Discomfort in Nursing Home Patients With Severe Dementia in Whom Artificial Nutrition and Hydration Is Forgone

Background  While there is debate about whether it may be better to forgo than to initiate artificial nutrition and hydration (ANH) in nursing home patients with severe dementia, the consequences of forgoing ANH in these patients, in particular their discomfort, have not yet been investigated.

Methods  In this prospective, longitudinal, observational study of 178 patients in Dutch nursing homes, discomfort was measured at all measurement times according to the observational Discomfort Scale–Dementia of Alzheimer Type. Furthermore, at all measurement times, plausible determinants of discomfort were registered. Data were analyzed with the statistical technique of generalized estimated equations.

Results  Decisions to forgo ANH were made most often in severely demented, female patients with an acute illness as the most important diagnosis at that time. The mean level of discomfort was highest at the time of the decision and decreased in the days thereafter. There were substantial differences in level of discomfort between patients. Dyspnea, restlessness, and physicians’ observations of pain and dehydration were associated with higher levels of discomfort. Furthermore, patients who were awake had higher levels of observed discomfort than patients who were asleep.

Conclusions  Forgoing ANH in patients with severe dementia who scarcely or no longer eat or drink seems, in general, not to be associated with high levels of discomfort. The individual differences emphasize the need for constant attention for possible discomfort.

Many patients with dementia develop difficulties with eating and drinking during the course of the disease.^1-7 When the intake of food and fluids becomes insufficient, a decision has to be made whether to start administration of artificial nutrition and/or hydration (ANH). Important issues often debated in relation to this decision are whether ANH causes discomfort and whether death due to dehydration is painful.

In recent years several reviews on tube feeding in patients with severe dementia have been published.^8-12 All authors conclude that feeding tubes may cause more suffering than comfort and can also cause burdensome complications such as aspiration and pressure sores, and patients often have to be restrained. Based on the findings that tube feeding might not prolong survival in patients with dementia,^8-16 the opinion that it may be better to forgo tube feeding in these patients seems to gain ground. However, there is lack of empirical data.

Discomfort in patients with severe dementia is difficult to measure, since most of these patients lose the ability to verbalize their feelings. However, a few instruments have been developed specifically to assess pain and/or discomfort in patients with dementia based on facial expressions, vocalizations, and muscle tension.^17-23

We conducted a prospective observational study in nursing home patients with dementia who scarcely or no longer ate or drank and for whom the decision was made to start or forgo ANH. We studied the characteristics of patients, the level and course of discomfort, and factors that are associated with discomfort.

This was a prospective observational study based on written questionnaires completed by nursing home physicians (NHPs). The managers of 70 nursing homes in 3 regions of the Netherlands were invited to participate in the study. In total, 39 nursing homes participated (with a total of more than 4000 patients with dementia). Reasons for nonparticipation most frequently mentioned were heavy workload and/or shortage of personnel (n = 14), already participating in other studies (n = 7), and reorganization (n = 4). The inclusion period lasted for 10 months (February 2000 to November 2000). In total, 190 patients in 32 different nursing homes were included: 178 for whom ANH was forgone and 12 for whom ANH was started. In the other 7 nursing homes, no patients were included either because of the heavy workload or because no patients met the inclusion criteria during the inclusion period. The present study focuses only on the 178 cases in which ANH was forgone because the number of patients for whom it was begun was too small to draw conclusions. The age, sex, and diagnosis of the included patients when admitted to the nursing home correspond with these demographic areas of the general Dutch nursing home population with dementia.²⁴ The study protocol was approved by the medical ethics committee of the VU University Medical Center.

The NHPs were requested to include patients with dementia when those patients scarcely or no longer ate or drank and a decision was made to start or forgo ANH (percutaneous endoscopic gastrostomy tube, nasogastric tube, hypodermoclysis, or intravenous infusion). The moment of decision can depend on several factors.²⁵ The goal was to study these decision-making processes and the consequences of the decision in daily practice. We did not use a more specific inclusion criterion to avoid influencing the daily practice.

