Associations between gastrointestinal symptoms and coexisting symptoms amongst patients in a palliative care unit: a PCOC-based longitudinal study

Abstract

Background

Gastrointestinal (GI) symptoms are both common and distressing in patients receiving palliative care, however their relationships with coexisting symptoms and clinical factors, including functional status, clinician-rated problem severity, and palliative care phase, remain unclear in hospice ward settings. The Palliative Care Outcomes Collaboration (PCOC) framework enables for standardized daily symptom assessment and longitudinal evaluation. This study aimed to longitudinally investigate the associations between GI symptoms and related symptoms and clinical domains among patients admitted to a hospice ward.

Methods

This retrospective cohort study included 1,008 unique adult patients admitted to the hospice ward of a tertiary medical center in central Taiwan between March 1, 2021 and September 30, 2023. Each patient contributed one hospice ward admission course and was followed from admission until death or transfer to home hospice care. Routine daily PCOC assessments collected throughout the admission course were analyzed longitudinally as repeated measurements within patients. Patients younger than 18 years, those with missing identification information, or those with incomplete PCOC data were excluded. Separate generalized estimating equation (GEE) models were constructed for the appetite, nausea, and bowel domains of the Symptom Assessment Scale (SAS), adjusting for demographic, clinical, functional, and symptom-related variables. Subgroup analyses were performed for cancer and non-cancer patients.

Results

Amongst the 1,008 patients, the mean age was 68.0 years, with 59.1% being male, and 86.5% having cancer. At admission, mean SAS scores were 2.45 for appetite loss, 1.91 for bowel discomfort, and 0.69 for nausea, indicating that appetite loss was the most prominent GI symptom. Nausea was significantly associated with sleep disturbance, appetite loss, bowel discomfort, and pain, while older age was associated with lower nausea severity. Appetite loss was associated with sleep disturbance, nausea, bowel discomfort, fatigue, pain, greater clinician-rated “other” and family problems, as well as deteriorating phase; referral from the general ward was associated with lower appetite distress. Bowel discomfort was associated with appetite loss, nausea, breathlessness, fatigue, pain, higher clinician-rated “other” problems, and higher Australia-modified Karnofsky Performance Status (AKPS). Subgroup analyses showed that symptom associations were generally more extensive in cancer patients than in non-cancer patients, particularly for appetite loss and bowel discomfort.

Conclusion

GI symptoms in hospice ward patients are closely linked to coexisting symptoms, functional status, clinician-rated problem severity, and palliative care phase, and should not be assessed in isolation. Daily standardized assessment using the PCOC framework may help support more individualized symptom management for palliative care patients.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12904-026-02242-6.

Background

With more than 21% of its population projected to be age 65 years or older by the end of 2026, Taiwan is expected to become a super-aged society [1]. Accordingly, issues related to terminal illness and palliative care have become increasingly important in healthcare. Patients with terminal illness are often diagnosed with complex underlying diseases and diverse symptoms involving multiple organ systems, which significantly affect their quality of life. These symptoms are often interrelated and may interact with one another, making palliative care both challenging and essential [2]. A better understanding of the relationships amongst symptoms may help relieve symptom burden and improve quality of life for this particular patient population.

Amongst the various symptoms experienced by patients with terminal illness, gastrointestinal (GI) symptoms are particularly common and distressing. These symptoms often coexist and may interact with one another, underscoring the need for comprehensive evaluation and management [3].

Nausea and vomiting are frequently encountered in patients with terminal illness, and even mild nausea may cause substantial distress [4]. In patients diagnosed with terminal cancer, nausea and vomiting may result from both the malignant process itself and its treatment. Various therapeutic approaches, including careful evaluation as well as pharmacological and nonpharmacological interventions, have been used to alleviate these symptoms [5]. However, their management remains challenging due to the underlying mechanisms being multifactorial, with the evidence supporting current treatment strategies still being limited [6].

Loss of appetite is also a major concern for those patients in palliative care. Severe appetite loss has been associated with a higher burden of concomitant symptoms and poorer performance status, although the condition may improve significantly during hospitalization [7]. Previous studies have shown that poor appetite in patients receiving palliative care is associated with shorter survival times, while improvement in appetite may be linked to better outcomes [8]. These findings highlight the importance of adequate nutritional assessment and support in palliative care whenever feasible.

Furthermore, bowel discomfort, including constipation, may worsen with disease progression in patients receiving palliative care services [9]. Uncontrolled constipation may lead to serious complications, such as bowel distension and even bowel perforation, which may in turn increase mortality [10]. In addition, opioids are frequently prescribed in palliative care settings and may increase the risk of opioid-induced constipation [11]. However, bowel discomfort in palliative care is not limited to constipation alone. Overuse of laxatives may lead to diarrhea. In severe cases, diarrhea may cause complications such as dehydration and electrolyte imbalance [12]. Therefore, early recognition and appropriate intervention remain essential towards preventing further complications.

