Factors related to withholding life-sustaining treatment in hospitalized elders

Abstract

Objectives

To look for predictors in the clinical records of orders for “limitation of life sustaining treatment” (LLST) or “do not attempt resuscitation” (DNAR) in hospitalized elders and to assess the relationship between the presence of these orders and the quality of end-of-life (EOL) care.

Design

Retrospective clinical record review.

Setting

Inpatients of an inner city elderly acute care unit (EACU) in Spain.

Participants

Of 103 hospitalized patients who died in the EACU during one year, 90 dying an expected death either from acute or chronic disease were included.

Measurements

Demographic, functional, cognitive, clinical, and end-of-life (EOL) parameters. The influence of identifying closeness to death and the number of LLST suborders on the quality of EOL-management were considered simultaneously using structural equation modelling with LISREL 8.30 software.

Results

LLST and specific DNAR orders were registered in 91.1% and 83.3% of patients, respectively. Failure of acute treatment, discussions with the patient/family, recognizing the presence of common EOL symptoms, and prescribing specific symptomatic treatment were recorded in 88.9%, 93.3%, 94.4%, and 86.7% of patients, respectively. LLST-orders were more likely to be documented if there was severe functional impairment prior to admission (p<0.001), advanced organ disease criteria were met (p=0.006), or closeness to death was acknowledged in writing (p<0.001). The quality of the EOL-management was better in patients for whom there were LLST-orders (p =0.01) and written acknowledgement of closeness to death (p<0.001).

Conclusions

LLST-orders were more likely to be written in an EACU for patients with previous severe impairment, co-morbidity, or advanced disease. Written acknowledgement of closeness to death and LLST-orders were predictors of better EOL-management.

Background

In 2005 cancer was responsible for 13% of the 58 million deaths worldwide while long-term illnesses accounted for 47%. By 2030 the annual number of deaths is expected to rise to 74 million worldwide, with conditions related to organ failure and physical and cognitive frailty responsible for most of this increase (1). Up to 77% of elderly people die in institutional settings in the United Sates (53% in hospitals) and only 20% at home (2). Deaths of elderly people are more likely to occur in hospital if there has not been holistic, coordinated assistance in all phases of illness (2, 3), a finding that has led to calls for the development of systems of continuing, comprehensive care for frail elders (2).

Limitation of life sustaining treatment (LLST) is a process or group of practices involving withholding or withdrawing treatments intended to prolong survival. LLST occurs in up to 50% of all expected hospital deaths (4, 5). The EURELD-Study examined LLST-decisions in six European countries and one of the main conclusions was that a rational framework for clinical decision-making near the end of life(EOL) could lead to increased respect for patient autonomy. Use of “do not attempt resuscitation” (DNAR) and LLST orders has increased in elderly patients (6, 7, 8) since the Study to Understand Prognosis and Preferences for Outcomes and Risks of Treatment (SUPPORT), which evaluated the low frequency of discussions about resuscitation and the poor response to specific training processes (9, 10). In Spain, the use of LLST orders and adequate EOL management has been reported to be low and specific training in EOL-issues in a general hospital did not improve EOL-management of end-stage dementia patients (11, 12).

Up to 80% of patients admitted to hospitals who die do so with a DNAR order in place (9), 75% of which were made during last admission and only 7% in a prior care plan. There is an agreement that a DNAR order is licit if resuscitation efforts are likely to be futile (13). However, before the implementation of protocols or guidelines currently used in the USA or UK, LLST orders were rarely registered in clinical notes (7, 8). Day-to-day use of these instructions is still heterogeneous, even in countries in which guidelines or protocols have been established (14, 15). In practice, most of health care practitioners might have an opinion on which of their patients should have a DNAR order. However, this is often not registered in the clinical notes either because of a lack of training (in confronting the possibility of death or analyzing the elements that would affect this decision) or for reasons such as fear of “therapeutic failure” or “destroying hope” (16).

