Participants

Participants were patients admitted to a tertiary care center. The recruitment period for participants was August 20, 2020 to August 19, 2024. Participants were treated for serious health care complications by the Palliative Medicine team and/or Family Medicine. Data on outcome measures (pneumonia, depression, and mortality) and feeding route (oral versus tube) were collected. Data were also collected as related to covariates: (a) mortality risk and (b) age. Moreover, descriptive and clinical data were collected as related to participants’ primary diagnoses, severity of illness, sex, race, and ethnicity. Diagnostic categories proposed for inclusion in this study were as follows: (a) neurological, (b) respiratory, (c) trauma, (d) immunosuppressive/viral, (e) wound/infection, (f) cancer, (g) cardiac, (h) renal disease, and (i) liver disease. Diagnostic categories were chosen based on their frequent alignment with dysphagia, morbidity, and mortality. Data were collected prospectively through observation and electronic chart query in a single-center, using a convenience sampling approach with pre-specified eligibility criteria.

Sample size and data collection

Researchers aimed to collect data over a 4-year period to provide an adequate sample size for statistical analysis. This study was not powered to detect a pre-specified effect size. The sample size estimation was driven by logistical and feasibility reasons. The study closest to the design of this research prospectively analyzed oral versus tube feeding amid advanced-stage dementia in 67 participants [3]. Participant consent was required. The Institutional Review Board (IRB) approved the study and associated consent forms. Written consent was obtained from each participant. The IRB mandated ethical rigor. All data were deidentified for the purpose of statistical analysis.

Researchers conducted electronic chart queries to collect data for this prospective observational analysis. The index disease (primary diagnosis) of each participant was determined based on the International Classification of Diseases (ICD)-10 code entered by the lead physician/health care provider. The outcome measure pneumonia was detected via ICD-10 coding if newly diagnosed. Physicians diagnosed pneumonia and entered this diagnosis by ICD-10 code based on radiology studies/chest imaging specific to pulmonary health, clinical symptoms, physical chest examinations, and laboratory markers/tests (e.g., sputum grain stain and culture, inflammatory markers such as procalcitonin, and blood cultures). The CURB (Confusion, Blood Urea Nitrogen [BUN], Respiratory Rate, Blood Pressure)-65 score was recorded during the participant’s hospital stay if pneumonia was present on admission (i.e., to determine worsening of pulmonary state or lack thereof, as related to feeding route). *Of note: the CURB-65 scores did not change throughout the hospital stay for the five participants diagnosed with pneumonia upon admission. The outcome measure depression was detected via ICD-10 coding if newly diagnosed. Physicians diagnosed depression and entered this diagnosis by ICD-10 code based on responses to the Patient Health Questionaire-9 (PHQ-9) and/or based on depression symptoms aligned with Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria: (a) depressed mood or anhedonia (i.e., loss of interest); (b) appetite/weight changes; (c) sleep disturbances; (d) fatigue or loss of energy; (e) feelings of being unworthy or extreme guilt; (f) psychomotor agitation or retardation; (g) disrupted ability to think or concentrate, and (g) suicidal ideation. Likert scales were developed to assess degree of depression via self-report if present on admission (i.e., to determine worsening of depression or lack thereof, as related to feeding route). Likert scales were created by the researcher and approved by the IRB. In theory, the investigator would have asked each participant who qualified to rate depression pre- and post-hospital admission from 1–10 with 10 equal to, “My depression is the worst it has ever been.” No patient met the criteria to participate in the depression inventory. To clarify, Likert scales were authorized but not used. *Of note: for this particular study, investigators elected to use Likert scales versus the PHQ-9 if depression was coded by ICD-10 in the electronic medical record (EMR) as an existing diagnosis. This avoided emergent management of responses (by researchers) to the final question on the PHQ-9 related to suicidal ideation. The hospital organization has existing policies that require treatment protocols for thoughts of suicide, and the primary medical team is required to enact these protocols. The outcome measure mortality was detected via manual chart review and confirmed via the hospital’s Information Technology (IT) Department of Analytics.