Importantly, GI symptoms may coexist with other distressing symptoms, such as dyspnea, fatigue and delirium, suggesting that they should be interpreted within broader symptom patterns rather than as isolated complaints [13]. Previous studies have examined symptom clusters in patients with cancer receiving palliative care and identified several distinct patterns of symptom coexistence [14]. Researchers also found that patients with cancer experienced an average of 11 to 13 concurrent symptoms [15]. Thus, focusing on single symptoms alone may be insufficient to either adequately relieve overall symptom burden or improve quality of life.

Despite these findings, the relationships between GI symptoms and other coexisting symptoms or clinical factors in patients admitted to hospice wards remain insufficiently understood. Previous studies have investigated symptom clusters in palliative care populations [14, 15]. However, comprehensive studies focusing specifically on GI symptoms and their associated risk factors are still lacking. The PCOC provides a standardized daily assessment of symptom burden, functional status, and palliative care phase, allowing repeated evaluation of symptom patterns over time [16]. In Taiwan, the PCOC has been implemented in palliative care units and shown to be significantly helpful in helping to maintain a high quality of care for those patients with terminal illness [17]. In the present study, appetite problems, nausea, and bowel problems were modeled separately because these GI symptom domains may have different clinical patterns and associations with coexisting symptoms and patient status. Therefore, this retrospective cohort study aimed to longitudinally investigate the associations between GI symptoms and coexisting symptoms, functional status, functional dependency, clinician-rated problem severity, palliative care phase, and other demographic and clinical characteristics among patients admitted to a hospice ward during a defined study period.

Methods

Study design

This study was designed as a retrospective cohort study of adult patients admitted to the hospice ward of a palliative care unit during a defined calendar period. Patients admitted to the palliative care unit between March 1, 2021 and September 30, 2023 were followed from the date of hospice ward admission until death or transfer to home hospice care. Each patient contributed one admission course, and routine daily PCOC assessments collected during the admission course were analyzed longitudinally as repeated measurements within patients. This study examined factors associated with gastrointestinal symptom severity among inpatients receiving palliative care. The study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of the medical center (IRB No.: CE22436A-1). Due to the retrospective nature of the study and use of de-identified data, informed consent was waived.

Participants

The study was conducted in the hospice ward of a palliative care unit at a tertiary medical center in central Taiwan. In our institution, this ward provides inpatient end-of-life care for patients with advanced terminal illnesses, with care goals focused on symptom relief, psychosocial and family support, and transition planning when appropriate. Eligible participants were adult inpatients who met the criteria for hospice palliative care under Taiwan’s Hospice Palliative Care Act. Terminal status was confirmed by two physicians prior to admission; therefore, all patients admitted to this ward were considered hospice patients.

Patients were excluded if they were younger than 18 years of age, had missing personal identification information, or had incomplete PCOC assessment data (Fig. 1). Incomplete PCOC data were defined as missing values in one or more required PCOC variables used in the final GEE models. Given the use of generalized estimating equations (GEE) to account for repeated measures within individuals [18], the adequacy of the sample size was considered in relation to the number of patients. In this study, a total of 1,008 unique patients were included, which was expected to provide stable and reliable parameter estimation in GEE models. No minimum number of daily PCOC assessments was required for inclusion. Given the end-of-life nature of the hospice ward population, requiring a minimum follow-up duration or minimum number of assessments could have excluded patients with short admissions or early death and introduced selection bias.

Fig. 1.

Flowchart of participant selection. * Incomplete PCOC data were defined as missing values in one or more required PCOC variables used in the final GEE models

Data collection and instruments

Routine PCOC assessments were conducted once daily at approximately 9:00 AM during each hospice ward admission course when patients remained in the ward and assessment data were available. Therefore, the repeated measurements analyzed in the GEE models referred to repeated once-daily assessments across hospitalization days, rather than multiple assessments within the same day. Length of hospice ward stay was reported in Table 1 to provide context regarding the observation period for each patient. The assessments conducted using the PCOC framework involved symptom measures (SAS), clinician-rated problem severity scores (PCPSS), functional assessments (AKPS, RUG-ADL), and palliative care phase classifications.

Table 1.

Characteristics of the patients (n = 1,008)

	Mean ± SD / N (%)
Age (years)	68.00 ± 14.70
Gender
Male	596 (59.13)
Female	412 (40.87)
Cancer
Non-cancer	136 (13.49)
Cancer	872 (86.51)
Without Metastasis	233 (26.72)
Metastasis	639 (73.28)
Referral type
Outpatient department	73 (7.24)
Emergency department	130 (12.9)
General ward	746 (74.01)
Home care service	59 (5.85)
Length of stay in palliative ward / observation period (days)	6.56 ± 7.16
In-hospital mortality	735 (72.92)
Symptom assessment scale
Sleep	2.55 ± 2.42
Appetite	2.45 ± 2.47
Nausea	0.69 ± 1.58
Bowel	1.91 ± 2.36
Breathlessness	2.99 ± 2.79
Fatigue	3.52 ± 2.89
Pain	2.60 ± 2.60
Palliative care problem severity score
Pain	0.89 ± 0.74
Other#	1.44 ± 0.65
Psychological	1.05 ± 0.58
Family	1.33 ± 0.59
Australia-modified Karnofsky performance status	32.08 ± 12.22
Resource utilization groups-activities of daily living
Mobility	3.98 ± 0.88
Toileting	4.10 ± 0.73
Transfer	4.14 ± 0.82
Eating	2.64 ± 0.58
Palliative care phase
1 (stable phase)	109 (10.81)
2 (unstable phase)	457 (45.34)
3 (deteriorating phase)	345 (34.23)
4 (terminal phase)	97 (9.62)