Adequately recording functional assessment and closeness to death might release acute medicine doctors from the “therapeutic contract”, help clarify the decision-making process regarding what should be done in the event of an emergency, rationalize treatment, and provide better quality EOL care (16, 17). Our hypothesis was that clinicians would be more likely to write LLST-orders if there was severe functional impairment, higher co-morbidity, or a diagnosis of an advanced disease and that there would be better EOL management when LLST-orders were made.

The objective of this study was to analyze whether a written record of LLST-orders was related to identifying patients in whom resuscitation efforts were likely to have led to poor outcomes or been futile, such as patients with severe impairment prior to admission (determined through geriatric assessment) or very advanced chronic diseases (as defined by the American National Hospice and Palliative Care Organization [NHPCO] advanced organ disease criteria (18, 19)) and to assess the relationship between the presence of these orders and the quality of EOL-care.

Methods

Data were gathered in a retrospective review of clinical records and prescription charts of all patients who died in an inner-city hospital elderly acute care unit (EACU) in Spain during one year (2004). Sociodemographic, clinical, functional, cognitive, and EOL-data were collected by three geriatricians. There is no specific protocol of current use for EOL-data collection in this setting.

Data were collected for

1. Functional and cognitive impairment tools were collected by geriatricians from a specific protocol in clinical notes. It included data about patient status both within 24 hours of admission and 15 days prior to the acute illness that led to hospitalization, to ascertain the impact of acute illness on functional/cognitive status. The Red Cross Hospital Scales for Functional (RCF) and Cognitive (RCM) Impairment have been validated in our country and population and correlated to the Barthel and Pfeiffer indexes; there are 6 grades (20), where grade 0 is normal, grade 5 is the highest impairment, and a score greater than 3 indicates severe impairment (20). Spearman’s correlation index between the RCF and Barthel Index in the EACU was 0.93 and between cognitive status and Pfeiffer’s Short Portable Mental Status Questionnaire was 0.72. Sensitivity and specificity were 0.72/0.96 for mild, 0.65/0.77 for moderate, and 0.83/0.93 for severe functional impairment (RCF-scale) and 0.81/0.90 for cognitive impairment (RCM-scale). Barthel Index (21) was also obtained from clinical notes; regarding data both of prior functional status and impairment on admission.

2. Co-morbidity. This was assessed using the Charlson Index (22).

3. Diagnosis of an advanced disease. The sample was divided in to two groups based on whether the patients met the criteria established in 1997 by the American NHPCO for advanced oncological and non-oncological organ diseases (18, 9). Even though the American NHPCO criteria for advanced diseases allow for a large margin of error in the prediction of prognosis, particularly in the elderly and even more so in patients with dementia (23), these criteria are an accepted method for defining the progression of chronic diseases that prevail in the elderly. The American NHPCO criteria were, therefore, used in this study for advanced clinical situations in which acute complications, when identified, may have indicated an irreversible failure of acute treatment. Two patients with Parkinson’s disease, Hoehn-Yahr stage-V (24), were included in the “patients with advanced disease” category as their condition was considered to be as severe as that of other patients meeting the American NHPCO criteria.

4. LLST and LLST-suborders. The presence of any of the following expressions written in clinical notes was considered to be an order for LLST: DNAR, not for hospital transfer, not for admission to the intensive care unit, or not for central line. The order “not for hospital transfer” relates to being sent to another hospital for complex care needs not available in this centre.

5. Identifying closeness to death. Any comment in clinical notes indicating recognition of closeness to death, pre-death phase or a last-days situation and the dating of this issue were noted.

6. Adequate EOL-management. Information recorded in writing about discussions of short-term-prognosis and decisions about LLST-orders, the presence of EOL-symptoms, prescribing specific symptomatic treatment, and withdrawing acute medical treatment were noted. We tried to assess adequate EOL-management using a scale whereby Level 0=inadequate management (no EOL symptoms were identified, no symptomatic relief was prescribed, no patient/family discussion of EOL-issues were held), Level 1=incomplete management (there was identification of symptoms and/or symptom treatment but no discussion with the patient/family), Level 2=failure of acute medical management identified, Level 3=Level 2 and discussion with patient/family, Level 4=Level 3 plus recognition of EOL-symptoms, and Level 5=Level 4 plus prescription of symptomatic treatment.