The covariate mortality risk and corresponding descriptive data point severity of illness scores were determined by an algorithm informed by each participant’s health care complications and diagnoses and provided to researchers by IT Analytics. Age, sex, race, and ethnicity were obtained from participants’ patient profiles in their digital charts. The first author manually inspected each chart and spoke with each participant or authorized patient representative/Medical Power of Attorney (MPOA) after consent for research participation to determine the primary route of feeding (route of longest duration). The first author also conducted manual chart analysis for each participant to detect presence or absence of dysphagia, determined by: (a) health care provider diagnosis by ICD-10 code or documented in provider digital chart notes, (b) clinical examination of swallowing, or (c) instrumental examination of swallowing.

A total of 69 participants were consented for participation in this study. Due to Covid-19 consent protocols in 2020, MPOAs provided virtual informed consent for four participants amid visitation restrictions. The IRB later advised the authors that in-person consent was preferred due to the sensitive nature of the research topic and the terminal nature of the patient population. Therefore, researchers elected to withdraw four participants, yielding a final sample size of 65 participants in the analysis dataset.

Inclusion criteria were defined a priori as follows

• all adult ages (18 years of age or older);
• presence of dysphagia;
• presence of life-limiting illness/diagnosis;
• admitted to UMC and treated by Palliative Medicine and/or Family Medicine;
• patient or authorized representative chooses nutrition by informed autonomous consent, and
• mortality risk and severity of illness scores accessible via IT Analytics related to diagnoses found in the electronic medical record.

Participants who were actively dying (imminent death) or did not desire nor accept nutritional support were excluded from this research study. It was important to exclude patients who were actively dying and thereby not receiving nutrition, because inclusion of these participants would have negatively impacted generalizability of results.

Design and statistical analysis

This study involved a prospective observational analysis of participants with life-limiting illness and dysphagia who had chosen either oral or enteral (tube) feeding for nutrition. Each participant received oral care four times per day, per the organization’s oral hygiene protocol. Participants who received tube feeding or participants who received oral feeding and had a history of gastrointestinal abnormalities, such as reflux, had positioning restrictions and were not reclined below 30 degrees per the organization’s protocols. The dependent variables were pneumonia, depression, and mortality. The independent variable was feeding route (oral or tube). Pneumonia was identified from ICD-10 scores or CURB-65 scores and defined as newly diagnosed/worsening during admission or absent. Depression was identified from ICD-10 scores or Likert self-report scales and defined as newly diagnosed/worsening during admission or absent. The outcome measure mortality was identified through manual chart review and confirmed through an electronic query. It was defined as discharge to hospice (i.e., prognosis of 6 months or less to live) or patient death.

Covariates were mortality risk and age. Mortality risk is defined as the “likelihood of dying” (p. 8) according to the Methodology Overview of the 3M^TM All Patient Refined Diagnosis Related Groups [38]. Age (dichotomous) was defined as years of living. Descriptive data point severity of illness is defined as “the extent of physiologic decompensation or organ system loss of function” (p. 8). Both mortality risk and severity of illness are derived from calculating the health impact of a patient’s medical diagnoses and health care complications, and they have been determined to be the most accurate comorbidity indices [39]. Mortality risk and severity of illness scores are defined as follows: (a) score of 1 = minor, (b) score of 2 = moderate, (c) score of 3 = major, and (d) score of 4 = extreme. The algorithm assigns metrics to each patient’s physiological stability or lack thereof (including and beyond the primary and secondary diagnoses). Resulting metrics are extrapolated from a vast array of health care complications, including but not limited to malnutrition, obesity, conditions resulting from hormonal imbalance, conditions resulting from metabolic disturbances, and need for surgical or invasive procedures [40].