Abbreviation: SD Standard Deviation

#PCPSS Other: Other symptoms include nausea, vomiting, bowel problems, dyspnea, and other distressing symptoms not classified in the pain or psychological categories

The Symptom Assessment Scale (SAS) was used to evaluate patient-reported symptom distress. The scale is a patient-reported instrument that uses a 0–10 numerical rating scale, where 0 indicates no symptom distress and 10 indicates the worst possible distress [19]. The scale includes eight symptom domains: pain, sleep, nausea, bowel problems, appetite problems, breathing difficulties, fatigue, and other symptoms. All SAS symptom domains were assessed as part of the same daily PCOC assessment; therefore, assessment frequency did not differ across symptom domains.

The Palliative Care Problem Severity Score (PCPSS) is a clinician-rated instrument used to assess the severity of key palliative care problems. It evaluates four domains—pain, other physical symptoms, psychological/spiritual problems, and family/carer problems—using a 4-point scale ranging from 0 (absent) to 3 (severe) [20].

Functional status was assessed using the Australia-modified Karnofsky Performance Status (AKPS), a 0–100 performance scale divided into 10-point increments, where higher scores indicate better functional status [21]. In addition to global performance status, functional dependency was assessed using the Resource Utilization Groups–Activities of Daily Living (RUG-ADL) scale, which evaluates four domains—bed mobility, toileting, transfer and eating. Here, the total score ranges from 4 to 18, with higher scores indicating greater dependency. The RUG-ADL is incorporated within the PCOC framework and derived from the Resource Utilization Groups, Version III (RUG-III) classification system [22].

In addition, clinical trajectory was classified using the Palliative Care Phase system, which categorizes inpatient patients into four phases of care: stable (phase 1), unstable (phase 2), deteriorating (phase 3), and terminal (phase 4). Each patient’s phase is determined by the interdisciplinary clinical team based upon changes in their care needs and clinical status and is routinely assessed within the PCOC framework [23].

All assessments were conducted within the standardized PCOC framework. The PCOC instruments were originally developed and psychometrically validated in Australia [24]. Subsequent validation studies in other Chinese-speaking populations have demonstrated acceptable reliability and construct validity [25], supporting their applicability in the present study setting.

Statistical analysis

Baseline demographic and clinical characteristics were summarized using descriptive statistics. Continuous variables are expressed as mean with standard deviation (SD), whereas categorical variables are expressed as counts and percentages.

To evaluate factors associated with gastrointestinal symptom distress over time, GEEs with an exchangeable working correlation structure were used to account for correlations among repeated once-daily PCOC assessments collected during each patient’s hospice ward admission course; each unique patient was treated as one cluster. Separate models were constructed for the appetite, nausea, and bowel domains of the SAS. Subgroup analyses were performed for both patients with and without cancer using the same modeling approach.

Covariates were selected based upon both clinical relevance and prior literature and were consistently included across all models. These comprised age, gender, diagnostic categories (non-cancer, cancer without metastasis, and cancer with metastasis), referral source, palliative care phase, functional status (AKPS), functional dependency (RUG-ADL), clinician-rated problem severity (PCPSS), and the remaining SAS symptom domain scores. Assessment day was not included as a separate independent covariate in the final GEE models; the models estimated population-averaged associations between each GI symptom domain and coexisting symptoms or clinical characteristics across available repeated daily assessments. In subgroup analyses, the diagnostic category variable was omitted; metastatic status was retained within the cancer subgroup.

In-hospital mortality was summarized descriptively to characterize the study population, but death was not modeled as an outcome in the GEE analyses. For the final GEE analyses, only daily assessment records with complete data for the outcome and all model covariates were included; records with missing values in any required variable were excluded using complete-case analysis, and no imputation was performed. To provide context for the potential impact of missing data, we summarized the basic demographic characteristics of patients excluded because of incomplete data in Supplementary Table 1.

Results are reported as regression coefficients with corresponding standard errors, 95% confidence intervals (CIs), and p-values. A two-sided p-value < 0.05 was considered statistically significant. All analyses were performed using SAS software, version 9.4 (SAS Institute Inc., Cary, NC, USA).

