- Email: [email protected]
Expectations and Outcomes of Gastric Feeding Tubes
The American Journal of Medicine (2006) 119, 527.e11-527.e16
CLINICAL RESEARCH STUDY
Expectations and Outcomes of Gastric Feeding Tubes Timothy S. Carey, MD, MPH,a,b Laura Hanson, MD, MPH,e Joanne M. Garrett, PhD,b Carmen Lewis, MD, MPH,b Nancy Phifer, MD,c Christopher E. Cox, MD, MPH,d Anne Jackman, MSWa a
The Cecil G. Sheps Center for Health Services Research and bDivision of General Internal Medicine and Clinical Epidemiology, University of North Carolina at Chapel Hill; cMoses Cone Hospital, Greensboro, NC; dDivision of Pulmonary and Critical Care Medicine, Duke University Medical Center, Durham, NC; eDivision of Geriatrics, University of North Carolina at Chapel Hill. ABSTRACT PURPOSE: To compare expected outcomes with actual outcomes from tube feeding in adult patients. SUBJECTS AND METHODS: This prospective cohort study was conducted in two North Carolina hospitals. Surrogates were interviewed shortly after feeding tube insertion and at 3- and 6-month follow-up; chart abstraction and death certificate review also were carried out. Participants were surrogate decisionmakers for consecutive adult patients who received new feeding tubes. RESULTS: There were 288 patients with surrogate decision-makers enrolled. Mean age was 65 years; 30% had a primary diagnosis of stroke, 16% neurodegenerative disorder, 20% head and neck cancer, and 30% other diagnoses. At 3 months, 21% of patients had died, and 6-month mortality was 30%. At 3 months, 38% of survivors were residing in a nursing home, and 27% had the feeding tube removed. Patients were impaired in most activities of daily living (ADLs) with little change over time. Medical complications were common: 25% of patients had decubitus ulcers at 3 months, and 24% had at least one episode of pneumonia. Perceived global quality of life was poor at 4.6 (on a 0-10 scale) at baseline, and surrogates anticipated this would improve to 8.0 with tube feeding. Family surrogates’ expectations for improvement from the feeding tube were very high at baseline and remained so at 3 and 6 months. CONCLUSIONS: Families’ high expectations of benefit from tube feeding are in contrast to clinical outcomes. Providers and families need better information about the outcomes of this common procedure. © 2006 Elsevier Inc. All rights reserved. KEYWORDS: Gastric feeding tube; End-of-life care
Gastric feeding tube insertion allows long-term enteral feeding for patients who are unable to hydrate or nourish themselves. Use of the procedure has increased substantially over the past 15 years since the introduction of percutaneous methods of tube insertion.1,2 The ease of insertion with percutaneous techniques may be contributing to increased use. This phenomenon may be analogous to the increased cholecystectomy rate after the introduction of laproscopic cholecystectomy in the late 1980s.3 In North Carolina, we found that the rate of feeding tube placement approximately doubled over an 11-year period.4 One recent report from the Requests for reprints should be addressed to Timothy S. Carey, MD, MPH, The Cecil G. Sheps Center for Health Services Research, 725 Airport Road, CB#7590, The University of North Carolina at Chapel Hill, Chapel Hill, NC 27599-7590. E-mail address: [email protected].
0002-9343/$ -see front matter © 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.amjmed.2005.11.021
Veterans Administration did indicate a recent decrease in feeding tube use among some populations.5 This utilization has occurred despite controversy about both the benefit and the ethics of the procedure. Several studies have indicated that patients who receive feeding tubes have a 50% 1-year mortality rate.6-10 Some authors recommend new decisionmaking approaches to tube feeding to discourage use of this procedure in patients with advanced dementia or terminal illness.10-13 One explanation for the continuing increase in feeding tube utilization is that patients and families who make decisions for cognitively impaired patients may have expectations for tube feeding that are not based on information about outcomes data, leading to an over-estimation of the prognosis for both survival and functional recovery.
527.e12
METHODS
The American Journal of Medicine, Vol 119, No 6, June 2006
nia; living longer; more freedom; better health; fewer coughing episodes; and greater enjoyment of food (included Patients who received an initial gastric feeding tube were as a distracter to assess global optimism). Higher scores recruited at two North Carolina Hospitals: a universityindicated higher expectations. We dichotomized the 4-level based 608-bed tertiary care hospital and a 500-bed commuresponse, which ranged from strong agreement to strong nity hospital about 45 miles away. Patients were identified disagreement for each question by daily checks of inpatient gasinto agree versus disagree; analytrointestinal procedures, intervenses using the numerical 4-point tional radiology schedules and CLINICAL SIGNIFICANCE scale gave similar results. We chart review. Inclusion criteria asked surrogates at 3 and 6 were: age ⬎17 years and ability of ● The families of patients who receive gasmonths whether their expectations patient or surrogate to speak Entric feeding tubes have very high expechad been met. We operationalized glish. Exclusion criteria were: tations about longevity and improved this with the following question: non-ENT (ear, nose and throat) functioning after feeding tube insertion. “We asked you about your expecmetastatic malignancy, major trauma, tations when it was decided that ● These high expectations persist over a noncompetent patient with no suryour [mother/father, etc.] would 6-month period in spite of an overall rogate decision-maker, ward of have a gastrostomy feeding tube. the state or patient death before lack of clinical improvement in recipiWe would now like to know how completion of consent. Early ents of gastric feeding tubes, with overyour experience has been after deaths and metastatic cancer were all high levels of dependence in activiyour [mother/father, etc.] had this excluded because our aims were ties of daily living, and 30% mortality. procedure . . .” Surrogates then to examine the medium-term (3answered identical items about ex● Over a 6-month period of observation to 6-month) functional outcomes pectation of benefit at this time. after feeding tube insertion, about one and fulfillment of expectations Estimated probability of surwith tube feeding. We excluded third of patients were able to resume vival was obtained from family trauma patients because they often nourishment without a feeding tube. surrogates at baseline. Functional have a very good prognosis, and ● Clinicians need to develop improved status was reported by surrogates the feeding tube is transient. To methods of communicating with famiat baseline and at 3 and 6 months ensure a uniform source of data, using a modification of the Katz lies of these