The NHPs filled out a questionnaire directly after the decision was made (baseline) and subsequently at 5 other time points (or fewer if a patient died): 2, 5, 9, and 14 days and 6 weeks after the decision was made to forgo ANH. Discomfort was measured at all measurement times according to the observational Discomfort Scale–Dementia of Alzheimer Type (DS-DAT).²⁰ Discomfort is defined by the developers of the scale as a negative emotional and/or physical state subject to variation in magnitude in response to internal or environmental conditions. The scale consists of nine 4-point items, and the summed scores range from 0 (no observed discomfort) to a maximum of 27 (high level of observed discomfort). The items are noisy breathing, negative vocalization, sad facial expression, frightened facial expression, frown, tense body language, fidgeting, content facial expression, and relaxed body language. Scores are based on frequency, intensity, and duration of the observed behavior over 5 minutes, and range from 0 (not observed) to 3 (present in high intensity and for almost the entire rating period).^20,26 To calculate summed scores, the scores of the 2 positive items are reversed.

The NHPs were instructed how to use the scale during a training session in which an instructional videotape was shown. Several studies have demonstrated acceptable reliability and validity of the scale,^20,26-30 and 1 of these studies was conducted with a patient group similar to ours.^26,28,30

At all measurement times, the presence of burdensome symptoms, average fluid intake, presence of dehydration, use of medication, and any other care received by the patient were registered. At baseline, the use of medication directly after the decision to forgo ANH was registered because treatment with most types of medication, except psychotropic medication and analgesics, was cancelled for most patients who were taking medication before the decision.

Patient characteristics (age, sex, dementia type, acute or gradual decline in food and/or fluid intake, and most important diagnosis at the time of the decision to forgo ANH) were registered at baseline (Table 1). The severity of the dementia was measured according to the Bedford Alzheimer Nursing Severity scale (BANS-S),³¹ which consists of seven 4-point items rating cognitive deficits, functional deficits, and occurrence of pathologic symptoms: (1-3) level of (in)dependency with dressing, eating, and walking; (4) (ir)regular sleep-wake cycle; (5) (in)ability to speak; (6) flexibility of muscles; and (7) maintenance of eye contact. Summed scores range from 7 (no impairment) to 28 (complete impairment). The BANS-S assesses the severity of dementia not only in patients severely affected by Alzheimer disease but also in patients with other types of dementia. In patients with Alzheimer disease, an average BANS-S score of 19.3 has been associated with most patients (92%) who score zero on the Mini–Mental State Examination, which means that they are severely demented. At baseline we retrospectively asked the patient’s condition 2 weeks prior to the time of the decision making to exclude temporary effects of diminished functioning due to an acute illness state.

The NHPs assessed presence of burdensome symptoms and dehydration on a 5-point scale, and level of consciousness during the DS-DAT on a 6-point scale. Both parameters were dichotomized for the analysis (burdensome symptom or dehydration present vs not present; patient awake vs asleep). All data are based on observations made by the attending NHPs.

Because the outcome variable (discomfort) is a repeated measure within the same person, the statistical technique of generalized estimated equations was used to account for correlated measures. The interpretation of the coefficient (B) is comparable to the regression coefficient in cross-sectional linear regression analysis. While the cross-sectional regression analysis is limited to differences between persons, the longitudinal regression technique is a combination of the differences between persons and the changes over time within persons.^32,33

To determine which variables were to be included in the final model for discomfort, the generalized estimated equations analysis was performed in 3 steps. First, all plausible determinants were analyzed separately (bivariate) to obtain information about their influence on discomfort without influence from other variables. Subsequently, all these variables, except symptoms and types of medication that were present in fewer than 10% of the patients, were entered into a model (multivariate). Symptoms, use of medication, and other treatments were entered as a group into the model, and the variables with the lowest z scores were eliminated 1 by 1 until all remaining variables were significant (backward). All variables that remained significant with P<.10 in these analyses composed the final model for discomfort. Finally, the level of consciousness during the observation period for the DS-DAT was entered into the model to control for its possible influence on the DS-DAT score.

At baseline, the patients had an average age of 85 years and were mostly women (79%). Almost half of the patients had Alzheimer disease, and most were severely demented. In almost half of the patients (43%) the decline in food and/or fluid intake started suddenly, and almost two thirds (62%) had an acute illness at the time of the decision making, such as a stroke, lower respiratory tract infection, or urinary tract infection. More than half of the patients (59%) died within 1 week after the decision was made to forgo ANH. Female patients, patients who were less demented, and patients without intercurrent disease survived longer (Table 1).