Results

A total of 1,008 unique patients, each contributing one hospice ward admission course, were included in the analysis. The mean age was 68.0 years, with 59.1% of patients being male. The majority had a cancer diagnosis (86.5%), amongst whom 73.3% were diagnosed with metastatic disease. The analytic dataset classified patients by cancer status and metastatic status; however, detailed primary diagnoses for non-cancer patients and primary cancer sites among cancer patients, including gastrointestinal cancers, were not available. Most patients were referred from the general ward (74.0%), while referrals from other sources were less common. The mean length of hospice ward stay, reflecting the observation period, was 6.56 ± 7.16 days, and the in-hospital mortality rate was 72.9%. At admission, the mean SAS scores were 2.45 ± 2.47 for loss of appetite, 0.69 ± 1.58 for nausea, and 1.91 ± 2.36 for bowel symptoms. The mean AKPS score was 32.08. Regarding palliative care phase, 45.3% of patients were classified as unstable, 34.2% as deteriorating, 10.8% as stable, and 9.6% as terminal (Table 1). Based on descriptive information, the age and sex distributions of patients excluded because of incomplete data did not appear markedly different from those of the final analytic cohort (Supplementary Table 1), suggesting that selection bias related to missing data was likely limited with respect to these baseline characteristics.

Multivariable GEE analysis demonstrated significant associations between several clinical variables and daily nausea scores (Table 2). Older age was significantly associated with lower nausea severity (β = − 0.008, p < 0.01). Higher nausea scores were significantly associated with greater symptom burden in other SAS domains, including sleep disturbance (β = 0.03, p = 0.04), appetite loss (β = 0.11, p < 0.01), bowel discomfort (β = 0.08, p < 0.01), and pain (β = 0.04, p < 0.01). No significant associations were observed for gender, diagnostic category, referral source, palliative care phase, AKPS, RUG-ADL, or palliative care problem severity scores.

Table 2.

The generalized estimating equations analysis for symptom assessment scale of Nausea domain score (n = 1,008)

	Estimate	Std.err	95% CI		p-value
Age (years)	-0.008	0.003	-0.014	-0.002	< 0.01
Gender
Male	REF
Female	0.02	0.08	-0.13	0.18	0.77
Non-Cancer	REF
Cancer	-0.15	0.13	-0.42	0.11	0.25
Cancer Without Metastasis	REF
Cancer With Metastasis	0.05	0.11	-0.16	0.26	0.62
Referral type
Outpatient department	REF
Emergency department	-0.02	0.23	-0.48	0.43	0.92
General ward	-0.20	0.21	-0.61	0.21	0.34
Home care service	0.00	0.27	-0.53	0.53	1.00
Symptom assessment scale
Sleep	0.03	0.01	0.00	0.06	0.04
Appetite	0.11	0.02	0.07	0.14	< 0.01
Bowel	0.08	0.02	0.05	0.11	< 0.01
Breathlessness	0.02	0.02	-0.02	0.05	0.32
Fatigue	0.01	0.01	-0.01	0.04	0.31
Pain	0.04	0.02	0.01	0.07	< 0.01
Palliative care problem severity score
Pain	-0.02	0.04	-0.10	0.06	0.61
Other#	0.05	0.04	-0.02	0.12	0.20
Psychological	0.06	0.06	-0.06	0.18	0.32
Family	-0.05	0.06	-0.16	0.06	0.36
Australia-modified Karnofsky performance status	0.0003	0.0035	-0.0065	0.0071	0.94
Resource utilization groups-activities of daily living
Mobility	-0.10	0.07	-0.23	0.03	0.14
Toileting	0.09	0.07	-0.05	0.23	0.23
Transfer	-0.07	0.07	-0.21	0.07	0.34
Eating	-0.07	0.07	-0.20	0.06	0.32
Palliative care phase
1 (stable phase)	REF
2 (unstable phase)	0.15	0.09	-0.02	0.33	0.09
3 (deteriorating phase)	0.13	0.08	-0.03	0.29	0.12
4 (terminal phase)	0.02	0.10	-0.18	0.21	0.87

Abbreviation: SD Standard Deviation

#PCPSS Other: Other symptoms include nausea, vomiting, bowel problems, dyspnea, and other distressing symptoms not classified in the pain or psychological categories

For loss of appetite scores (Table 3), referrals from the general ward were associated with lower appetite distress when compared with outpatient referrals (β = − 0.60, p = 0.01), while referrals from the emergency department showed a similar trend that did not reach any statistical significance (β = − 0.53, p = 0.06). Higher loss of appetite scores was significantly associated with increased symptom burden in several other SAS domains, including sleep disturbance (β = 0.15, p < 0.01), nausea (β = 0.24, p < 0.01), bowel discomfort (β = 0.15, p < 0.01), fatigue (β = 0.20, p < 0.01), and pain (β = 0.07, p < 0.01). With respect to clinician-assessed problems, greater PCPSS “other” (β = 0.14, p = 0.02) and family-related concerns (β = 0.18, p = 0.02) were associated with higher appetite distress. Regarding palliative care phase, appetite scores were elevated during the deteriorating stage (phase 3) and compared with the stable stage (phase 1) (β = 0.24, p = 0.02). Other covariates were not significantly associated with appetite scores.

Table 3.