seriously ill patients in we attempted to interview a surroADL (Activities of Daily Living) order to convey realistic expectations gate in all cases. We initially con14 scale. We excluded eating from tacted the family member who and encourage appropriate shared the ADL scale because, by definisigned the procedure consent for decision-making. tion, tube-fed patients are unable the feeding tube, confirming that to eat independently. Baseline they were the family member data were not fed back to respon“most familiar with the patient’s dents at 3- and 6-month interviews. Information also was care.” Most interviews were conducted several days to 2 collected on other health outcomes such as nursing home weeks after feeding tube insertion. residence, feeding tube removal, incidence of pneumonia, Follow-up interviews were conducted at 3 and 6 months. decubitus ulcers, and use of restraints. Surrogates were If the patient had died before the follow-up interviews, a asked about these conditions for the 3 months before feedmodified surrogate interview was conducted approximately ing tube insertion, as well as in the interval to each fol1 month after the patient’s death. Family surrogates of low-up interview. Most of the baseline surveys were conpatients who died before the 3-month interview were not ducted in person; follow-up interviews were by telephone. contacted at 6 months. If the feeding tube had been removed Analyses were performed using STATA 8.0 (StataCorp, before the time of the 3-month interview, the 6-month College Station, Tex). interview was redundant and therefore not conducted. MorWe measured outcomes at both 3 and 6 months, includtality was ascertained through surrogate interview and ing survival, ADL score, residence in a nursing home, and search of the North Carolina Death Index. feeding tube removal. We compared several factors (eg, The content of the interviews included an index of items patient age, functional status, race, diagnosis) with each of reflecting possible expectations for improvement from the these outcomes using t tests for continuous variables and gastrostomy tube that was generated from review of the Pearson chi-squared for categorical comparisons. We then outcomes literature and from open-ended interviews with developed multivariable logistic regression models to deterexperienced physicians. The final index was composed of mine which set of factors was predictive of outcomes, such the sum of 14 items reflecting expectations for: improved as residence in a nursing home or mortality at 3 and 6 nutrition; being more comfortable; less pain; fewer probmonths, and removal of the feeding tube at each survey lems eating; greater independence; not feeling hungry or thirsty; quality of life; having a decreased risk of pneumointerval. Finally, we performed survival analysis using a
Carey et al
Gastric Feeding Tubes
Cox proportional hazard model. The University of North Carolina Committee on the Protection of Human Subjects approved the research.
RESULTS During enrollment, 722 patients had gastrostomy, and 306 of these were ineligible. The main reasons for exclusion were metastatic cancer (77), diagnosis of trauma (44), repeat feeding tube procedure (34), and ward of the state (13); 29 patients died before initial contact, with the remainder miscellaneous reasons. Of the 416 eligible patients, we successfully enrolled surrogate decision-makers for 288 patients with new gastrostomy feeding tubes (69% enrollment of eligibles). Among eligibles who were not enrolled, the main reasons for nonenrollment were: 35 refusals; in 15 cases the physician thought the patient was not appropriate; and in 59 cases we were unable to contact the surrogate respondent despite repeated attempts . The surrogates were generally close relatives, with 38% a spouse and 37% an adult child. Patient characteristics and outcomes were similar between the two clinical sites. We had complete baseline information from surrogates on 288 patients. We completed 228 surrogate 3-month interviews; 51 surrogates refused or could not be contacted despite repeated attempts. Six patients died shortly after consent; 3 patients had their feeding tubes removed shortly after consent. For these 9 patients, the baseline survey provided information on the total experience of the patient and surrogate with tube feeding. The completion rate among those eligible for the 3-month interview was 82%. At 6 months, a total of 107 patients did not require further follow-up because they had either had the feeding tube removed before the 3 month interview (n ⫽ 61) or had died before the 3 month interview (n ⫽ 55). Of the 121 patients eligible for the 6-month interview, 39 surrogates could not be contacted, for a completion rate of eligibles of 68%. Ascertainment of mortality was complete because we used the North Carolina Death Index. We completed 82 surrogate 6-month interviews. Given the high numbers of patients who died, had feeding tubes removed, or who were lost to follow-up, the generalizability of the 6-month follow-up is somewhat limited. Demographic and clinical characteristics of patients lost to follow-up did not differ significantly from surviving patients retained in the cohort. The baseline characteristics of the patients are in Table 1. The mean age was 65.8 years; 39% were aged over 74 years, and 41% were under 65. The age distribution within diagnostic categories did vary, with 59% of ENT cancer patients under age 65 and only 29% of patients with neurodegenerative disorders under that age. Patients were 54% men, and one third were black. Half had income ⬍$10, 000 per year. The underlying clinical conditions were varied, with 30% having stroke; 16% a neurodegenerative disease, including dementia; 20% head and neck cancer; and 34% a wide variety of other serious medical conditions. The mean Charlson comorbidity score was 2.7 (SD ⫽ 1.7), indicating
527.e13 Table 1
Baseline Characteristics (n ⫽ 288)
Age in years (SD) Male Race White Black Other Education ⬍ HS Income ⬍ $10 000 Underlying diagnosis Stroke Neurodegenerative ENT cancer Other ADL (SD) Nursing home residence Before procedure Decubitus ulcers in the 2 weeks before tube insertion Quality of life (SD)
65.8 (16.6) 54% 61% 37% 2% 36% 50% 30% 16% 20% 34% 11.8 (3.75) 36% 23% 4.6 (2.31)
the presence of multiple comorbid illnesses.15 When we asked the surrogate to rate the patient’s baseline quality of life on a 0-10 scale, it was relatively low, at 4.6. The patients were partially or totally dependent in all 5 activities of daily living (toileting, continence, walking, dressing, bathing) at the time of the baseline interview. When we assigned 1-3 points to levels of independence in activities of daily living (1 ⫽ independent, 2 ⫽ needing assistance to accomplish the task, 3 ⫽ completely dependent) and then summed them (range 5-15), the mean baseline ADL score was 11.8, indicating a severe level of dependency. Thirty-six percent had resided in a nursing home before insertion of the feeding tube.
