| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||
1 Department of Community Health Sciences and 2 Centre on Aging, University of Manitoba, Winnipeg, Manitoba, Canada.
Address correspondence to Verena Menec, PhD, University of Manitoba, Community Health Sciences, 750 Bannatyne Ave., Winnipeg, Manitoba, R3E 0W3 Canada. E-mail: menec{at}cc.umanitoba.ca
 |
A |
|---|
 Top Abstract MethodsResultsDiscussionReferences |
|---|
Methods. The study included all adults 65 years old or older who died in Manitoba, Canada in 2000 (N = 7678). Measures were derived from administrative data files and included location of death, hospitalizations, intensive care unit (ICU) admission, long-term care (LTC) use, physician visits, and prescription drug use in the last 30 days versus 180 days before death, respectively.
Results. Individuals 85 years old or older had increased odds of being in a LTC institution and also dying there than did individuals 65–74 years old. They had, correspondingly, lower odds of being hospitalized and being admitted to an ICU. Although some statistically significant age differences emerged for physician visits, the effects were small. Prescription drug use did not vary by age.
Conclusions. These findings indicate that very elderly individuals tended to receive care within LTC settings, with care that might be considered aggressive declining with increasing age. However, health care use among all age groups was substantial. A critical issue that needs to be examined in future research is how to ensure quality end-of-life care in a variety of clinical contexts and care settings for individuals of all ages.
Apart from concerns with where people die, concern over end-of-life health care has also been raised because of the substantial increase in health care use at the end of life (15–17). The increase in hospital use is particularly dramatic in the last month before death—a 250% increase compared to the second-to-last month of life in one study focusing on frail elderly individuals (15). Total health care resource use (as measured in dollars) more than doubled in the last month before death compared to 12 months or 24 months prior to death, with hospital use being a major component of the resources used (15).
A controversial issue is whether health care use and cost at the end of life may be too high and due to unnecessary treatment. In particular, the issue has been raised as to whether very elderly individuals receive overly aggressive treatment at the end of life (18–22). Contrary to this possibility, however, research indicates that aggressive treatment, including intensive care unit (ICU) admissions, cardiac catheterization, dialysis, ventilators, pulmonary artery monitors, and chemotherapy decreases with very old age (19–21). Yet few studies have systematically examined potential age variation across a wider range of health services. The few studies that exist suggest that when total health care use and costs are considered, use and cost indeed increase with age, which is primarily due to the heavy use of nursing homes among the very old (18,22).
The purpose of the present study was to further explore the question of how health care use at the end of life varies across age and, in particular, to examine whether the health care use of elderly individuals 85 years old or older differs from that of younger individuals. We examined this issue using a range of health care indicators. Location of death was assessed because it is sometimes used as an indicator of appropriateness of care at the end of life (2,6). However, location of death does not reflect the varying patterns of health care use. For example, whereas one individual may have died in hospital after a brief hospital stay, another person might have spent considerable time in hospital, only to be transferred to a long-term care (LTC) setting shortly before death. Thus, we also examined hospital use and LTC use in the last 30 days and 180 days before death, respectively. As an indication of the aggressiveness of treatment, we examined more specifically use of ICUs. Lastly, we examined patterns of physician visits and prescription drug use.
Context of the Present Study
The present study was conducted in Manitoba, a midwestern Canadian province with a population of about 1.1 million people, 13.5% of whom are 65 years old or older. Although ethnically diverse, older adults in Manitoba are predominantly of European/Eastern European descent. Manitoba has little in or out migration, particularly among seniors, with < 1% of individuals 65 years old or older leaving the province per year (23).
Given Canada's universal health care system, health care services (hospital, physician, and home care) are provided at no charge to patients. Nursing home use is partially funded, subject to a residential charge determined based on individuals' income. Receipt of home care and admission to nursing homes occurs on the basis of a single-access, standardized assessment that is based solely on need. Care for terminally ill patients who are registered as "palliative" (i.e., deemed by their physician as having a life expectancy of < 6 months and who have agreed to forego curative treatment) is also provided free of charge. Palliative care for these terminally ill individuals is provided in hospitals, special inpatient hospital units, or the patient's own residence, depending on the person's preference and/or required level of care, with care in the home being provided through the provincial home care program. As patients registered as palliative could not reliably be identified in the administrative data, the care provided for them was not examined in the present study.
|
METHODS |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
Study Population
The study population was a complete cohort of residents of Manitoba 65 years old or older who died in 2000 (7678 individuals), as identified from Vital Statistics data.