The generalized estimating equations analysis for symptom assessment scale of Appetite domain score (n = 1,008)

	Estimate	Std.err	95% CI		p-value
Age (years)	-0.002	0.004	-0.011	0.006	0.61
Gender
Male	REF
Female	0.04	0.12	-0.20	0.27	0.75
Non-Cancer	REF
Cancer	0.40	0.21	-0.02	0.82	0.06
Cancer Without Metastasis	REF
Cancer With Metastasis	-0.03	0.15	-0.32	0.26	0.85
Referral type
Outpatient department	REF
Emergency department	-0.53	0.28	-1.07	0.02	0.06
General ward	-0.60	0.24	-1.07	-0.13	0.01
Home care service	-0.48	0.34	-1.14	0.18	0.15
Symptom assessment scale
Sleep	0.15	0.02	0.11	0.20	< 0.01
Nausea	0.24	0.03	0.17	0.30	< 0.01
Bowel	0.15	0.02	0.11	0.19	< 0.01
Breathlessness	0.01	0.02	-0.03	0.05	0.53
Fatigue	0.20	0.02	0.15	0.24	< 0.01
Pain	0.07	0.03	0.02	0.12	< 0.01
Palliative care problem severity score
Pain	-0.13	0.07	-0.27	0.02	0.08
Other#	0.14	0.06	0.02	0.26	0.02
Psychological	-0.04	0.08	-0.20	0.12	0.65
Family	0.18	0.08	0.03	0.33	0.02
Australia-modified Karnofsky performance status	0.00	0.01	-0.01	0.01	0.93
Resource utilization groups-activities of daily living
Mobility	0.09	0.08	-0.07	0.24	0.28
Toileting	0.05	0.10	-0.15	0.25	0.63
Transfer	-0.03	0.08	-0.20	0.13	0.68
Eating	-0.08	0.09	-0.26	0.11	0.40
Palliative care phase
1 (stable phase)	REF
2 (unstable phase)	0.16	0.12	-0.07	0.39	0.16
3 (deteriorating phase)	0.24	0.10	0.04	0.43	0.02
4 (terminal phase)	-0.14	0.14	-0.41	0.14	0.32

Abbreviation: SD Standard Deviation

#PCPSS Other: Other symptoms include nausea, vomiting, bowel problems, dyspnea, and other distressing symptoms not classified in the pain or psychological categories

Regarding bowel symptom distress (Table 4), higher bowel scores were significantly related to greater appetite loss (β = 0.15, p < 0.01), nausea (β = 0.19, p < 0.01), breathlessness (β = 0.07, p < 0.01), fatigue (β = 0.12, p < 0.01), and pain (β = 0.12, p < 0.01). In addition, bowel distress was significantly related to higher PCPSS “other” scores (β = 0.18, p < 0.01), while other clinician-rated domains showed no significant associations. Interestingly, better functional status as measured by AKPS was also associated with higher bowel scores (β = 0.02, p < 0.01). No additional significant associations were identified.

Table 4.

The generalized estimating equations analysis for symptom assessment scale of Bowel domain score (n = 1,008)

	Estimate	Std.err	95% CI		p-value
Age (years)	-0.0014	0.0043	-0.0099	0.007	0.75
Gender
Male	REF
Female	0.09	0.12	-0.14	0.33	0.44
Non-Cancer	REF
Cancer	-0.32	0.20	-0.71	0.07	0.11
Cancer Without Metastasis	REF
Cancer With Metastasis	0.03	0.14	-0.24	0.30	0.82
Referral type
Outpatient department	REF
Emergency department	-0.44	0.25	-0.94	0.06	0.09
General ward	-0.31	0.23	-0.76	0.14	0.17
Home care service	-0.37	0.31	-0.98	0.24	0.23
Symptom assessment scale
Sleep	0.02	0.02	-0.02	0.06	0.27
Appetite	0.15	0.02	0.11	0.20	< 0.01
Nausea	0.19	0.03	0.13	0.26	< 0.01
Breathlessness	0.07	0.02	0.02	0.11	< 0.01
Fatigue	0.12	0.02	0.08	0.16	< 0.01
Pain	0.12	0.03	0.07	0.17	< 0.01
Palliative care problem severity score
Pain	0.01	0.07	-0.13	0.14	0.93
Other#	0.18	0.07	0.05	0.32	< 0.01
Psychological	0.03	0.09	-0.14	0.20	0.72
Family	-0.09	0.07	-0.23	0.05	0.20
Australia-modified Karnofsky performance status	0.02	0.00	0.01	0.02	< 0.01
Resource utilization groups-activities of daily living
Mobility	-0.03	0.08	-0.18	0.12	0.73
Toileting	0.09	0.09	-0.09	0.27	0.32
Transfer	-0.006	0.08	-0.17	0.16	0.95
Eating	0.07	0.07	-0.07	0.21	0.35
Palliative care phase
1 (stable phase)	REF
2 (unstable phase)	0.13	0.12	-0.10	0.37	0.26
3 (deteriorating phase)	0.03	0.10	-0.16	0.22	0.77
4 (terminal phase)	0.00	0.13	-0.26	0.25	0.98

Abbreviation: SD Standard Deviation

#PCPSS Other: Other symptoms include nausea, vomiting, bowel problems, dyspnea, and other distressing symptoms not classified in the pain or psychological categories

Subgroup analyses stratified by cancer status are summarized in Supplementary Tables 3–5. Baseline characteristics stratified by cancer status are shown in Supplementary Table 2. Compared with cancer patients, non-cancer patients were older (78.71 ± 14.55 vs. 66.33 ± 14.00 years) and had lower functional status, as reflected by lower AKPS scores (27.65 ± 12.31 vs. 35.57 ± 13.17).