Sex (dichotomous) was defined as biological identity. As related to descriptive clinical data, index diseases ultimately included in the study (assigned to diagnostic categories neurological, respiratory, trauma, immunosuppressive/viral, cancer, and renal disease) corresponded to an ICD-10 code and were defined as dichotomous (existing or absent). Race was defined as: (a) American Indian or Alaska native, (b) Asian, (c) Black or African American, (d) Native Hawaiian or Other Pacific Islander, or (e) White. Ethnicity was defined as: (a) Hispanic or Latino or (b) Not Hispanic or Latino.

The primary aim of this study was to evaluate the association between feeding route (oral versus tube) and clinical outcomes among patients who desire nutrition amid a life-limiting illness and dysphagia. The outcome measures were pneumonia, depression, and mortality. The secondary aim was to contribute to shifting the nutritional decision-making paradigm in Palliative Medicine.

Descriptive statistics were utilized for baseline demographic and clinical characteristics. Continuous variables were summarized as mean (standard deviation), median (interquartile range), and range (minimum-maximum), while categorical variables were presented as frequency (percentage). Differences in distribution of continuous variables across the two study arms were tested for statistical significance using the Wilcoxon rank-sum test. Bivariate association between study arms and categorical variables were tested for statistical significance using either the Chi-squared test or Fisher’s exact test.

The association between feeding routes and the three clinical outcomes was assessed separately using logistic regression modeling. Logistic regression was selected because it estimates the odds of an outcome while adjusting for potential confounders, allowing for direct assessment of the independent association between feeding route and clinical outcomes. Given the low number of events in the analysis dataset, we decided to limit the number of confounders to two to avoid model overparameterization [41]. Based on findings from the exploratory data analysis, there were significant differences in age, mortality risk scores, and illness severity scores between the two study groups. Additional analysis further revealed that mortality risk scores and illness severity scores were significantly correlated. Hence, we decided to include age and the mortality risk index as confounders in the logistic regression model. Logistic regression modeling was done sequentially as follows: 1) first, an unadjusted analysis was completed with the feeding route as the independent variable; 2) secondly, an adjusted analysis was done by including age and mortality risk scores into the model from step 1.

We conducted an additional statistical analysis with propensity matched data to obtain two comparable groups to assess the effect of feeding route on clinical outcomes. While logistic regression adjusts for measured confounders within the regression model, propensity score matching creates a pseudo-randomized dataset by balancing baseline characteristics between groups prior to outcome assessment. Patients who were tube-fed were matched to patients who were oral-fed using cardinality matching [42]. The matched dataset had 30 patients (15 in tube-fed group and 15 in oral-fed group). Matching was employed to optimize comparability between groups, yielding a matched dataset that achieved balance not only in terms of matching variables (mortality risk, age, and diagnosis), but also with respect to sex, race, ethnicity, and severity of illness.

A two-sided p-value of 0.05 was used to determine statistical significance. The uncertainty in parameter estimates was reported using a two-sided 95% confidence interval. The analysis was done using R version 4.4.1.

Primary outcomes

The incidence of pneumonia was significantly higher in tube feeding group compared to the oral feeding group (79% vs 12%, p < 0.001) (see Table 3). Logistic regression analysis showed that tube feeding was associated with significantly increased odds of pneumonia (adjusted OR = 19.28, 95% CI: 4.5–109.6, p < 0.01), even after controlling for age and mortality risk (see Table 4). Depression was significantly higher among those receiving tube feeding (50% vs 9.8%, p < 0.001) compared to oral feeding. Adjusted logistic regression indicated a strong association between tube feeding and depression (adjusted OR = 17.25, 95% CI: 3.13–158.78, p < 0.01) (see Table 5). Mortality was observed in 18 participants (28%), with a higher death rate in the tube feeding group (46%) compared to the oral feeding group (17%) (p = 0.012). Although the unadjusted odds ratio suggested an increased risk of mortality with tube feeding (OR = 4.11, 95% CI: 1.34–13.47, p = 0.015), the adjusted model did not reach statistical significance (adjusted OR = 2.78, 95% CI: 0.71–11.7, p = 0.147) (see Table 6). The incidence of at least one adverse event (i.e., pneumonia, depression, or mortality) was significantly higher among participants receiving tube feeding (96% vs 37%, p < 0.001) (see Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Clinical Outcomes – Descriptive Statistics.