Measures
Predictor variables.--
Three age groups were created: 65–74, 75–84, and 85+ years. Demographic characteristics included in all analyses were: gender, marital status, and region of residence (see Table 1 for descriptive information).
|
Outcome variables.-- Location of death was identified by linking administrative data files, which allowed us to differentiate among four categories: (i) hospital; (ii) LTC facility; (iii) home while receiving home care; and (iv) home without home care. Note that some of the deaths in categories iii and iv might have occurred outside of the home, such as on the road due to a car accident.
Given data limitations, we were not able to specifically identify patients who were registered as palliative. This means that some individuals classified as having died in a hospital could have died in a specialized inpatient hospital unit for terminally ill patients. We estimate that about 5% of decedents died in an inpatient palliative care unit, about half of whom would therefore be included in our hospital deaths, as one unit is located within an acute care hospital. The other half would be included in our LTC deaths, as the second palliative care unit is located in a chronic care hospital, which was classified here as an LTC facility. Some individuals identified as having died on home care also might have received palliative rather than regular home care. This likely represents a small proportion of the 2000 decedent cohort used here, as increased emphasis on providing palliative care for terminally ill individuals at home has been placed in Manitoba in more recent years. For example, the Palliative Drug Access Program, which provides free drug coverage for patients who are registered as palliative who want to die at home, was introduced in 2002.
Hospital use (excluding ICUs) was derived from the hospital discharge abstract, and we determined whether individuals were hospitalized (i.e., incurred any hospital days) within 30 days and 180 days before death, respectively. Because of the large number of people that had no hospital admission during these time frames and therefore incurred no hospital days, the variables were subsequently dichotomized into 0 days versus 1+ days. Hospital days were also examined for those persons hospitalized at least once, that is, those persons who incurred one or more hospital days.
ICU use was derived from the hospital discharge abstract for the 30 and 180 days before death and was dichotomized because the majority of decedents were never admitted to an ICU. LTC use was identified from the LTC file for the 30 and 180 days before death and was subsequently also dichotomized into no use versus some use, given the large number of individuals who were never in an LTC institution. Physician visits were derived from physician claims data and included visits to general/family physicians and specialists. Visits to nursing home residents are included in this measure. Prescription drug use was derived from the drug file, which contains data for all prescription drugs filled, both in the community and in LTC settings. It does not contain drugs dispensed in hospitals, nor does it capture over-the-counter drugs.
Analyses
Multinomial regressions were used to examine the relationship between age and location of death, and were appropriate because there were four categories of the outcome variable. Hospital use and LTC use were examined using logistic regressions, given that these were dichotomous variables. Hospital days for those persons hospitalized at least once, physician visits, and prescription drug use were treated as continuous outcome measures and were analyzed from a generalized linear model (GLM) perspective using a negative binomial distribution. This distribution provided an excellent fit for these skewed measures.
|
RESULTS |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
|
|
|
Table 4 shows that the odds of being hospitalized were increased for individuals aged 75–84 years relative to those aged 65–74 years. No significant difference emerged between individuals aged 85 years or older and their 65- to 74-year-old counterparts. In contrast, the analysis for hospital days among those hospitalized at least once did not reveal an age effect (data not shown). Only cause of death and region of residence were significantly related to number of days. Individuals who died of injuries, circulatory diseases, and nervous or sensory organ diseases incurred fewer days than those who died of cancer (relative rate =.70,.85, and.81, respectively), and rural residents incurred more days than did urban residents (relative rate = 1.08).
|
As shown in Table 5, physician visits did not differ between persons aged 75–84 years relative to those aged 65–74 years. The effect for individuals aged 85 years or older differed depending on the time frame examined. Although persons aged 85 years or older incurred more visits than did the individuals aged 65–74 years within 30 days of death, they made fewer visits when focusing on a 6-month time frame. Although statistically significant, these effects were quite small, however. No age effects were evident for prescription drug use.
|
|
DISCUSSION |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
Consistent with the findings for location of death, individuals aged 85 years or older had 5 times the odds of being in an LTC institution 1 month and 6 months prior to death, relative to individuals aged 65–74 years. Moreover, although 75- to 84-year-old individuals had greater odds of being hospitalized than did 65- to 74-year-olds, those individuals aged 85 years or older were no more likely to be hospitalized within the last month or last 6 months of death than were those persons in the youngest age group. Among hospitalized persons, days in hospital also did not increase with age. Noteworthy, however, is the very clear age gradient for ICU admissions. Individuals aged 85 years or older had considerably reduced odds of being admitted to an ICU relative to those aged 65–75 years. This finding is consistent with previous research that shows that intensive treatment decreases in very old age (19–21).