In stratified analyses of nausea scores (Supplementary Table 3), age showed a small but significant inverse association with nausea severity in both non-cancer and cancer patients (β = −0.02, p = 0.046; β = −0.01, p = 0.047, respectively). Amongst non-cancer patients, female gender was associated with lower nausea scores (β = −0.60, p = 0.002), while bowel discomfort remained positively associated with nausea (β = 0.12, p = 0.016). In cancer patients, nausea was more strongly linked to co-occurring symptom burden, including sleep disturbance (β = 0.03, p = 0.033), appetite loss (β = 0.11, p < 0.001), bowel discomfort (β = 0.09, p < 0.001), and pain (β = 0.04, p = 0.020).

For appetite distress (Supplementary Table 4), the pattern differed between groups. In cancer patients, appetite scores remained closely associated with multiple co-occurring symptoms—sleep disturbance, nausea, bowel discomfort, fatigue and pain (all p ≤ 0.03)—and were additionally related to clinician-rated PCPSS “other” problems (β = 0.16, p = 0.012) as well as family-related concerns (β = 0.19, p = 0.029). Referral from the general ward was associated with lower appetite distress when compared with outpatient referrals (β = −0.55, p = 0.018), while appetite scores were higher during the deteriorating palliative care phase (phase 3) (β = 0.21, p = 0.046). In contrast, within the non-cancer subgroup, appetite distress was primarily associated with sleep disturbance (β = 0.33, p = 0.001) and pain (β = 0.37, p = 0.002), with functional dependency measures showing significant associations, particularly in toileting (β = 1.22, p = 0.001) and eating (β = −1.43, p < 0.001).

In analyses of bowel symptom distress (Supplementary Table 5), symptom clustering remained evident in both subgroups but was more extensive amongst cancer patients. In the cancer subgroup, bowel scores were significantly associated with appetite loss (β = 0.15, p < 0.001), nausea (β = 0.21, p < 0.001), breathlessness (β = 0.08, p = 0.001), fatigue (β = 0.11, p < 0.001), and pain (β = 0.12, p < 0.001), and were also positively related to AKPS (β = 0.01, p = 0.007). In the non-cancer subgroup, bowel distress was most strongly associated with nausea (β = 0.46, p < 0.001) and showed additional associations with fatigue (β = 0.24, p = 0.007) and pain (β = 0.25, p = 0.035), with a borderline association for AKPS (β = 0.04, p = 0.050).

Supplementary Figs. 1–3 present spaghetti plots of nausea, appetite loss, and bowel discomfort trajectories during the first 7 days of hospice ward admission. This 7-day window was selected because it approximated the mean length of hospice ward stay in this cohort and represented the typical observation period. Beyond 7 days, the number of patients remaining in the ward decreased substantially, and trajectories were more likely to reflect a small number of longer-stay patients rather than the overall cohort. Across the first 7 days, overall mean SAS scores decreased from 0.9 to 0.5 for nausea, from 2.9 to 1.8 for appetite loss, and from 2.4 to 1.5 for bowel discomfort. Although overall mean scores remained within the mild range, individual trajectories varied, supporting the value of repeated daily symptom assessment.

Discussion

The main findings of this study are outlined here. First, amongst the three GI symptoms, loss of appetite was the most prominent at baseline, followed by bowel discomfort and nausea. Second, all three GI symptoms were associated with other coexisting symptoms and clinical factors, suggesting that they should not be assessed in isolation. Third, subgroup analysis showed that the pattern and correlations of GI symptoms differed between cancer and non-cancer patients. A better understanding of these factors may help guide individualized treatment strategies so as to better alleviate symptom burden and improve quality of life in patients with terminal illness. The high in-hospital mortality rate of 72.9% observed in this cohort reflects the advanced end-of-life status of patients admitted to our hospice ward and should be considered when interpreting the symptom burden and longitudinal symptom patterns observed in this study.

Although several associations reached statistical significance, the effect sizes were generally small and should be interpreted cautiously. The relatively low average SAS scores may reflect ongoing symptom monitoring and management within the hospice ward. In the PCOC framework, SAS scores of 1–3 generally reflect mild distress, whereas scores of 4–7 indicate moderate distress [16]. Therefore, small changes within the lower score range may still be relevant for daily quality monitoring and early identification of symptom deterioration, but these findings should be viewed as identifying potential symptom interrelationships rather than demonstrating large clinically meaningful effects.