Expectations of Benefit from Feeding Tube Placement The surrogate responses to each of the questions on expectations are shown in Table 2. At baseline, these expectations were very high (mean ⫽ 3.23 on a 4-point scale). For example, 79% “agreeing” or “agreeing a lot” that the patient would be more comfortable after the gastrostomy; 56% expected to have greater independence; 87% expected to have better quality of life; 66% expected to have decreased risk of pneumonia; and 93% expect the patient’s overall health to be better. Responses to experience with expected benefits remained high at 3 months (mean ⫽ 3.06), with little change from baseline. Among surrogates, expectation of benefit remained similarly high at 6 months for the minority of patients who survived and still had a feeding tube in place. Surrogates of decedents did have lower assessments, but excluding their responses did not significantly change the overall rating of expectations (data not shown). When we queried the surrogates at baseline about the life expectancy of the patient, 61% thought there was a ⬎90%
527.e14
The American Journal of Medicine, Vol 119, No 6, June 2006
Table 2 Surrogate Expectations about Benefit of Gastrostomy Feeding (% Agree or Strongly Agree with Each Statement) Expectation
Baseline n ⫽ 288
3 month n ⫽ 228
a. Improvement in nutrition b. Be more comfortable c. Have less pain d. Fewer problems eating e. Have greater independence f. Not to feel hungry/thirsty g. Better quality of life h. Decreased risk of pneumonia i. Live longer j. Enjoy food k. Have more freedom l. Health to be better m. Fewer coughing episodes n. Fewer choking episodes
96 (96)* 79 (79) 64 (65) 83 (83) 56 (55) 69 (70) 87 (87) 66 (66) 90 (90) 39 (40) 54 (54) 93 (93) 67 (68) 77 (79)
87 71 60 78 54 67 80 65 91 29 48 83 62 74
*The percentages in parentheses are the baseline values for those subjects interviewed at 3 months (n ⫽ 228).
probability that the patient would live more than 1 year. At 3 months, 67% of surrogates for survivors thought the patient would survive for at least 1 more year, and at 6 months, 61% did so. The surrogate-assessed mean global quality of life (0-10) did change slightly from 4.6 at baseline to 6.0 when rated at 3 months for surviving patients. When we asked surrogates at the baseline interview what they would expect their family member’s quality of life to be at 6 months, they estimated a high quality of 8.0.
Health Outcomes At 3 months, 21% of the patients had died, and 6-month mortality was 30%; 82 patients survived beyond 3 months and had complete interviews. The mortality rate was very similar across the four diagnostic strata (stroke, ENT cancer, neurodegenerative disease, and other serious medical conditions) after we controlled for baseline age and functional status as assessed by ADLs. In a Cox proportional hazard model we examined age, race, sex, educational status, presence of an ENT cancer diagnosis, nursing home status at baseline, ability to eat somewhat at baseline, and ADLs at baseline. Only age and baseline ADL predicted mortality over the 6 months. For each decade of increased age, the hazard ratio increased by 17%. For each increased point in ADL from 5-15 (indicating increasing dependence), the hazard ratio increased 9%. In a logistic regression model examining 6-month mortality and the above independent variables, we estimated that a 30-year-old had a 15% probability of dying over the 6-month period, and a 90-year-old had a 40% probability. An individual with a baseline ADL score of 5 (least dependent score) had an estimated 19% 6-month mortality, which rose to 33% with a score of 15. The baseline functional status score was worse (higher) among those who eventu-
ally died, compared with those alive at 3 months (14.1 vs 9.8). Although ENT cancer patients had similarly high mortality during follow-up, their baseline and follow-up functional status differed from other diagnostic groups. ENT cancer patients had a much better ADL score of 6.6, compared with 12.5 for the other diagnostic strata. Only 5% of ENT cancer patients resided in a nursing home before insertion of the feeding tube, compared with 42% of patients in the other strata. At 3 months, 77% of ENT cancer patients were independent in two or more ADLs, compared with 26% of patients in other strata. Thirty-two percent of subjects had the feeding tube removed between baseline and 3 months, and 63% took at least some nutrition by mouth, although families often told us that this feeding was without medical approval. In multivariable logistic regression analysis we found that significant baseline variables predictive of feeding tube removal at 3 months were younger age (34% at age 50, 15% at age 90), white race (31% white race, 18% black), male sex (34% men, 18% women), diagnosis other than ENT cancer (only 9% of cancer patients had tubes removed at 3 months), and less dependent baseline ADL (probability decreased from 41% at an ADL of 5 to 19% for an ADL of 15). The incidence of pneumonia was 24% between baseline and 3 months, and 12% among surviving patients from 3 to 6 months (Table 3). One in four patients had decubitus ulcers at the 3-month interview. We did ask the surrogates at baseline about pneumonia and decubiti in the 3 months before feeding tube insertion, and the prevalence before feeding tube insertion was similar to the 3 months after insertion.
DISCUSSION This multi-site, prospective cohort study found that families have high expectations for improvements in both longevity
Table 3
Other Health Outcomes
Mean ADL (range 515, higher more dependent) Fully independent in 2 or more ADLs (%) Residence in a nursing home (%) Feeding tube removed and patient alive (%) One or more episodes of pneumonia (%) Decubitus ulcers (%) Restraint use (%)
Baseline n ⫽ 288
3 months n ⫽ 228
6 months* n ⫽ 82
11.8
10.8
10.3
20
31
34
36
38
37
–
32
23
11
24
12
23 30
25 27
25 14
*Subset of patients whose feeding tube still in place at 3 months.