Thus, in combination, the present results suggest that very elderly individuals (85 years or older) received less aggressive treatment at the end of life than their younger counterparts received. This conclusion is also supported by the finding that prescription drug use did not vary across age. Moreover, although we found some statistically significant age differences for physician visits, these effects were small, suggesting that there was little variation in physician use across age groups.
Although the present findings suggest that the care of very elderly individuals (85 years or older) may not be overly medicalized, at least not relative to that of younger individuals (aged 65–74 years), they do raise the question of what kind of end-of-life care is provided to seniors, and particularly very elderly individuals who represent the largest proportion of decedents. The need for geriatric palliative care, which bridges the fields of palliative care and geriatrics, has been identified (28). Currently, in Manitoba, the Palliative Care Program focuses on individuals who are in the final stages of a terminal illness, predominantly cancer. Yet the majority of very elderly individuals do not die of cancer. In the present decedent cohort, only 15% of individuals aged 85 years or older died of cancer, compared to 39% of those aged 65–74 years. Moreover, many elderly individuals have a number of health problems and, therefore, require multiple treatments; this is not the case in typical palliative care for those persons who are dying (28). Thus, the question of how to ensure quality end-of-life care for elderly individuals in a variety of clinical contexts and settings represents an important area of research for the future.
Apart from the age variation in health care use at the end of life observed here, the extent of health care use in the last months of life among all age groups is noteworthy. For example, almost half of decedents spent some time in hospital during the last 30 days of their lives. Much of the hospital use was concentrated into the last month of life, with hospital days increasing by almost 300% in the last month before death, compared to the five previous months, an increase that is similar to what has been found in previous research in the United States (15).
Prescription drug use was also substantial in the last month and last 6 months of life. Unlike for hospitalizations and ICU admission, use was not concentrated into the last month before death, however. This finding may suggest that the death trajectory of a significant proportion of older adults extended over at least a 6-month time span, rather than displaying a very sudden decline. Alternatively, much of the drug use at the end of life may have been incurred in hospitals and thus would no longer be captured in the prescription drug file. On average, individuals had about four prescriptions filled in each of the 6 months before death, resulting in a total of (on average) 24 prescriptions. This is substantially higher than what has been reported for the general population in the province of Manitoba, where the rate of prescription drug use is about six prescriptions per Manitoba resident in a given year (29). It underscores again the extent to which health care use increases at the end of life.
Several limitations of this study must be acknowledged at this point. Most importantly, although administrative data allow us to examine health care use at the end of life, we cannot determine the quality of care. Nor can we determine whether the care received or the location of death reflected individuals' wishes. For example, the question of whether those persons who were hospitalized at the end of life might have been better cared for in alternative settings (such as LTC institutions) is a complex one that cannot be answered with this study, although comparison with previous research suggests that patterns of findings are at least in line with those from other jurisdictions. For example, a study from the United Kingdom showed that, similar to the present study, 50% of individuals 75 years old or older with cancer died in hospital (26).
Second, we were not able to examine issues around palliative care services provided to terminally ill patients, as the data did not allow us to identify reliably patients registered as palliative. We were therefore not able to focus more specifically on individuals in palliative inpatient hospital units or differentiate between persons who died at home and received palliative care services through the home care program versus persons who received regular home care. Moreover, we were not able to specifically examine prescription drug use among these individuals.
Third, although we controlled for cause of death in our analyses, we did not examine more specific conditions. In this respect, characterizing the health care use of different functional trajectories—terminal illness, sudden death, organ failure, and frailty (30,31)—would be particularly useful. Lastly, although we examined overall patterns of health care use at the end of life, more in-depth analyses are now needed to examine more specifically issues such as reasons for hospitalizations or physician visits, and types of prescription drugs used.
However, the present study also highlights the strengths of using administrative data to examine patterns of health care use at the end of life. It allowed us to examine a variety of health care use indicators for an entire population, to an extent that has not been the case in previous research. Thus, the inclusion of not only hospital use, but also LTC use, physician visits, and prescription drug use provides a more comprehensive picture of the end-of-life health care use of older adults than was provided in previous research. Routine reporting of end-of-life indicators, such as location of death, hospital use, proportion of deaths in ICUs, and prescription drug use prior to death could effectively be used to monitor a jurisdiction's progress in end-of-life care. Such routine reporting might, for example, help prioritize targets for improvements to palliative care services for terminally ill individuals. In this respect, home death has been proposed as a useful benchmark for palliative care progress.
|
F |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
Received March 8, 2006
Accepted August 21, 2006
|
References |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
| ||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
