Causes of Death for Patients
With Community-Acquired Pneumonia
Results From the Pneumonia Patient
Outcomes Research Team Cohort Study
Arch Intern Med. 2002;162:1059-1064
Eric M. Mortensen, MD, MSc; Christopher M. Coley, MD; Daniel E. Singer,
MD; Thomas J. Marrie, MD; D. Scott Obrosky, MSc; Wishwa N. Kapoor, MD, MPH;
Michael J. Fine, MD, MSc
Background To our knowledge, no previous study has systematically examined
pneumonia-related and pneumonia-unrelated mortality. This study was performed
to identify the cause(s) of death and to compare the timing and risk factors
associated with pneumonia-related and pneumonia-unrelated mortality.
Methods For all deaths within 90 days of presentation, a synopsis of all
events preceding death was independently reviewed by 2 members of a 5-member
review panel (C.M.C., D.E.S., T.J.M., W.N.K., and M.J.F.). The underlying and
immediate causes of death and whether pneumonia had a major, a minor, or no
apparent role in the death were determined using consensus. Death was defined
as pneumonia related if pneumonia was the underlying or immediate cause of
death or played a major role in the cause of death. Competing-risk Cox proportional
hazards regression models were used to identify baseline characteristics
associated with mortality.
Results Patients (944 outpatients and 1343 inpatients) with clinical and
radiographic evidence of pneumonia were enrolled, and 208 (9%) died by 90 days.
The most frequent immediate causes of death were respiratory failure (38%),
cardiac conditions (13%), and infectious conditions (11%); the most frequent
underlying causes of death were neurological conditions (29%), malignancies
(24%), and cardiac conditions (14%). Mortality was pneumonia related in 110
(53%) of the 208 deaths. Pneumonia-related deaths were 7.7 times more likely to
occur within 30 days of presentation compared with pneumonia-unrelated deaths.
Factors independently
associated with pneumonia-related mortality were hypothermia, altered mental
status, elevated serum urea nitrogen level, chronic liver disease, leukopenia,
and hypoxemia. Factors independently associated with pneumonia-unrelated
mortality were dementia, immunosuppression, active cancer, systolic
hypotension, male sex, and multilobar pulmonary infiltrates. Increasing age and
evidence of aspiration were independent predictors of both types of mortality.
Conclusions For patients with community-acquired pneumonia, only half of all
deaths are attributable to their acute illness. Differences in the timing of
death and risk factors for mortality suggest that future studies of
community-acquired pneumonia should differentiate all-cause and
pneumonia-related mortality.
Arch Intern Med. 2002;162:1059-1064
PNEUMONIA COMBINED with influenza is the sixth
leading cause of death in the United States.1 Although the mortality
rate from pneumonia decreased sharply with the introduction of antibiotic
therapy in the 1940s, since 1950, the overall mortality rate for this illness
has either remained stable or increased.2 In a meta-analysis3 of studies of
prognosis, the short-term mortality of patients hospitalized with community-acquired
pneumonia ranged from 5.1% for patients treated in an ambulatory or hospital
setting to 36.5% for patients treated in an intensive care unit.
Prior studies4-6 of pneumonia
prognosis focused almost exclusively on short-term mortality and assessed risk
factors for all-cause mortality. To our knowledge, no previous studies have
examined the causes of death of patients with community-acquired pneumonia or
the role that pneumonia played in the cause of death. The goals of this study
were as follows: (1) to identify the underlying and immediate causes of death
for patients with community-acquired pneumonia, (2) to determine the role that
community-acquired pneumonia played in the cause of death, and (3) to compare
the risk factors associated with pneumonia-related and pneumonia-unrelated
mortality in patients with this illness.