Two statistical theoretical models were predicted (Figure 1)

Figure 1.

Statistical Theoretical Predicted Models

1. Theoretical model using categorical variables. All result-variables were categorical. The U-Mann-Whitney test was used to compare ranges. The Chi-square test was used for qualitative variables with the Yate’s correction and Fisher’s exact test when needed. A 95% confidence interval was used for statistical significance. Statistical analyses were performed using SPSS, version 13.0 (SPSS Inc., Chicago, Illinois).

2. The influence of functional impairment prior to admission, impact on functional status of severity of the acute illness, and presence of co-morbidity on the number of LLST-suborders and early recognition of closeness to death and the importance of these in adequate EOL-management were considered simultaneously using structural equation modelling with LISREL 8.30 software (25). Not all the variables of the study suit this model, which tries to achieve parsimonious models in which redundant variables, which have high correlation coefficients, are deleted. Since the variables were not normally distributed, the Unweighted Least Squares was used as the estimation method.

Table 1 shows the polychoric correlation matrix. Variables introduced into the model were the number of days between closeness-to-death being identified and death occurring, the Barthel index score assessing functional status prior to admission, the Barthel (21) index score on admission reflecting functional impact of severity of illness, the Charlson (22) co-morbidity index, the number of LLST-suborders (how many therapies were not-for-attempt), and the level of adequacy of the EOL-management.

Table 1.

Correlation matrix to be analyzed

	Days to death	Number of LLST orders	Adequate management	Charlson index	Barthel previous to admission	Barthel on admission
Days to death	1.00
Number of LLST orders	0.02	1,00
Adequate management	0.25	0.34	1.00
Charlson index	0.10	0.18	0.14	1.00
Barthel previous to admission	0.14	-0.14	0.08	-0.02	1.00
Barthel on admission	0.08	0.02	0.00	0.31	0.59	1.00

Results

During 2004, 1140 patients were admitted to the EACU, 97% of who came from the emergency department. One hundred and three patients died (9%): the average age was 86.5 years (±6.6); 72.2% were females; 76.7% previously lived at home; and the average length of stay in hospital was 13.12 ±12.83 days. Ninety clinical records were included for analysis; patients who died within the first 24 hours following admission (n=7) or suddenly (n=2), those who were transferred (n=2), and those whose data were unavailable (n=1) were excluded.

Patients were divided into two groups

Group I consisted of 43 frail elderly patients without end-stage disease who died of acute illness, such as infections (39.5%) or cardiovascular deaths including acute stroke (34.9%).

Group II included 47 patients who met American NHPCO-criteria for advanced diseases. Of these patients, 20 had end-stage dementia, 15 had cancer in an advanced stage, and 12 had other diseases (heart disease[5], liver disease[3], respiratory disease[2], and Parkinson’s disease, Hoehn-Yahr stage-V[2]).

The most frequent problem on admission was dehydration (73.3%), followed by infection (52.2% overall) (63.8% in group II compared to 39.5% in group I), then heart failure (27.8%) and acute stroke (18.9%), which was much more frequent in group I (34.9 versus 4.3%, p<0.001).

One or more complications different from the cause of hospitalization occurred in 73.3% of the patients, of which inhospital infections were the most frequent(32.2%), followed by heart failure(15%).

The most frequent cause of death overall was infection (38 patients, 42.2%) followed by the progression or complications of a neoplasm and heart problems (each with 12 patients, 13.3%), then acute stroke (8 patients, 8.9%).

The prevalence of severe functional impairment prior to admission was 61.1%. The prevalence of severe cognitive impairment prior to hospitalization was 33.3%. Baseline characteristics of the groups are described in Table 2.