Most symptoms were present in a minority of the patients, with the exception of apathy, which was present in 79% of all patients (Table 2). The average number of symptoms was 2.1 at baseline, and this decreased during the follow-up measurements. At baseline, almost one third of the patients were taking opioids and/or other analgesic medication, and 1 in 5 patients was taking benzodiazepines or other psychotropic medication. At baseline and in the first few days after the decision was made to forgo ANH, about two thirds received mouth care.

Figure 1 shows the average fluid intake of patients, classified according to survival. Patients who survived for 2 weeks had a much higher fluid intake at all measurement times except at baseline than the patients who died within 2 weeks after the decision was made to forgo ANH. Based on clinical observations made by the NHPs, two thirds of the patients were dehydrated at the time of the decision.

Table 2 lists the scores on the discomfort scale at all measurement times. The level of discomfort was highest at baseline (mean [SD], 8.11 [5.63]) but decreased in the first 5 days after the decision was made to forgo ANH (mean [SD] score, 5.63 [4.65] at day 5) and then increased again after 5 days.

Figure 2 shows the course of the discomfort scores, classified according to survival, after the decision was made to forgo ANH. The level of discomfort decreased by at least 1 point at day 2 in all groups. In patients who died within 2 weeks after baseline, the average level of discomfort decreased until they died. In patients who survived longer than 2 weeks after baseline, the average level of discomfort decreased in the first 5 days after the decision was made and increased again after day 5, but it did not reach baseline level (mean score of 8.3 at baseline and 6.8 at day 42). In patients in whom discomfort was measured latest, 1 day before death (n = 63), a significant decrease in level of discomfort was found (P = .02) (data not shown).

Most variables did not have a significant longitudinal influence on discomfort, either in the bivariate or in the multivariate analysis. However, vomiting, myoclonus, and seizures did have a significant influence on discomfort in the bivariate analyses (DS-DAT score differences when present, 2.8, 2.4, and −1.2, respectively), but these symptoms were not included in the multivariate analysis because they were present in fewer than 10% of the patients. The average number of symptoms (maximum of 10) was also significantly related to discomfort in the bivariate analysis: discomfort increased with an increase in the total number of symptoms (1 point per extra symptom) (B, 1.02 [95% CI, 0.57-1.47]; P<.001) (data not shown). However, this was not included in the multivariate analysis because the longitudinal influence of each symptom was studied separately.

Table 3 lists the variables that were significant in the bivariate analyses and were included in the final model for discomfort. On average, men had a higher level of discomfort (a score of 1.7 points higher on the DS-DAT) than women. Furthermore, patients who had symptoms of dyspnea and/or restlessness had a substantially higher level of discomfort than patients who did not have these symptoms (DS-DAT scores 2.6 and 3.3 points higher, respectively). Also patients in whom the NHP observed pain or dehydration had a higher level of discomfort (1.5 and 1.3 points higher, respectively) than patients who had no observed pain and dehydration. Finally, the level of consciousness had a substantial influence on the DS-DAT score. Patients who were awake during the DS-DAT had a score that was almost 3 points higher than the scores of patients who were asleep.

The effect of fluid intake, use of benzodiazepines and other psychotropic medication, and provision of mouth care did not have a significant influence on discomfort in the final model.

In Dutch nursing homes, for patients with severe dementia, decisions to forgo ANH are made much more often than decisions to start ANH. This decision is most often made for female patients with severe dementia and with an acute illness. Most patients die relatively soon after the decision. This can imply that the patients were already in the dying process. The fact that in a quarter of the patients all medication treatment except that with opioids was canceled directly after the decision indicates that NHPs expected that the patient would die soon or that a more “general” decision to withhold possible life-sustaining treatment (including ANH) was made.

The DS-DAT has been used in several research projects.^26,30,34 In comparison with the findings of these studies, the patients in the present study, on average, did not have a high level of discomfort. Patients with dementia in whom ANH is forgone have lower levels of discomfort than both (1) nursing home patients with dementia and pneumonia²⁶ and (2) patients with dementia enrolled in traditional long-term care in the United States who have no intercurrent disease. However, patients with dementia in whom ANH is forgone have higher levels of discomfort in the first 2 days after AHN is forgone than patients with dementia in special dementia care units who have no intercurrent disease. The care given in these special dementia care units emphasizes patient comfort over maximum survival and is comparable with the care given in units for patients with dementia in Dutch nursing homes.³⁴

An important finding is that the level of discomfort decreases in the days after the decision is made to forgo ANH. In patients who died within 2 weeks after the decision, the discomfort level decreased until they died. Most patients are probably asleep or comatose shortly before death, and it was found that patients who are asleep have much lower levels of discomfort, according to the DS-DAT. It is possible that the decrease in discomfort is caused more by the dying process than by forgoing ANH. However, we do not have data to support this because we do not have a sufficient reference group.