Previous studies have revealed that more than 30% of patients receiving palliative care experience both loss of appetite and constipation. The same study also reported that approximately 95% of palliative care patients had a diagnosis of cancer [26]. Findings also support the fact that gastrointestinal symptoms are highly prevalent amongst hospitalized patients with advanced cancer [27]. Although primary cancer sites were not available in the present dataset, previous institutional data suggest that gastrointestinal malignancies are common among terminal cancer patients receiving palliative care consultation services at our center. These data were used only as contextual information and were not used to infer cancer-site distribution in the present cohort [28]. Consistent with these findings, most patients in our study had cancer. At baseline, loss of appetite was the most prominent GI symptom on the SAS, followed by bowel discomfort, whereas nausea had the lowest mean score amongst the three GI symptoms. Furthermore, use of the PCOC framework allowed repeated and standardized symptom assessments, making it possible to evaluate not only baseline symptom burden but also symptom trajectories over time [29]. This is particularly relevant in hospice care, where symptom burden may fluctuate dynamically during admission.

Amongst the three GI symptoms, loss of appetite had the highest baseline mean score among the three GI symptom domains and showed statistically significant associations with several coexisting symptoms in the GEE analysis, particularly nausea, fatigue, bowel discomfort, sleep disturbance and pain. These findings suggest that appetite loss in hospice ward patients may reflect a broader pattern of multidimensional symptom burden rather than an isolated GI complaint [30]. Interestingly, patients referred from the general ward had lower appetite scores than those referred from the outpatient department. One possible explanation for this is that patients transferred from the general ward may have already received preliminary symptom management, such as hydration, medication adjustment, or supportive care, prior to admission to the hospice ward [31].

Bowel discomfort was also associated with several coexisting symptoms, particularly loss of appetite, nausea, fatigue and pain, although these associations appeared to be less extensive than those observed for loss of appetite. Better AKPS was statistically associated with slightly higher bowel discomfort scores, but the magnitude of this association was small. This modest association suggests that functional status may not fully reflect subjective bowel-related symptom burden. Previous studies have reported only low-to-moderate concordance between patient-reported outcomes and functional performance measures, indicating that the two assessments provide distinct clinical information [32].

Compared with loss of appetite and bowel discomfort, nausea showed the lowest baseline severity amongst the three GI symptoms. Nevertheless, it showed statistically significant associations with several coexisting symptoms, particularly loss of appetite, bowel discomfort, sleep disturbance and pain. This pattern suggests that even when nausea is not the most prominent GI complaint at baseline, it may still represent part of a broader symptom cluster rather than an isolated symptom and may therefore require management strategies tailored to the specific symptom cluster [33]. Previous studies in patients with cancer have shown that symptoms often occur in clusters, and a gastrointestinal cluster involving nausea and vomiting has been reported in the available literature [34]. In addition, younger ages were associated with slightly higher nausea scores. Although the magnitude of this association was small, it may indicate that the experience or reporting of nausea differs across age groups.

Another noteworthy finding was that pain was significantly associated with all three GI symptoms. This suggests that GI symptom burden in hospice ward patients may be closely linked to overall physical distress rather than confined to a single organ system. Pain may contribute to reduced oral intake, decreased mobility, and greater subjective discomfort, thereby exacerbating appetite loss, bowel discomfort and nausea [35]. In addition, patients with more severe pain may be more likely to receive higher doses of opioid analgesics, which may subsequently contribute to a greater burden of GI symptoms [36]. These findings underscore the importance of comprehensive symptom assessment and integrated management in palliative care.

In the subgroup analysis, baseline GI symptom burden was generally higher in patients with cancer than in those without. Notably, loss of appetite was the leading GI symptom in the cancer group, whereas bowel discomfort was most prominent in the non-cancer group. This difference may be related to the unique clinical features of cancer, such as cachexia and tumor burden, as well as any prior anticancer treatment. In addition, opioid analgesics frequently used for cancer-related pain may further increase GI symptom burden by inducing bowel dysfunction, nausea, vomiting, reflux-related symptoms and constipation [37].

Compared with the non-cancer group, the cancer group showed a more consistent and extensive pattern of associations between GI symptoms and other coexisting symptoms. By contrast, the pattern observed in non-cancer patients was less extensive and appeared to involve more heterogeneous functional and symptom-related factors. Amongst the three GI symptoms, appetite loss appeared to play a particularly central role in the cancer group, being associated with multiple concurrent symptoms and broader palliative care concerns. In addition, bowel discomfort in patients with cancer also showed a more extensive pattern of associations with other coexisting symptoms. This difference may partly reflect the distinct clinical characteristics of cancer patients, including greater exposure to opioid analgesics for pain control, which may contribute to opioid-induced bowel dysfunction [38]. These findings suggest that the clinical context and correlations of GI symptoms may differ between cancer and non-cancer patients, and that symptom assessment and treatment strategies may therefore need to be tailored according to the underlying disease category [39].