Table 2.

Baseline characteristics of patients and end of life issues

	Sample N=90	No NHPCO* N=43	NHPCO* N=47	p
Demographic Parameters:
- Age	86.5 ±6.6	87.7 ±6.7	85.5 ±6.4	0.1
- Sex female	65 (72.2%)	32 (74.4%)	33 (70.2%)	0.8
- Previous Location
Home	69 (76.7%)	32 (74.4%)	37 (78.7%)	0.5
Nursing Home	19 (21.1%)	10 (23.3%)	9 (19.1%)
Geriatric Assessment Results:
Severe functional impairment:
- Previous	55 (61.1%)	18 (41.8%)	37 (78.7%)	<0.001
- On admission	86 (95.5%)	41(95.3%)	45(95.7%)	0.9
Severe cognitive impairment:
- Previous	30 (33.3%)	9(20.9%)	21(44.7%)	0.017
- On admission	49(54.4%)	23(53.5%)	26(55.3%)	0.8
Comorbidity	4.8±2.4	4+±1.9	5.5±2.7	0.006
Home caregiver
- Non professional	39 (43.3%)	15 (34.9%)	24 (51%)	0.6
- Professional	13 (14.4%)	6 (14%)	7 (14.9%)
End of life issues:
LLST ORDERS1	82 (91.1%)	35 (81.4%)	47 (100%)	0.002
Type of order regarding
- DNAR2	75 (83.3%)	30 (69.8%)	45 (95.7%)	0.001
- No Central line	54 (60%)	24 (55.8%)	30 (63.8%)	0.5
- Not Intensive Care Unit	68 (75.6%)	29 (67.4%)	39 (83%)	0.1
- Not for Hospital transfer	57 (63.3%)	26 (60.5%)	31 (66%)	0.6
Information to patient/family	84 (93.3%)	38 (88.4%)	46 (97.9%)	0.1
Closeness to death/”label dying”	80 (88.9%)	36 (83.7%)	44 (93.6%)	0.1
Days to death from withdrawal	2.1±1.8	2 ±1.9	2.2±1.8	0.5
Recognize frequent symptoms	85 (94.4%)	40 (93%)	45 (95.7%)	0.7
Specific treatment prescription	78 (86.7%)	35 (81.4%)	43 (91.5%)	0.3

* NHO: American National Hospice Organization Criteria for end-stage diseases; 1. LLST orders: limitation of life sustaining treatment orders; 2. DNAR: do not attempt resuscitation. Data showed in Media + Standard Deviation or Frequency (Percentage)

There were written records regarding LLST-orders for 82 patients (100% in Group II compared to 81.4% in Group I; p=0.002). These included specific DNAR orders for 75 patients, which were more frequently written for Group II (95.7% versus 69.9% for Group I patients; p=0.001). The rest of the EOL recorded parameters are described in Table 2 and there were no differences between the groups.

Written LLST-orders were more likely to be present if there was established severe functional (p<0.001) or cognitive (p=0.03) impairment prior to hospitalization but were not related to the presence of severe functional (p=0.3) or cognitive (p=0.5) impairment on admission (Table 3 presents predictors for LLST-orders in both groups and combined sample). 21.1% of patients had a diagnosis of delirium on admission. Written LLST-orders were more likely to occur in the presence of advanced chronic disease (p=0.002) but not the presence of an acute illness or severe effects from an acute illness (p=0.6 and p=0.3, respectively). LLST-orders were also more likely to have been made when closeness to death was recognized (p<0.001), specific EOL-symptoms were acknowledged (p=0.06), or when specific comfort measures or treatments were prescribed (p=0.01) (Figure 3).

Table 3.