Carey et al
Gastric Feeding Tubes
and functioning when interviewed after the placement of a gastrostomy feeding tube. The mortality in this study was high (30% over 6 months) with little functional improvement among survivors. The mortality rate was similar across the 4 diagnostic groups after adjustment for age and baseline functional status. The high baseline expectations of benefit did not significantly change over time. Feeding tube use remains very common, with a recent study by Mitchell et al demonstrating that over one third of severely cognitively impaired residents in US nursing homes have feeding tubes.16 In another secondary data analysis, Rabeneck and colleagues examined outcomes of patients receiving gastrostomy through VA hospitals.17 They found a median survival of 8 months, with increasing mortality with greater age and malignant diagnosis. Our outcome analysis is most similar to a community-based study conducted by Callahan and colleagues in a single-site study of 150 patients with feeding gastrostomy in Indiana.6 He found a 1-year mortality rate of nearly 50% and similar rates of decubitus ulcers to those found in our study, but did not examine surrogate expectations of benefit relative to these outcomes. Why are families so optimistic about the feeding tube, and how does this optimism persist in the face of little functional improvement over time? Most of these patients had long hospital courses and chronic illness—the feeding tube often represented one in a long line of procedures, though one that was perhaps appealing because it was associated with nutrition. It is also possible that families were not adequately informed about what to expect after tube insertion. A cross-sectional survey of about 94 surrogates of tube-fed nursing home patients in Boston and Ottawa found that the most common perceived reasons for tube feeding were to prolong life (84%) and prevent aspiration (67%), indicating ongoing need for greater informed decision-making on these issues.18 In other data from the surrogates in this cohort, we found that most surrogates reported having had discussions with health care providers about what life would be like with and without a feeding tube before tube placement.4 Despite these discussions, expectations for improvement were high and persistent, with low levels of regret in open-ended questions. Our study had several limitations. All but 18 of our patients could not report their own preferences, because they were enrolled after illnesses that interfered with their ability to participate in the interviews. We included patients in our study after the feeding tube was inserted; we cannot compare this group with expectations and outcomes of those who considered but refused tube feeding. In addition, relatively high death rates after the procedure create high dropout over time, limiting comparison of expectations data at 6 months. Alternatively, longer follow-up of the small number of survivors might lead to more perspective on patient outcomes. Our focus in this study was the association of surrogate expectations at the time of initiation of gastric
527.e15 tube feeding with patient outcomes assessed over time. While we describe the outcomes after feeding tube insertion, we cannot use these data to compare the outcomes of gastric tube feeding with possible outcomes after other feeding methods such as assisted feeding or naso-gastric tube feeding. The strengths of our study include the prospective study design and primary data collection in both academic and community hospitals, as well as examination of patient functional status and family preferences in addition to survival. Given our finding of high surrogate expectations and their lack of substantial decrease in expectations over time, we believe that physicians and other caregivers should conduct more detailed and standardized discussion with the families of patients who are about to receive gastric feeding tubes. The patient’s overall prognosis should be explicitly addressed, and expectations such as those from our study should be elicited. The expectations should be addressed one by one, so that families may make informed decisions. For example, 24% of patients with feeding tubes have respiratory infection in the 3 months following feeding tube insertion, likely related to aspiration. Severely dependent patients are likely to remain dependent, and functional dependency and advanced age at baseline are predictive of worse outcomes. We can, however, predict that patients who are more independent at baseline have a 40% likelihood of being alive and able to live outside a nursing home setting at 6 months. Decisions about life-sustaining treatment, and especially semi-permanent artificial nutrition, are among the most difficult in medicine. For many families, these are the most difficult decisions they will ever face. While a substantial minority of patients who receive gastric feeding tubes can resume oral nutrition over a period of months, functional status and quality of life remain poor, with a high rate of institutionalization and substantial mortality. For some patients, such enteral feeding may provide a bridge to an independent recovery, and health care providers should support them. For others, we recommend that more discussions take place with families about the true outcomes of this commonly employed procedure.
ACKNOWLEDGMENTS This work was supported by the National Institute of Nursing Research Grant: R01 NR05018 and by The Duke Endowment.
References 1. Gauderer MW, Ponsky JL. A simplified technique for constructing a tube feeding gastrostomy. Surg Gynecol Obstet. 1981;152:83-85. 2. Verrilli D, Welch HG. The impact of diagnostic testing on therapeutic interventions. JAMA. 1996;275:1189-1191. 3. Legorreta AP, Silber JH, Costantino GN, et al. Increased cholecystectomy rate after the introduction of laparoscopic cholecystectomy. JAMA. 1993;270:1429-1432.
527.e16 4. Lewis C, Cox CE, Garrett J, et al. Trends in the use of percutaneous feeding tubes in North Carolina hospitals. J Gen Intern Med. 2004; 19:1034-1038. 5. Braun UK, Rabeneck L, McCullough LB, et al. Decreasing use of percutaneous endoscopic gastrostomy tube feeding for veterans with dementia—racial differences remain. J Am Geriatr Soc. 2005;53:242-248. 6. Callahan CM, Haag KM, Weinberger M, et al. Outcomes of percutaneous endoscopic gastrostomy among older adults in a community setting. J Am Geriatr Soc. 2000;48:1048-1054. 7. Clarkston WK, Smith OJ, Walden JM. Percutaneous endoscopic gastrostomy and early mortality. South Med J. 1990;83:1433-1436. 8. Finocchiaro C, Galletti R, Rovera G, et al. Percutaneous endoscopic gastrostomy: a long-term follow-up. Nutrition. 1997;13:520-523. 9. Fisman DN, Levy AR, Gifford DR, et al. Survival after percutaneous endoscopic gastrostomy among older residents of Quebec. J Am Geriatr Soc. 1999;47:349-353. 10. Kaw M, Sekas G. Long-term follow-up of consequences of percutaneous endoscopic gastrostomy (PEG) tubes in nursing home patients. Dig Dis Sci. 1994;39:738-743. 11. Finucane TE, Christmas C, Travis K. Tube feeding in patients with advanced dementia: a review of the evidence. JAMA. 1999;282:1365-1370.
The American Journal of Medicine, Vol 119, No 6, June 2006 12. Sanders DS, Carter MJ, D’Silva J, et al. Survival analysis in percutaneous endoscopic gastrostomy feeding: a worse outcome in patients with dementia. Am J Gastroenterol. 2000;95:1472-1475. 13. Angus F, Burakoff R. The percutaneous endoscopic gastrostomy tube. medical and ethical issues in placement. Am J Gastroenterol. 2003; 98:272-277. 14. Katz S, Akpom CA. 12. Index of ADL. Med Care. 1976;14(5 Suppl): 116-118. 15. D’Hoore W, Bouckaert A, Tilquin C. Practical considerations on the use of the Charlson comorbidity index with administrative data bases. J Clin Epidemiol. 1996;49:1429-1433. 16. Mitchell SL, Teno JM, Roy J, et al. Clinical and organizational factors associated with feeding tube use among nursing home residents with advanced cognitive impairment. JAMA. 2003;290:73-80. 17. Rabeneck L, Wray NP, Petersen NJ. Long-term outcomes of patients receiving percutaneous endoscopic gastrostomy tubes. J Gen Intern Med. 1996;11:287-293. 18. Mitchell SL, Berkowitz RE, Lawson FM, et al. A cross-national survey of tube-feeding decisions in cognitively impaired older persons. J Am Geriatr Soc. 2000;48:391-397.