In the present study, dyspnea and restlessness were associated with higher levels of discomfort. This was also found in an earlier pneumonia study.²⁶ However, it is not surprising that these symptoms were significant in the analysis because they might be associated with specific items on the DS-DAT (noisy breathing and fidgeting). Vomiting and myoclonus were also associated with higher levels of discomfort (a difference of 2.8 and 2.4 points in bivariate analysis, respectively), but these symptoms were present in only a small number of patients. However, it is important to prevent or treat these symptoms because that will decrease the level of discomfort substantially.

Observed pain and dehydration were also associated with higher levels of discomfort. However, it is possible that the observations of pain and dehydration influenced the NHPs in scoring the DS-DAT, leading to overestimating the effect. On the other hand, the items of the DS-DAT are not identical to these variables.

Furthermore, patients who were asleep had substantially lower levels of observed discomfort than patients who were awake. It is reasonable to think that patients who are asleep or (sub)comatose have lower levels of discomfort because being comfortable may be a prerequisite for being able to sleep. On the other hand, there are also reasons to think that the DS-DAT is not equally valid for patients who are asleep and patients who are awake. The developers of the DS-DAT pointed out that some items, such as sad facial expression, cannot be assessed in a sleeping patient.²⁰ In addition, it is possible that there is an important difference between just being asleep and being (sub)comatose (due to the dying process). Unfortunately our data are not suitable for investigating these points.

Although providing mouth care was not significantly associated with a lower level of discomfort in the final model (in contrast to the bivariate analysis), specific attention should be paid to providing adequate mouth care in daily practice, especially since other studies have found a positive effect of mouth care after forgoing artificial hydration in terminally ill cancer patients and since providing mouth care is easy to implement in daily practice.^9,35,36

Finally, it is remarkable that analgesic medication (including opioids) and psychotropic medication, usually given to improve the comfort of patients, had no significant longitudinal effect on discomfort. More research is needed on this aspect, especially since psychotropic medication did have an affect in the bivariate analysis.

The DS-DAT is a well-developed measurement instrument²⁰ that has shown acceptable psychometric properties in several studies.^20,26-30 In the present study, its internal consistency was also high (0.85 at baseline). The DS-DAT seems to be an appropriate instrument to measure discomfort in patients with severe dementia. However, until now, no attention has been paid to the effect of the level of consciousness on the DS-DAT scores, and it is also not clear why patient sex influences discomfort, not in our study, nor in other studies.^26,37 Therefore, more research is needed to investigate the psychometric properties of the DS-DAT.

Furthermore, the observations in this study were all made by the attending NHPs, which might affect the objectivity of the observations. In the ideal situation, the person who scores the DS-DAT does not know the patient and is not familiar with the patient’s behavior.

Finally, the study was designed to compare a group of patients in whom ANH was forgone with a group of patients in whom ANH was started. However, the second group was too small to include in the analysis (only 12 patients), so no suitable reference group was available.

In conclusion, this is the first empirical study to investigate discomfort in patients with severe dementia for whom the decision is made to forgo ANH. Because of the observational nature of the study and lack of reference group, conclusions should be drawn carefully. However, it seems that forgoing ANH in patients with severe dementia who scarcely or no longer eat or drink is not associated with high levels of discomfort. This is in concordance with data on forgoing artificial hydration in terminal cancer patients.^{9,35,36,38,39} With regard to discomfort, forgoing ANH therefore seems to be an acceptable decision for patients with severe dementia. However, the data are based on average scores, and there are substantial individual differences. These differences emphasize the need for constant attention for possible discomfort, certainly since the recent finding of Mitchell et al⁴⁰ that patients with advanced dementia in nursing homes do not always receive optimal palliative care.

Correspondence: H. Roeline W. Pasman, PhD, VU University Medical Center, Department of Public and Occupational Health, Van der Boechorststraat 7, 1081 BT, Amsterdam, the Netherlands (hrw.pasman@vumc.nl).

Financial Disclosure: None.

Accepted for Publication: January 15, 2005.