Factors related to limitation of life sustaining treatment (LLST*) orders

Combined Sample	Group I: Not Advanced Chronic Disease	Group II. Advanced Chronic Disease: 100% LLST-orders: most frequent suborders
Severe functional impairment	Severe functional Impairment	Oncological disease DNAR*
Severe cognitive impairment Co-morbidity	Co-morbidity	Non oncological-non dementia→ Not for Intensive Care Unit
Advanced organ disease	Failure of acute medicine identified In-Hospital Infection (as an acute Complication)	Dementia→ Not for Intensive Care Unit/Not for Hospital transfer.

Statistically significant predictors of Limitation of Life-Sustaining-Treatment orders in an Elderly Acute Care Unit (p<0.05). *DNAR= Do Not Attempt Resuscitation.

Figure 3.

Statistical Models Tested

Specific symptoms were recorded in 94.4% (85 patients). The most frequent was dyspnoea, followed by noisy breathing and pain, although the most treated was pain (100% patients received morphine). Morphine was prescribed for dyspnoea in 88% and midazolam for additional control in 35%. Treatment percentages did not differ between the groups, (p<0.3) (Figure 2). Prescription of symptomatic treatment was more likely to occur when there was a written note acknowledging closeness to death (p<0.001) and when there was written acknowledgement of the presence of EOL-symptoms (p<0.016).

Figure 2.

Prescription of specific symptomatic treatment

Drug access was for all cases a subcutaneous line and prescription details are showed in Figure 2. The patients with cancer in group II were frail elders who had not been referred for specific oncological treatment when stage IV advanced neoplasms were diagnosed and in whom attempts for intravenous access were made for the treatment of acute illness. When maintaining an intravenous line presented problems (e.g., repeated attempts at cannulation) and treatment could be provided subcutaneously, this was considered to be a criterion of quality EOL-management. None of them had a port-a-cath.

In 84 patients (93.3%), all active, non-symptomatic-treatments were withdrawn when the failure of acute non-symptomatic treatment was recognized. Pre-death was identified in 88.9% of patients (83.7% of those dying from acute illness and 93.6% of those with chronic illness) a median of 2 days before death (0–6 days in chronic disease patients, average 2 +1.8 days, and 0–8 days in those dying from acute disease, average 2.2 +1.9 days); the figures were 2.1 ±1.9 days before death in those who had LLST-orders and 1.5 ±0.7 days before in those without LLST-orders (p=0.10).

Not stopping treatment for acute illness was related to the absence of severe functional impairment (0% in those with severe functional impairment versus 12% in those whose functional impairment was not severe, p=0.04). All subtypes of LLST-orders were related to the acknowledgement pre-death or using the label “dying” (p<0.001 for DNAR, p=0.013 for no central line, p<0.001 for not for the intensive care unit, and p=0.004 for not for hospital transfer). Statistically significant relations are showed in Figure 3.

Structural equation modelling: A theoretical model of decision-making variables was tested (Figure 1). Table 1 shows the correlation matrix of the variables included in the model. Earlier identification of closeness to death was influenced by the presence of co-morbidity (β=0.15) and previous functional impairment (β=0.23), while the influence of the impact of acute disease on functional status was smaller (β=–0.11). The number of LLST-suborders was influenced by co-morbidity (β=0.18) and previous functional status (β=-0.14). Adequate EOL-management was related to earlier identification of closeness to death (β=0.25) and to the number of written LLST-suborders (β=0.34). The number of LLST-suborders was not influenced by earlier identification of closeness to death (β=0.019) or the impact of acute disease on functional status (β=0.035).

In a second stage of analysis, to obtain a parsimonious model, all paths of the initial model smaller than 0.10 were deleted. The final model is presented in Figure 3. Various goodness-of-fit indexes calculated for the final model indicated that the estimated model provided good fit to the data. The chi-square fit index corrected for non-normality was statistically non-significant, which is consistent with an excellent model fit (X2=2.06, p=0.78). The goodness-of-fit-index(GFI) and the adjusted-goodness-of-fit-index(AGFI), which range between 0 and 1, were high, which is associated to being a good fit (GFI=1.0/AGFI=0.99).