https://doi.org/10.1371/journal.pone.0333895.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Association between Pneumonia and Feeding Mechanism.

https://doi.org/10.1371/journal.pone.0333895.t004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Association between Depression and Feeding Mechanism.

https://doi.org/10.1371/journal.pone.0333895.t005

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Association between Mortality and Feeding Mechanism.

https://doi.org/10.1371/journal.pone.0333895.t006

A composite outcome variable was created to assess whether any of the three conditions—pneumonia, depression, or mortality (PDM)—were present. This composite variable was created to shift the lens and view participants’ physiological and psychological responses to feeding route holistically, recognizing that the occurrence of even one of these outcomes represents a meaningful decline in patient health and may warrant changes in clinical management. An adverse outcome was defined as increased odds of either pneumonia, depression, or mortality occurring based on feeding route (oral or tube). An adverse outcome (i.e., at least one outcome measure increasing as related to feeding route) was observed in 38 participants (58%), with a strikingly high prevalence in the tube feeding group (96%) compared to the oral feeding group (37%) (p < 0.001). Logistic regression analysis revealed that tube feeding was associated with a significantly increased risk of experiencing at least one adverse effect (adjusted OR = 55.64, 95% CI: 7.59–1235.15, p < 0.01), demonstrating a significant association between feeding method and negative clinical aftermaths (see Table 7).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 7. Association between PDM (at least one of Pneumonia/Depression/Mortality) and Feeding Mechanism.

https://doi.org/10.1371/journal.pone.0333895.t007

Propensity-matched analysis

A propensity-matched dataset of 30 participants (15 oral, 15 tube) was analyzed (see Table 8). Results from propensity-matched data reinforced previous findings, showing significantly higher rates of pneumonia (73% vs 13%, p < 0.001) and depression (47% vs 6.7%, p = 0.035) in the tube feeding group (see Table 9). Mortality remained higher but was not statistically significant in the matched cohort (40% vs 27%, p = 0.4). The persistence of significant differences even after controlling for potential confounders further supports the observed trends and associations.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 8. Distribution of Baseline Variables in Propensity Matched Data.

https://doi.org/10.1371/journal.pone.0333895.t008

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 9. Clinical Outcomes – Bivariate Association.

https://doi.org/10.1371/journal.pone.0333895.t009

Summary of key findings

This study provides compelling statistical evidence that tube feeding is associated with significantly higher risks of pneumonia and depression, with mortality also trending higher. The PDM analysis further reflects the increase in adverse outcomes associated with tube feeding. Findings show trends toward harm with tube feeding versus oral feeding in the context of serious illness and difficulty swallowing, supporting the need for further exploration of reported associations. Findings challenge the long-held belief that tube feeding is a safer alternative, suggesting instead that it may contribute to greater clinical deterioration in patients with dysphagia and life-limiting illnesses. A paradigm shift in medical decision-making regarding feeding routes in palliative care may need to be considered, especially if continued research with a larger sample size and multi-site approach replicates results.

Comparison to previous research

The findings align with previous studies that have reported increased harm from tube feeding as related to oral feeding in specific patient populations (e.g., advanced dementia, stroke, COPD). Cintra et al. [3] investigated tube versus oral feeding in the context of advanced-stage dementia and found that pneumonia and mortality increased amid tube feeding compared to oral feeding. Hanners Gutierrez et al. [5] reported pneumonia to be 10.14 times more likely to occur with tube versus oral feeding. Increased incidence of depression (2.79) and mortality (3.02) were also noted to be worse with tube feeding compared to oral feeding. In a content-rich review of ethical principles at the end of life, Akdeniz et al. [43] reported increased aspiration pneumonia and gastrointestinal complications with tube feeding. Johnston et al. [44] reported that 43% of patients who received a percutaneous endoscopic gastrostomy (PEG) tube placement died within 1 week and 70% of those patients died due to respiratory disease. In contrast, careful hand oral feeding as an alternative to tube feeding has been found to improve both psychosocial and physiological aspects of nutrition.