CLINICAL RESEARCH STUDY
Expectations and Outcomes of Gastric Feeding Tubes Timothy S. Carey, MD, MPH,a,b Laura Hanson, MD, MPH,e Joanne M. Garrett, PhD,b Carmen Lewis, MD, MPH,b Nancy Phifer, MD,c Christopher E. Cox, MD, MPH,d Anne Jackman, MSWa a
The Cecil G. Sheps Center for Health Services Research and bDivision of General Internal Medicine and Clinical Epidemiology, University of North Carolina at Chapel Hill; cMoses Cone Hospital, Greensboro, NC; dDivision of Pulmonary and Critical Care Medicine, Duke University Medical Center, Durham, NC; eDivision of Geriatrics, University of North Carolina at Chapel Hill. ABSTRACT PURPOSE: To compare expected outcomes with actual outcomes from tube feeding in adult patients. SUBJECTS AND METHODS: This prospective cohort study was conducted in two North Carolina hospitals. Surrogates were interviewed shortly after feeding tube insertion and at 3- and 6-month follow-up; chart abstraction and death certificate review also were carried out. Participants were surrogate decisionmakers for consecutive adult patients who received new feeding tubes. RESULTS: There were 288 patients with surrogate decision-makers enrolled. Mean age was 65 years; 30% had a primary diagnosis of stroke, 16% neurodegenerative disorder, 20% head and neck cancer, and 30% other diagnoses. At 3 months, 21% of patients had died, and 6-month mortality was 30%. At 3 months, 38% of survivors were residing in a nursing home, and 27% had the feeding tube removed. Patients were impaired in most activities of daily living (ADLs) with little change over time. Medical complications were common: 25% of patients had decubitus ulcers at 3 months, and 24% had at least one episode of pneumonia. Perceived global quality of life was poor at 4.6 (on a 0-10 scale) at baseline, and surrogates anticipated this would improve to 8.0 with tube feeding. Family surrogates’ expectations for improvement from the feeding tube were very high at baseline and remained so at 3 and 6 months. CONCLUSIONS: Families’ high expectations of benefit from tube feeding are in contrast to clinical outcomes. Providers and families need better information about the outcomes of this common procedure. © 2006 Elsevier Inc. All rights reserved. KEYWORDS: Gastric feeding tube; End-of-life care
Gastric feeding tube insertion allows long-term enteral feeding for patients who are unable to hydrate or nourish themselves. Use of the procedure has increased substantially over the past 15 years since the introduction of percutaneous methods of tube insertion.1,2 The ease of insertion with percutaneous techniques may be contributing to increased use. This phenomenon may be analogous to the increased cholecystectomy rate after the introduction of laproscopic cholecystectomy in the late 1980s.3 In North Carolina, we found that the rate of feeding tube placement approximately doubled over an 11-year period.4 One recent report from the Requests for reprints should be addressed to Timothy S. Carey, MD, MPH, The Cecil G. Sheps Center for Health Services Research, 725 Airport Road, CB#7590, The University of North Carolina at Chapel Hill, Chapel Hill, NC 27599-7590. E-mail address: [email protected].
0002-9343/$ -see front matter © 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.amjmed.2005.11.021
Veterans Administration did indicate a recent decrease in feeding tube use among some populations.5 This utilization has occurred despite controversy about both the benefit and the ethics of the procedure. Several studies have indicated that patients who receive feeding tubes have a 50% 1-year mortality rate.6-10 Some authors recommend new decisionmaking approaches to tube feeding to discourage use of this procedure in patients with advanced dementia or terminal illness.10-13 One explanation for the continuing increase in feeding tube utilization is that patients and families who make decisions for cognitively impaired patients may have expectations for tube feeding that are not based on information about outcomes data, leading to an over-estimation of the prognosis for both survival and functional recovery.
527.e12
METHODS
The American Journal of Medicine, Vol 119, No 6, June 2006
nia; living longer; more freedom; better health; fewer coughing episodes; and greater enjoyment of food (included Patients who received an initial gastric feeding tube were as a distracter to assess global optimism). Higher scores recruited at two North Carolina Hospitals: a universityindicated higher expectations. We dichotomized the 4-level based 608-bed tertiary care hospital and a 500-bed commuresponse, which ranged from strong agreement to strong nity hospital about 45 miles away. Patients were identified disagreement for each question by daily checks of inpatient gasinto agree versus disagree; analytrointestinal procedures, intervenses using the numerical 4-point tional radiology schedules and CLINICAL SIGNIFICANCE scale gave similar results. We chart review. Inclusion criteria asked surrogates at 3 and 6 were: age ⬎17 years and ability of ● The families of patients who receive gasmonths whether their expectations patient or surrogate to speak Entric feeding tubes have very high expechad been met. We operationalized glish. Exclusion criteria were: tations about longevity and improved this with the following question: non-ENT (ear, nose and throat) functioning after feeding tube insertion. “We asked you about your expecmetastatic malignancy, major trauma, tations when it was decided that ● These high expectations persist over a noncompetent patient with no suryour [mother/father, etc.] would 6-month period in spite of an overall rogate decision-maker, ward of have a gastrostomy feeding tube. the state or patient death before lack of clinical improvement in recipiWe would now like to know how completion of consent. Early ents of gastric feeding tubes, with overyour experience has been after deaths and metastatic cancer were all high levels of dependence in activiyour [mother/father, etc.] had this excluded because our aims were ties of daily living, and 30% mortality. procedure . . .” Surrogates then to examine the medium-term (3answered identical items about ex● Over a 6-month period of observation to 6-month) functional outcomes pectation of benefit at this time. after feeding tube insertion, about one and fulfillment of expectations Estimated probability of surwith tube feeding. We excluded third of patients were able to resume vival was obtained from family trauma patients because they often nourishment without a feeding tube. surrogates at baseline. Functional have a very good prognosis, and ● Clinicians need to develop improved status was reported by surrogates the feeding tube is transient. To methods of communicating with famiat baseline and at 3 and 6 months ensure a uniform source of data, using a modification of the Katz lies of these seriously ill patients in we attempted to interview a surroADL (Activities