oriented care when prognostic uncertainty is high [22]. Although national initiatives have accelerated hospice and palliative care development since 2017, service availability, workforce capacity, and integration within acute hospitals and critical care remain uneven [23]. Together, these contextual factors may contribute to persistently high Aggressive EOL care even among older or multimorbid patients; strengthening earlier, structured goals-of-care communication and timely palliative involvement may help reduce potentially non-beneficial escalation and improve goal-concordant care [6].
In our adjusted models, emergency admission and non-cancer status as the strongest factors associated with higher aggressive EOL care intensity. Emergency admission should be interpreted primarily as a marker of crisis trajectories and acute clinical instability rather than a causal driver; unmeasured physiologic severity and limited time for communication may partly explain the observed association [24]. Nonetheless, the emergency department and early hospitalization represent pragmatic opportunities to initiate timely goals-of-care discussions and consider earlier palliative care involvement [3, 25]. Even after adjustment, non-cancer decedents remained substantially more likely to receive high-intensity care, consistent with less predictable illness trajectories and greater prognostic uncertainty outside oncology. Associations for sex and CCI were comparatively modest and should be interpreted cautiously given residual confounding and incomplete measurement of baseline severity and functional status. Palliative care was uncommon—particularly among non-cancer decedents—and was inversely associated with intensity [10]. Aggressive EOL care nevertheless remained common among cancer decedents, which is clinically salient given evidence that outcomes after in-hospital CPR are poor in advanced cancer, underscoring the need for earlier goals-of-care discussions and timely palliative involvement [26].
We observed increasing treatment intensity over calendar time, with evidence that trends differed by cancer status. The overall increase may reflect expanding availability of organ-support technologies and ICU beds, evolving practice norms in acute care, and changes in case mix over time [27, 28]. The divergent trajectories by cancer status suggest that temporal changes are shaped by different care pathways: oncology services may have experienced earlier diffusion of structured goals-of-care practices, whereas non-cancer pathways—often routed through emergency and critical care—may remain more prone to default escalation under uncertainty [29]. The COVID-19 pandemic may have further influenced hospitalization pathways and the use of ICU care and organ-support therapies major waves in China, but we lacked reliable indicators of COVID-19 infection status and physiologic severity to disentangle pandemic-related effects from secular trends [30].
Our findings should be interpreted in light of several limitations. First, this single-center retrospective study of in-hospital decedents may limit generalizability to community hospitals, other regions, and decedents who die outside hospital. Second, intervention ascertainment and timing were derived from routinely collected electronic records and may be affected by incomplete capture, variable documentation practices, and imprecise timestamps. Third, timing analyses were based on the most proximal documented occurrence, which characterizes the final documented episode before death but does not reflect the full duration or frequency of earlier treatment exposure. Fourth, we lacked direct measures of patient preferences, functional status, and standardized acute physiologic severity (e.g., SOFA/NEWS) and granular disease stage, which may introduce indication bias and residual confounding in the observed associations and limit attribution of calendar-time patterns (including potential pandemic-related case-mix shifts). Finally, palliative care (Z51.5) and cancer status (C00–C97 during the terminal hospitalization) were defined from ICD-10 codes; under-ascertainment and misclassification are possible and would likely attenuate diagnosis-group differences.
Conclusion
Aggressive EOL care was common among in-hospital decedents and occurred predominantly close to death. Non-cancer decedents had higher multi-domain treatment intensity and lower documented palliative care use than cancer decedents, with differences persisting after adjustment. Emergency admission showed the strongest association with higher intensity, whereas documented palliative care was inversely associated with intensity. Using complementary timing metrics (≤ 7-day population occurrence and recipient-level most proximal episode), this study delineates when aggressive care is delivered during terminal hospitalization and supports earlier, pathway-based palliative integration to promote more equitable, goal-concordant EOL care.
Supplementary Information
Acknowledgements
Not applicable.