Discussion

This analysis was performed as a pilot study of a multi-centre project being conducted to describe EOL-decision-making processes in hospitalized, vulnerable elders in Spain. Decision-making is a complex process, and in geriatric medicine EOL-care is a continuum in which disease-modifying treatments are replaced by EOL-care when closeness to death is identified. The decision to treat or withhold treatment is complex and involves evidence-based analysis of the burdens and benefits. It is important to be sure what factors are relevant when making these decisions.

Withdrawal of treatments at EOL has been described in six European countries and Australia but no data were available from Spain (4, 26). The elderly patients studied in our project were similar to those admitted to any general hospital medical acute ward in our country but the frequencies of LLST-orders and patient/family-discussion obtained were higher, despite a similar death rate (9.9 vs 9.4 for patients >75yr) (11, 12, 27). A Spanish study describing LLST in elderly patients who died in hospital with advanced organ disease found lower frequencies of recorded orders (11); specific DNAR orders were only recorded in 37% compared to 91.7% of end-stage patients in our study. Our data are similar to those reported by Hesse et al (28) who found DNAR suborders in 95% of patients older than 80 years of age in whom closeness to death had been identified. In the general sample in our study, DNAR written suborders were recorded in 83.3% of the clinical notes, which is in agreement with the 85% reported by Tschann (29) in hospitalized patients and the 83.8% reported by Lamberg et al(30) in nursing home residents with end-stage dementia at the time of death.

Studies have found variability in the frequency of LLST-orders, death rates and EOL-issues across different settings and some of the factors accounting for this variability have been identified; for example, frequency of LLST-orders might be related to patient characteristics or physician characteristics such as socio-cultural factors, age, gender, life-stance, religious beliefs, training in palliative care medicine, and the number of terminal patients attended in the previous year (31).

In the same way that decisions to forgo hospitalization for nursing home residents may not be made until death is imminent(30), EOL-decisions in our setting were made when failure of acute medicine management and a pre-death status were identified, which can clearly be improved once seen that the label dying improves EOL-management. Care plans and DNAR orders were rarely documented prior to the final admission in one study of elderly patients in the last year of life(6), which may reflect an emphasis on the treatment of acute illness. This means that geriatric assessment should lead to EOL and care plans-discussions earlier in the evolution of illnesses in order to improve quality of care.

In the classical SUPPORT study, DNAR orders were recorded in 81% of patients who died during hospital admission and were related to patient preferences, estimated survival and age (32). When analyzed specifically, DNAR orders involved consideration of functional status and quality of life, reflecting important prognostic factors for survival during hospital admission, as shown in the Hospitalized Elderly Longitudinal Project Study (9).

In our sample, LLST-orders were more likely to be present when criteria for advanced disease were met or there was severe functional impairment prior to admission, and these have been described as quality indicators on which to base the start of palliative measures in vulnerable elders (33). However, it is possible that the effect of advanced disease on decision-making is overwhelmed by severe functional impairment, which is the better predictor of outcome and was linked to the presence of LLST-orders in both our groups and analytical models, whether American NHPCO criteria were met or not. In our study, severe established functional impairment prior to admission was significantly related to the presence of LLST-orders but the influence was smaller in the structural modelling equation. Categorizing functional impairment into “severe” versus “not severe” may help explain this: only when functional status was very severely impaired (CFR>3 or Barthel index<20) did it affect written decisions about LLST. Intermediate impairment might have no effect. Measuring functional loss can help make physicians aware of pre-death situations, and this awareness was related in both our statistical models to adequate EOL-management. In those with severe impairment at baseline, identification of closeness to death occurred later. The impact of a severe illness, however, was very low, reflecting an initial attempt at acute treatment and a lower sensitivity of physicians to smaller ranges of functional change. Co-morbidity also played an important role in making LLST-orders in both statistical models. However, concerns about functional autonomy have been identified by patients as crucial factors when considering EOL-decisions (34).