of Daily Living) order to convey realistic expectations gate in all cases. We initially con14 scale. We excluded eating from tacted the family member who and encourage appropriate shared the ADL scale because, by definisigned the procedure consent for decision-making. tion, tube-fed patients are unable the feeding tube, confirming that to eat independently. Baseline they were the family member data were not fed back to respon“most familiar with the patient’s dents at 3- and 6-month interviews. Information also was care.” Most interviews were conducted several days to 2 collected on other health outcomes such as nursing home weeks after feeding tube insertion. residence, feeding tube removal, incidence of pneumonia, Follow-up interviews were conducted at 3 and 6 months. decubitus ulcers, and use of restraints. Surrogates were If the patient had died before the follow-up interviews, a asked about these conditions for the 3 months before feedmodified surrogate interview was conducted approximately ing tube insertion, as well as in the interval to each fol1 month after the patient’s death. Family surrogates of low-up interview. Most of the baseline surveys were conpatients who died before the 3-month interview were not ducted in person; follow-up interviews were by telephone. contacted at 6 months. If the feeding tube had been removed Analyses were performed using STATA 8.0 (StataCorp, before the time of the 3-month interview, the 6-month College Station, Tex). interview was redundant and therefore not conducted. MorWe measured outcomes at both 3 and 6 months, includtality was ascertained through surrogate interview and ing survival, ADL score, residence in a nursing home, and search of the North Carolina Death Index. feeding tube removal. We compared several factors (eg, The content of the interviews included an index of items patient age, functional status, race, diagnosis) with each of reflecting possible expectations for improvement from the these outcomes using t tests for continuous variables and gastrostomy tube that was generated from review of the Pearson chi-squared for categorical comparisons. We then outcomes literature and from open-ended interviews with developed multivariable logistic regression models to deterexperienced physicians. The final index was composed of mine which set of factors was predictive of outcomes, such the sum of 14 items reflecting expectations for: improved as residence in a nursing home or mortality at 3 and 6 nutrition; being more comfortable; less pain; fewer probmonths, and removal of the feeding tube at each survey lems eating; greater independence; not feeling hungry or thirsty; quality of life; having a decreased risk of pneumointerval. Finally, we performed survival analysis using a
Carey et al
Gastric Feeding Tubes
Cox proportional hazard model. The University of North Carolina Committee on the Protection of Human Subjects approved the research.
RESULTS During enrollment, 722 patients had gastrostomy, and 306 of these were ineligible. The main reasons for exclusion were metastatic cancer (77), diagnosis of trauma (44), repeat feeding tube procedure (34), and ward of the state (13); 29 patients died before initial contact, with the remainder miscellaneous reasons. Of the 416 eligible patients, we successfully enrolled surrogate decision-makers for 288 patients with new gastrostomy feeding tubes (69% enrollment of eligibles). Among eligibles who were not enrolled, the main reasons for nonenrollment were: 35 refusals; in 15 cases the physician thought the patient was not appropriate; and in 59 cases we were unable to contact the surrogate respondent despite repeated attempts . The surrogates were generally close relatives, with 38% a spouse and 37% an adult child. Patient characteristics and outcomes were similar between the two clinical sites. We had complete baseline information from surrogates on 288 patients. We completed 228 surrogate 3-month interviews; 51 surrogates refused or could not be contacted despite repeated attempts. Six patients died shortly after consent; 3 patients had their feeding tubes removed shortly after consent. For these 9 patients, the baseline survey provided information on the total experience of the patient and surrogate with tube feeding. The completion rate among those eligible for the 3-month interview was 82%. At 6 months, a total of 107 patients did not require further follow-up because they had either had the feeding tube removed before the 3 month interview (n ⫽ 61) or had died before the 3 month interview (n ⫽ 55). Of the 121 patients eligible for the 6-month interview, 39 surrogates could not be contacted, for a completion rate of eligibles of 68%. Ascertainment of mortality was complete because we used the North Carolina Death Index. We completed 82 surrogate 6-month interviews. Given the high numbers of patients who died, had feeding tubes removed, or who were lost to follow-up, the generalizability of the 6-month follow-up is somewhat limited. Demographic and clinical characteristics of patients lost to follow-up did not differ significantly from surviving patients retained in the cohort. The baseline characteristics of the patients are in Table 1. The mean age was 65.8 years; 39% were aged over 74 years, and 41% were under 65. The age distribution within diagnostic categories did vary, with 59% of ENT cancer patients under age 65 and only 29% of patients with neurodegenerative disorders under that age. Patients were 54% men, and one third were black. Half had income ⬍$10, 000 per year. The underlying clinical conditions were varied, with 30% having stroke; 16% a neurodegenerative disease, including dementia; 20% head and neck cancer; and 34% a wide variety of other serious medical conditions. The mean Charlson comorbidity score was 2.7 (SD ⫽ 1.7), indicating
527.e13 Table 1
Baseline Characteristics (n ⫽ 288)
Age in years (SD) Male Race White Black Other Education ⬍ HS Income ⬍ $10 000 Underlying diagnosis Stroke Neurodegenerative ENT cancer Other ADL (SD) Nursing home residence Before procedure Decubitus ulcers in the 2 weeks before tube insertion Quality of life (SD)
65.8 (16.6) 54% 61% 37% 2% 36% 50% 30% 16% 20% 34% 11.8 (3.75) 36% 23% 4.6 (2.31)
the presence of multiple comorbid illnesses.15 When we asked the surrogate to rate the patient’s baseline quality of life on a 0-10 scale, it was relatively low, at 4.6. The patients were partially or totally dependent in all 5 activities of daily living (toileting, continence, walking, dressing, bathing) at the time of the baseline interview. When we assigned 1-3 points to levels of independence in activities of daily living (1 ⫽ independent, 2 ⫽ needing assistance to accomplish the task, 3 ⫽ completely dependent) and then summed them (range 5-15), the mean baseline ADL score was 11.8, indicating a severe level of dependency. Thirty-six percent had resided in a nursing home before insertion of the feeding tube.