Clinical implications, strengths, and limitations

The significant association between tube feeding and adverse outcomes reinforces the need for an improved approach to nutritional decision-making in palliative care. Evidence supports prioritization of patient autonomy and quality of life when discussing feeding options with patients who have dysphagia and are pleading to eat by mouth. The results of this groundbreaking pilot study reveal that oral feeding or careful hand oral feeding may be a viable and less harmful alternative to tube feeding for patients with dysphagia and life-limiting illness. Past studies have revealed the need for greater emotional support as dependence for cares increases [45], and interactive, careful hand feeding offers one way to provide this type of support [46]. With 55.64 greater odds of developing at least one adverse event (pneumonia, depression, mortality) due to tube feeding as compared to oral feeding, health care providers must heavily weigh the choice to feed someone enterally (by tube). Tube feeding should not be implemented by default for patients exhibiting swallowing problems when they are terminally ill.

This study has several notable strengths. The researchers completed a pilot study to investigate a key element of end-of-life care which required fortitude necessary to consent participants while they were enduring serious emotional and physical challenges. Several participants felt that they could leave a legacy through their participation. Offering this type of psychological support and establishing a connection with participants in this manner was and is indeed a strength. This is the first study to compare the impact of oral or tube feeding on pneumonia, depression, and mortality across multiple life-limiting diagnoses. The prospective nature of the study supported systematic data collection. Moreover, the use of logistic regression and propensity score matching enhanced the robustness of the findings and isolated potential confounding variables, such as age and mortality risk, from the treatment effect. Strong statistical significance was achieved across analyses, raising confidence in results. Lastly, the ICC values (1.0) indicated excellent interrater and intrarater agreement, demonstrating excellent consistency in data collection. These results enhance the credibility of the findings.

There are limitations that must be acknowledged. This pilot, exploratory study had a relatively small sample size (N = 65). However, the sample size aligns with the only other known prospective study to compare oral and tube feeding in patients with dysphagia at the end of life (N = 67) [3]; thus, contributing to a meaningful body of evidence and providing a foundation for future research. Additionally, the study was conducted in a single tertiary care center. These limitations present an opportunity for similar research to be conducted on a broader scale across multiple institutions, increasing the sample size and diversifying the settings. Expanding the scope of future studies may replicate these findings and enhance generalizability.

Future research and conclusion

Further research is needed to explore the impact of feeding route (oral or tube) on outcomes (pneumonia, depression, mortality) when a patient desires nutrition amid a life-limiting illness and dysphagia. Larger-scale studies should aim to replicate these findings in diverse patient populations and investigate additional factors such as caregiver burden. Additionally, qualitative research exploring patient and family perspectives on nutritional decision-making could provide valuable insights into the psychosocial aspects of nutritional management at the end of life. It would be fruitful to explore protocols and perspectives of health care providers as related to understanding and disclosure of risks and benefits of different feeding routes and compare the perspectives of patients, MPOAs, or authorized patient representatives.

The results of this study provide compelling evidence that tube feeding in patients with dysphagia and life-limiting illness is associated with increased risks of pneumonia and depression, with a strong trend toward higher mortality. These findings challenge traditional assumptions regarding the safety and efficacy of tube feeding in palliative care and highlight the need for an evidence-based change in clinical practice. Findings support increased educational efforts to ensure health care providers are armed with well-grounded and recent evidence related to selection of a feeding route in the context of serious or terminal illness. An infrastructural health care shift toward a standardized feeding route selection protocol may be fruitful to ensure best practices are enacted (e.g., adherence to biomedical ethical principles, informed autonomous consent) versus default use or overuse of tube feeding in the context of dysphagia, and legislation should be proposed to solidify educational and clinical paradigm changes. Future research should build upon this foundation to refine evidence-based guidelines that prioritize biomedical ethics, psychosocial health/support, physiological comfort, and reduction of suffering in end-of-life nutrition decision-making.