This make-up of this study possesses several strengths. First, it applied the PCOC framework to the daily assessment of GI symptoms and their associated risk factors in hospice ward patients, providing a relatively novel longitudinal perspective on GI symptom burden in palliative care. The use of validated instruments, including the SAS, PCPSS, AKPS and RUG-ADL, allowed for a multidimensional evaluation of symptom burden, functional status and palliative care needs [16]. Second, repeated once-daily assessments enabled us to examine symptom trajectories over time rather than relying solely on baseline measurements. Supplementary Figs. 1–3 present spaghetti plots showing that GI symptoms changed dynamically during admission, with an overall tendency toward lower scores but persistent fluctuations, particularly in the non-cancer group. This highlights the clinical value of ongoing daily symptom assessment during hospice care. Third, the use of GEE accounted for within-patient variation over time and thus enhanced the robustness of the analysis. Finally, the inclusion of both cancer and non-cancer patients allowed for subgroup comparisons which subsequently increased the clinical relevance of the findings.

However, several limitations within the study should be acknowledged. First, this was a retrospective cohort study conducted at a single medical center in central Taiwan, which may limit the generalizability of the findings to other institutions or healthcare systems. Second, the analytic dataset classified diagnoses by cancer status and metastatic status but did not include detailed primary diagnoses for non-cancer patients, primary cancer sites among cancer patients, or several clinically important potential confounders, including opioid use, antiemetics, laxatives, prior anticancer treatment, and nutritional support. Therefore, we were unable to report the main diagnoses among non-cancer patients, identify the number of patients with gastrointestinal cancers, directly evaluate the impact of gastrointestinal cancers on GI symptom burden, or adjust for medication and nutritional interventions that may influence appetite loss, nausea, and bowel discomfort. This limitation may substantially affect disease-specific and symptom-specific interpretation, particularly for bowel symptoms. Future studies incorporating detailed diagnostic, medication, and nutritional intervention data may help clarify whether GI symptom patterns and their associations with coexisting symptoms differ across specific cancer sites, non-cancer disease categories, and potentially modifiable treatment-related factors. Finally, although several associations were statistically significant, the effect sizes were small and the average GI symptom scores were relatively low. Therefore, the clinical significance of these associations should be interpreted cautiously. In addition, the imbalance between cancer and non-cancer patients may limit the interpretation of subgroup findings, particularly for non-cancer patients. Future studies with larger and more balanced disease groups and predefined thresholds for clinically meaningful change are needed to better determine the clinical relevance of these associations.

This study has demonstrated that distressing symptoms in patients receiving palliative care are highly interconnected and should not be assessed in isolation. In particular, GI symptoms such as loss of appetite, nausea and bowel discomfort were closely associated with other coexisting symptoms and clinical factors, reflecting the complex and multidimensional nature of symptom burden in hospice ward patients [40]. By applying the PCOC framework, we used standardized and multidimensional tools for daily symptom assessment, which allowed not only evaluation of baseline symptom burden but also dynamic observation of symptom changes over time [24]. A better understanding of symptom clusters and their associated risk factors may help clinicians provide more individualized symptom management and more comprehensive palliative care, ultimately improving quality of life for patients with terminal illness [41].

Conclusion

This study highlights the complex and multidimensional nature of GI symptoms in hospice ward patients receiving palliative care in Taiwan. Using the PCOC framework for daily symptom monitoring, we identified key coexisting symptoms and clinical factors associated with loss of appetite, nausea and bowel discomfort. These GI symptoms were not isolated complaints but rather were closely intertwined with other aspects of symptom burden and differed according to underlying disease category. Our findings support the importance of standardized and repeated symptom assessment in palliative care, as this approach may help clinicians detect dynamic symptom changes, recognize high-risk patients, and ultimately provide more individualized symptom management. Further research remains warranted in order to validate these findings in broader clinical settings as well as to clarify the influence of treatment-related factors on GI symptom burden.

Abbreviations

AKPS

Australia—modified Karnofsky Performance Status

CI

Confidence Interval

GEE

Generalized Estimating Equations

GI

Gastrointestinal

PCOC

Palliative Care Outcomes Collaboration

PCPSS

Palliative Care Problem Severity Score

RUG-ADL

Resource Utilization Groups—Activities of Daily Living

RUG-III

Resource Utilization Groups, Version III

SAS

Symptom Assessment Scale

SD

Standard Deviation

Authors’ contributions

Conceptualization: WMC. Formal analysis: JET. Methodology: WMC and JET. Writing—original draft: CJL and WMC. Writing—review and editing: CFL, CLW, LCL and PSH. All authors approved the final version for submission.

Funding

This research did not receive any specific grant from funding agencies within the public, commercial or non-profit sectors.

Data availability

The dataset is available upon request to the corresponding author.

Declarations

Ethics approval and consent to participate

The study was conducted in accordance with the principles of the Declaration of Helsinki and approved by the Institutional Review Board of Taichung Veterans General Hospital (Protocol code: TCVGH-IRB CE22436A-1), which granted permission to access deidentified patient data for this retrospective study, thereby obviating the need for consent to participate.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.