Expectations of Benefit from Feeding Tube Placement The surrogate responses to each of the questions on expectations are shown in Table 2. At baseline, these expectations were very high (mean ⫽ 3.23 on a 4-point scale). For example, 79% “agreeing” or “agreeing a lot” that the patient would be more comfortable after the gastrostomy; 56% expected to have greater independence; 87% expected to have better quality of life; 66% expected to have decreased risk of pneumonia; and 93% expect the patient’s overall health to be better. Responses to experience with expected benefits remained high at 3 months (mean ⫽ 3.06), with little change from baseline. Among surrogates, expectation of benefit remained similarly high at 6 months for the minority of patients who survived and still had a feeding tube in place. Surrogates of decedents did have lower assessments, but excluding their responses did not significantly change the overall rating of expectations (data not shown). When we queried the surrogates at baseline about the life expectancy of the patient, 61% thought there was a ⬎90%
527.e14
The American Journal of Medicine, Vol 119, No 6, June 2006
Table 2 Surrogate Expectations about Benefit of Gastrostomy Feeding (% Agree or Strongly Agree with Each Statement) Expectation
Baseline n ⫽ 288
3 month n ⫽ 228
a. Improvement in nutrition b. Be more comfortable c. Have less pain d. Fewer problems eating e. Have greater independence f. Not to feel hungry/thirsty g. Better quality of life h. Decreased risk of pneumonia i. Live longer j. Enjoy food k. Have more freedom l. Health to be better m. Fewer coughing episodes n. Fewer choking episodes
96 (96)* 79 (79) 64 (65) 83 (83) 56 (55) 69 (70) 87 (87) 66 (66) 90 (90) 39 (40) 54 (54) 93 (93) 67 (68) 77 (79)
87 71 60 78 54 67 80 65 91 29 48 83 62 74
*The percentages in parentheses are the baseline values for those subjects interviewed at 3 months (n ⫽ 228).
probability that the patient would live more than 1 year. At 3 months, 67% of surrogates for survivors thought the patient would survive for at least 1 more year, and at 6 months, 61% did so. The surrogate-assessed mean global quality of life (0-10) did change slightly from 4.6 at baseline to 6.0 when rated at 3 months for surviving patients. When we asked surrogates at the baseline interview what they would expect their family member’s quality of life to be at 6 months, they estimated a high quality of 8.0.
Health Outcomes At 3 months, 21% of the patients had died, and 6-month mortality was 30%; 82 patients survived beyond 3 months and had complete interviews. The mortality rate was very similar across the four diagnostic strata (stroke, ENT cancer, neurodegenerative disease, and other serious medical conditions) after we controlled for baseline age and functional status as assessed by ADLs. In a Cox proportional hazard model we examined age, race, sex, educational status, presence of an ENT cancer diagnosis, nursing home status at baseline, ability to eat somewhat at baseline, and ADLs at baseline. Only age and baseline ADL predicted mortality over the 6 months. For each decade of increased age, the hazard ratio increased by 17%. For each increased point in ADL from 5-15 (indicating increasing dependence), the hazard ratio increased 9%. In a logistic regression model examining 6-month mortality and the above independent variables, we estimated that a 30-year-old had a 15% probability of dying over the 6-month period, and a 90-year-old had a 40% probability. An individual with a baseline ADL score of 5 (least dependent score) had an estimated 19% 6-month mortality, which rose to 33% with a score of 15. The baseline functional status score was worse (higher) among those who eventu-
ally died, compared with those alive at 3 months (14.1 vs 9.8). Although ENT cancer patients had similarly high mortality during follow-up, their baseline and follow-up functional status differed from other diagnostic groups. ENT cancer patients had a much better ADL score of 6.6, compared with 12.5 for the other diagnostic strata. Only 5% of ENT cancer patients resided in a nursing home before insertion of the feeding tube, compared with 42% of patients in the other strata. At 3 months, 77% of ENT cancer patients were independent in two or more ADLs, compared with 26% of patients in other strata. Thirty-two percent of subjects had the feeding tube removed between baseline and 3 months, and 63% took at least some nutrition by mouth, although families often told us that this feeding was without medical approval. In multivariable logistic regression analysis we found that significant baseline variables predictive of feeding tube removal at 3 months were younger age (34% at age 50, 15% at age 90), white race (31% white race, 18% black), male sex (34% men, 18% women), diagnosis other than ENT cancer (only 9% of cancer patients had tubes removed at 3 months), and less dependent baseline ADL (probability decreased from 41% at an ADL of 5 to 19% for an ADL of 15). The incidence of pneumonia was 24% between baseline and 3 months, and 12% among surviving patients from 3 to 6 months (Table 3). One in four patients had decubitus ulcers at the 3-month interview. We did ask the surrogates at baseline about pneumonia and decubiti in the 3 months before feeding tube insertion, and the prevalence before feeding tube insertion was similar to the 3 months after insertion.
DISCUSSION This multi-site, prospective cohort study found that families have high expectations for improvements in both longevity
Table 3
Other Health Outcomes
Mean ADL (range 515, higher more dependent) Fully independent in 2 or more ADLs (%) Residence in a nursing home (%) Feeding tube removed and patient alive (%) One or more episodes of pneumonia (%) Decubitus ulcers (%) Restraint use (%)
Baseline n ⫽ 288
3 months n ⫽ 228
6 months* n ⫽ 82
11.8
10.8
10.3
20
31
34
36
38
37
–
32
23
11
24
12
23 30
25 27
25 14
*Subset of patients whose feeding tube still in place at 3 months.