Authors’ contributions
Bin Wu: Methodology; Formal analysis; Investigation; Data curation; Writing—original draft. Yingping Yi: Data curation; Investigation; Resources; Writing—review & editing. Pengfei Yu: Data curation; Software; Validation; Writing—review & editing. Haowen Luo: Data curation; Resources; Visualization; Writing—review & editing. Jianmo Liu: Data curation; Investigation; Writing—review & editing. Yuhuai Peng: Supervision; Investigation (clinical guidance); Writing—review & editing. Yu Cao: Formal analysis; Methodology; Visualization; Writing—review & editing. Yongping Zhang: Validation; Visualization; Writing—review & editing. Hao Wang: Software; Validation; Writing—review & editing. Wei Gao: Conceptualization; Methodology; Supervision; Funding acquisition; Writing—review & editing.
Funding
This project was supported by grants from Senior Talent Startup Fund of Nanchang University (28170120/9167).
Data availability
Data will be available to the contact author upon reasonable request.
Declarations
Ethics approval and consent to participate
This study was conducted in accordance with the Declaration of Helsinki and in compliance with relevant local regulations. This study was approved by the Medical Ethics Committee of Nanchang University (NCUREC202504004). Owing to its retrospective design and the use of de-identified, anonymized electronic medical records, the requirement for written informed consent was waived by the committee.
Consent for publication
Not applicable (data were anonymized and not identifiable).
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Weng L, Hu Y, Sun Z, et al. Place of death and phenomenon of going home to die in Chinese adults: A prospective cohort study [J]. Lancet Reg Health West Pac. 2022;18:100301. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Zhang J, Cao Y, Su M, et al. The experiences of clinical nurses coping with patient death in the context of rising hospital deaths in China: a qualitative study [J]. BMC Palliat Care. 2022;21(1):163. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Earle CC, Park ER, Lai B, et al. Identifying potential indicators of the quality of end-of-life cancer care from administrative data [J]. J Clin Oncol. 2003;21(6):1133. [DOI] [PubMed] [Google Scholar]
- 4.Luta X, Maessen M, Egger M, et al. Measuring intensity of end of life care: a systematic review [J]. PLoS ONE. 2015;10(4):e0123764. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.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 [J]. EClinicalMedicine, 2024, 71: 102561. [DOI] [PMC free article] [PubMed]
- 6.Wright AA, Zhang B, Ray A, et al. Associations between end-of-life discussions, patient mental health, medical care near death, and caregiver bereavement adjustment [J]. JAMA. 2008;300(14):1665. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Lunney JR, Lynn J, Foley DJ, et al. Patterns of functional decline at the end of life [J]. JAMA. 2003;289(18):2387. [DOI] [PubMed] [Google Scholar]
- 8.Lyngaa T, Christiansen CF, Nielsen H, et al. Intensive care at the end of life in patients dying due to non-cancer chronic diseases versus cancer: a nationwide study in Denmark [J]. Crit Care. 2015;19:413. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Ko YT, Ko MC, Huang CM, et al. Trends of Utilization of Palliative Care and Aggressive End-of-Life Care for Patients Who Died of Cancers and Those Who Died of Noncancer Diseases in Hospitals [J]. J Pain Symptom Manage. 2020;60(6):1136. [DOI] [PubMed] [Google Scholar]
- 10.Quinn KL, Wegier P, Stukel TA, et al. Comparison of Palliative Care Delivery in the Last Year of Life Between Adults With Terminal Noncancer Illness or Cancer [J]. JAMA Netw Open. 2021;4(3):e210677. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Chung H, Harding R, Guo P. Palliative Care in the Greater China Region: A Systematic Review of Needs, Models, and Outcomes [J]. J Pain Symptom Manage. 2021;61(3):585. [DOI] [PubMed] [Google Scholar]
- 12.Earle CC, Neville BA, Landrum MB, et al. Evaluating claims-based indicators of the intensity of end-of-life cancer care [J]. Int J Qual Health Care. 2005;17(6):505. [DOI] [PubMed] [Google Scholar]