In our study, age was not related to withholding life-sustaining treatment, reflecting a management approach in which baseline characteristics were relevant factors for decision-making (an alternative explanation is that all patients were over 75, which was the limit of age reported for significance in SUPPORT, that found a linear association between age and DNAR-decisions. The SUPPORT findings may reflect the effect of baseline features not included for general analysis or the influence of other factors such as attending physicians’ approaches (32)).

Older patients in Spain may not be as willing to receive life-sustaining therapies as patients in other countries. Kessel et al (35) reported that 87.2% of geriatric Spanish patients who were asked, declined resuscitation (88.38% of healthy elderly, 86.81% of those suffering from chronic illnesses, and 84.62% of those who met advanced-organ-disease criteria) but in other European countries higher percentages of patients desired resuscitative measures (36).

Identifying closeness to death seems to be of great relevance, as all other EOL-parameters were related to a written record of the label “dying” in both our models. Flaming (16) suggested that the in-hospital use of the label “dying” or written acknowledgement of closeness to death might improve the quality of assistance for patients with EOL-needs. Goodlin et al. found that less aggressive measures were used for patients designated for “comfort measures only” but this did not correlate with initiating measures to improve EOL-care (i.e., recognizing or treating symptoms) (37). Our findings reveal the importance of recording closeness to death in writing earlier (identified in 88.9% of those with advanced organ disease and 83% of those without American NHPCO criteria but only a median of two days before death). Important measures relating to EOL-care, such as specific discussion with the patient/family regarding the short-term prognosis and documentation and treatment of specific symptoms, which occur frequently at the end-of-life, were related to identifying and recording in writing closeness to death in both statistical models. Veerbeek et al. also found that recognition of the dying phase reduced the number of undesirable diagnostic tests (38). In our structural equation model, there was a clear link between identifying earlier a pre-death phase, LLST-orders and adequate EOL-management.

Symptoms were identified and specific symptomatic treatment was started in similar percentages of both groups, whether advanced diseases criteria were met or not. Compared to the EURELD Study (4), there was a similar frequency of specific symptomatic treatment and a higher rate of withdrawal of active, non-symptomatic treatment, perhaps because of the greater age and/or differences in baseline functional features of the patients in our sample.

A written record of a specific discussion with the patient/family regarding short-term prognosis and comfort measures was present in 93% of cases compared to 57% and 49% in other studies (11, 35) and was also related to the written use of the label “dying” and LLST-orders, reflecting the importance of family involvement in EOL-hospital care (29).

In geriatric medicine an evidence-based decision-making process avoiding both failure of adequate assessment and unnecessary aggressiveness may lead to improved EOL-treatment. Through a comprehensive and evidence-based geriatric and clinic assessment, we can identify these factors and therefore apply proportioned care when necessary, avoiding futile measures, such as intensive care or resuscitation. EOL-care of vulnerable elders may be effectively improved through an accurate assessment of needs and access to aggressive symptom management when necessary.

Although retrospective studies in EOL-care are interesting (39), they have limitations. In particular, our study was small and looked only at written records, which may not thoroughly document the decision-making process. Also, many aspects of the decision-making process are difficult to operationalize. Spiritual considerations, for example, were not analyzed and the degree of patient participation in decision-making was not examined, except for documentation of discussions with the patient/family about the measures applied. Further research is being conducted to determine whether these findings are reproducible in other acute care settings in this country and to uncover any potential confounding factors.

In conclusion, our study shows that the likelihood of written LLST-orders in an EACU was related to the severity of patient impairment, presence of co-morbidity, and meeting criteria for advanced diseases and that better EOL-management was related to documenting closeness to death and LLST-orders. EOL-discussions based on comprehensive and evidence-based accurate assessment of needs should be held early in care plans to improve EOL-care of vulnerable elders.

Previous submissions: The study in which these data are based was presented as an oral communication to the Spanish Geriatric Society Annual Meeting in Málaga, Spain, June 2005. The authors declare no conflict of interests.

Financial disclosure: None of the authors had any financial interest or support for this paper.