Carey et al
Gastric Feeding Tubes
and functioning when interviewed after the placement of a gastrostomy feeding tube. The mortality in this study was high (30% over 6 months) with little functional improvement among survivors. The mortality rate was similar across the 4 diagnostic groups after adjustment for age and baseline functional status. The high baseline expectations of benefit did not significantly change over time. Feeding tube use remains very common, with a recent study by Mitchell et al demonstrating that over one third of severely cognitively impaired residents in US nursing homes have feeding tubes.16 In another secondary data analysis, Rabeneck and colleagues examined outcomes of patients receiving gastrostomy through VA hospitals.17 They found a median survival of 8 months, with increasing mortality with greater age and malignant diagnosis. Our outcome analysis is most similar to a community-based study conducted by Callahan and colleagues in a single-site study of 150 patients with feeding gastrostomy in Indiana.6 He found a 1-year mortality rate of nearly 50% and similar rates of decubitus ulcers to those found in our study, but did not examine surrogate expectations of benefit relative to these outcomes. Why are families so optimistic about the feeding tube, and how does this optimism persist in the face of little functional improvement over time? Most of these patients had long hospital courses and chronic illness—the feeding tube often represented one in a long line of procedures, though one that was perhaps appealing because it was associated with nutrition. It is also possible that families were not adequately informed about what to expect after tube insertion. A cross-sectional survey of about 94 surrogates of tube-fed nursing home patients in Boston and Ottawa found that the most common perceived reasons for tube feeding were to prolong life (84%) and prevent aspiration (67%), indicating ongoing need for greater informed decision-making on these issues.18 In other data from the surrogates in this cohort, we found that most surrogates reported having had discussions with health care providers about what life would be like with and without a feeding tube before tube placement.4 Despite these discussions, expectations for improvement were high and persistent, with low levels of regret in open-ended questions. Our study had several limitations. All but 18 of our patients could not report their own preferences, because they were enrolled after illnesses that interfered with their ability to participate in the interviews. We included patients in our study after the feeding tube was inserted; we cannot compare this group with expectations and outcomes of those who considered but refused tube feeding. In addition, relatively high death rates after the procedure create high dropout over time, limiting comparison of expectations data at 6 months. Alternatively, longer follow-up of the small number of survivors might lead to more perspective on patient outcomes. Our focus in this study was the association of surrogate expectations at the time of initiation of gastric
527.e15 tube feeding with patient outcomes assessed over time. While we describe the outcomes after feeding tube insertion, we cannot use these data to compare the outcomes of gastric tube feeding with possible outcomes after other feeding methods such as assisted feeding or naso-gastric tube feeding. The strengths of our study include the prospective study design and primary data collection in both academic and community hospitals, as well as examination of patient functional status and family preferences in addition to survival. Given our finding of high surrogate expectations and their lack of substantial decrease in expectations over time, we believe that physicians and other caregivers should conduct more detailed and standardized discussion with the families of patients who are about to receive gastric feeding tubes. The patient’s overall prognosis should be explicitly addressed, and expectations such as those from our study should be elicited. The expectations should be addressed one by one, so that families may make informed decisions. For example, 24% of patients with feeding tubes have respiratory infection in the 3 months following feeding tube insertion, likely related to aspiration. Severely dependent patients are likely to remain dependent, and functional dependency and advanced age at baseline are predictive of worse outcomes. We can, however, predict that patients who are more independent at baseline have a 40% likelihood of being alive and able to live outside a nursing home setting at 6 months. Decisions about life-sustaining treatment, and especially semi-permanent artificial nutrition, are among the most difficult in medicine. For many families, these are the most difficult decisions they will ever face. While a substantial minority of patients who receive gastric feeding tubes can resume oral nutrition over a period of months, functional status and quality of life remain poor, with a high rate of institutionalization and substantial mortality. For some patients, such enteral feeding may provide a bridge to an independent recovery, and health care providers should support them. For others, we recommend that more discussions take place with families about the true outcomes of this commonly employed procedure.
ACKNOWLEDGMENTS This work was supported by the National Institute of Nursing Research Grant: R01 NR05018 and by The Duke Endowment.
References 1. Gauderer MW, Ponsky JL. A simplified technique for constructing a tube feeding gastrostomy. Surg Gynecol Obstet. 1981;152:83-85. 2. Verrilli D, Welch HG. The impact of diagnostic testing on therapeutic interventions. JAMA. 1996;275:1189-1191. 3. Legorreta AP, Silber JH, Costantino GN, et al. Increased cholecystectomy rate after the introduction of laparoscopic cholecystectomy. JAMA. 1993;270:1429-1432.
527.e16 4. Lewis C, Cox CE, Garrett J, et al. Trends in the use of percutaneous feeding tubes in North Carolina hospitals. J Gen Intern Med. 2004; 19:1034-1038. 5. Braun UK, Rabeneck L, McCullough LB, et al. Decreasing use of percutaneous endoscopic gastrostomy tube feeding for veterans with dementia—racial differences remain. J Am Geriatr Soc. 2005;53:242-248. 6. Callahan CM, Haag KM, Weinberger M, et al. Outcomes of percutaneous endoscopic gastrostomy among older adults in a community setting. J Am Geriatr Soc. 2000;48:1048-1054. 7. Clarkston WK, Smith OJ, Walden JM. Percutaneous endoscopic gastrostomy and early mortality. South Med J. 1990;83:1433-1436. 8. Finocchiaro C, Galletti R, Rovera G, et al. Percutaneous endoscopic gastrostomy: a long-term follow-up. Nutrition. 1997;13:520-523. 9. Fisman DN, Levy AR, Gifford DR, et al. Survival after percutaneous endoscopic gastrostomy among older residents of Quebec. J Am Geriatr Soc. 1999;47:349-353. 10. Kaw M, Sekas G. Long-term follow-up of consequences of percutaneous endoscopic gastrostomy (PEG) tubes in nursing home patients. Dig Dis Sci. 1994;39:738-743. 11. Finucane TE, Christmas C, Travis K. Tube feeding in patients with advanced dementia: a review of the evidence. JAMA. 1999;282:1365-1370.
The American Journal of Medicine, Vol 119, No 6, June 2006 12. Sanders DS, Carter MJ, D’Silva J, et al. Survival analysis in percutaneous endoscopic gastrostomy feeding: a worse outcome in patients with dementia. Am J Gastroenterol. 2000;95:1472-1475. 13. Angus F, Burakoff R. The percutaneous endoscopic gastrostomy tube. medical and ethical issues in placement. Am J Gastroenterol. 2003; 98:272-277. 14. Katz S, Akpom CA. 12. Index of ADL. Med Care. 1976;14(5 Suppl): 116-118. 15. D’Hoore W, Bouckaert A, Tilquin C. Practical considerations on the use of the Charlson comorbidity index with administrative data bases. J Clin Epidemiol. 1996;49:1429-1433. 16. Mitchell SL, Teno JM, Roy J, et al. Clinical and organizational factors associated with feeding tube use among nursing home residents with advanced cognitive impairment. JAMA. 2003;290:73-80. 17. Rabeneck L, Wray NP, Petersen NJ. Long-term outcomes of patients receiving percutaneous endoscopic gastrostomy tubes. J Gen Intern Med. 1996;11:287-293. 18. Mitchell SL, Berkowitz RE, Lawson FM, et al. A cross-national survey of tube-feeding decisions in cognitively impaired older persons. J Am Geriatr Soc. 2000;48:391-397.