- 13.Wang CL, Hsu CY, Lin CY. Hospice Use and Aggressive End-of-Life Care Among Cancer and Non-Cancer Decedents in Taiwan [J]. J Pain Symptom Manage. 2026;71(1):144. [DOI] [PubMed] [Google Scholar]
- 14.Au DH, Udris EM, Fihn SD, et al. Differences in health care utilization at the end of life among patients with chronic obstructive pulmonary disease and patients with lung cancer [J]. Arch Intern Med. 2006;166(3):326. [DOI] [PubMed] [Google Scholar]
- 15.Spångfors M, Molt M, Samuelson K. In-hospital cardiac arrest and preceding National Early Warning Score (NEWS): A retrospective case-control study [J]. Clin Med (Lond). 2020;20(1):55. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Weissman DE. Decision making at a time of crisis near the end of life [J]. JAMA. 2004;292(14):1738. [DOI] [PubMed] [Google Scholar]
- 17.Murray SA, Kendall M, Boyd K, et al. Illness trajectories and palliative care [J]. BMJ. 2005;330(7498):1007. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Murray SA, Boyd K, Kendall M, et al. Dying of lung cancer or cardiac failure: prospective qualitative interview study of patients and their carers in the community [J]. BMJ. 2002;325(7370):929. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Rosenwax L, Spilsbury K, McNamara BA, et al. A retrospective population based cohort study of access to specialist palliative care in the last year of life: who is still missing out a decade on? [J]. BMC Palliat Care. 2016;15:46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Ding CQ, Zhang YP, Wang YW, et al. Death and do-not-resuscitate order in the emergency department: A single-center three-year retrospective study in the Chinese mainland [J]. World J Emerg Med. 2020;11(4):231. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Pei Y, Qi X, Zhu Z, et al. The informal discussion of advance care planning among Chinese older adults: Do education and social media use matter? [J]. Geriatr Nurs. 2024;55:1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Liao J, Wu B, Mao J et al. Preference for Aggressive End-of-Life Care among Advanced Cancer Patients in Wuhan, China: A Cross-Sectional Study [J]. Int J Environ Res Public Health, 2020, 17(18). [DOI] [PMC free article] [PubMed]
- 23.Lu Y, Yu W, Zhang J, et al. Advancements in hospice and palliative care in China: A five-year review [J]. Asia Pac J Oncol Nurs. 2024;11(3):100385. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Kim JS, Lee SY, Lee MS, et al. Aggressiveness of care in the last days of life in the emergency department of a tertiary hospital in Korea [J]. BMC Palliat Care. 2022;21(1):105. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Smith AK, McCarthy E, Weber E, et al. Half of older Americans seen in emergency department in last month of life; most admitted to hospital, and many die there [J]. Health Aff (Millwood). 2012;31(6):1277. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Reisfield GM, Wallace SK, Munsell MF, et al. Survival in cancer patients undergoing in-hospital cardiopulmonary resuscitation: a meta-analysis [J]. Resuscitation. 2006;71(2):152. [DOI] [PubMed] [Google Scholar]
- 27.Yuan L, Xu S, Xu J, et al. Correction: Variation in intensive care unit beds capacity in China from 2007 to 2021 [J]. Intensive Care Med. 2024;50(4):616. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Jain S, Long JB, Rao V, et al. Trends in use of intensive care during hospitalizations at the end-of-life among older adults with advanced cancer [J]. J Am Geriatr Soc. 2024;72(12):3840. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Karim S, Levine O, Simon J. The Serious Illness Care Program in Oncology: Evidence, Real-World Implementation and Ongoing Barriers [J]. Curr Oncol. 2022;29(3):1527. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Tu WJ, Liu Y, Zeng X. Critical care capacity during the omicron wave of the COVID-19 pandemic in China: far from enough [J]. Lancet Reg Health West Pac. 2023;31:100689. [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
Data Availability Statement
Data will be available to the contact author upon reasonable request.