PATIENTS AND METHODS
PATIENT RECRUITMENT
The Pneumonia Patient Outcomes Research Team cohort study was conducted at 5
medical institutions in 3 geographic locations between October 12, 1991, and
March 31, 1994. These were the University of Pittsburgh Medical Center, a
942-bed university teaching hospital, and St Francis Medical Center, a 427-bed
community teaching hospital, in Pittsburgh, Pa; Massachusetts General Hospital,
an 899-bed university teaching hospital, and Harvard Community Health
Plan–Kenmore Center, a staff-model health maintenance organization, in Boston,
Mass; and Victoria General Hospital, a 637-bed university teaching hospital, in
Halifax, Nova Scotia. Outpatients (defined as those initially treated in an
outpatient setting) and inpatients were enrolled from each of the 4
hospital-based sites (University of Pittsburgh Medical Center, St Francis
Medical Center, Massachusetts General Hospital, and Victoria General Hospital);
only outpatients were enrolled from the Harvard Community Health Plan–Kenmore
Center.
Potential study subjects were identified by
research assistants through daily reviews of admitting and radiology department
logs and records of patients presenting to the emergency departments and
clinics affiliated with the participating sites. Inclusion criteria were as
follows: 18 years of age or older, 1 or more symptoms suggestive of
community-acquired pneumonia, radiographic evidence of community-acquired
pneumonia not known to be chronic, and informed consent for baseline and follow-up
interviews. Exclusion criteria were as follows: discharge from an acute-care
facility within 10 days of presentation, known seropositivity for the human
immunodeficiency virus, or pulmonary symptoms secondary to another diagnosis
(eg, lung cancer). Patients were only enrolled once during the study; those who
presented with community-acquired pneumonia on more than 1 occasion were not
subsequently enrolled.
BASELINE ASSESSMENT
For all study patients, baseline sociodemographic information and clinical data
were assessed at presentation by direct interview by a study nurse and medical
record review. If unable to obtain information directly from the patient
because of mental status changes or language or communication barriers, a proxy
respondent was used. Clinical data examined included medical history, physical
examination results, laboratory values, chest radiographic findings, and
microbiologic results. Historical information obtained included 5 common
respiratory symptoms (cough, dyspnea, sputum production, pleuritic chest pain,
and hemoptysis) and 14 common nonrespiratory symptoms (fatigue, fever,
anorexia, chills, sweats, headache, myalgias, nausea, sore throat, confusion,
inability to eat, vomiting, diarrhea, and abdominal pain). Physical examination
data collected included vital signs and an evaluation of mental status.
Laboratory data collected, when available, included white blood cell count;
hematocrit; levels of serum urea nitrogen, serum sodium, liver enzymes, and
arterial blood gases; and pulse oximetry readings. Radiographic data included
location of the infiltrate, pattern of the infiltrate (predominantly alveolar,
predominantly interstitial, miliary, or mixed alveolar and interstitial), and
presence of pleural effusion.
When ordered by the physicians caring for these
patients, the following microbiologic tests were abstracted: sputum gram stains
and bacterial cultures obtained within 2 days of presentation, blood cultures
drawn before initiating antimicrobial therapy, pleural fluid cultures, and
short-term (
1 week of presentation) and
convalescent (1-8 weeks after presentation) serologic tests. Results of these
tests were reviewed and a microbiologic cause was assigned, as previously
described.7
Copies of the initial chest radiographs used for
the diagnosis of pneumonia at each study site were independently reviewed by a
3-member panel of attending radiologists who had no patient-specific clinical
information. Pleural effusion was quantified by the maximum present in either
lung as follows: none, minimal (costophrenic angle blunting only), moderate
(less than one third of the pleural space), and large (one third or more of the
pleural space).8 Aspiration pneumonia
was diagnosed by the clinical committee based on radiographic data and synopses
of clinical data. Aspiration pneumonia was diagnosed in patients with a
disorder known to alter consciousness, the normal gag reflex, or the swallowing
mechanism in whom the chest radiograph revealed an infiltrate involving the
superior or basilar segments of the lower lobes or the posterior segments of
the upper lobes.9
The severity of illness at presentation was
quantified using the validated Pneumonia Patient Outcomes Research Team
prediction rule for 30-day mortality and medical complications in patients with
community-acquired pneumonia.10 This rule is based on
3 demographic characteristics, 5 comorbid illnesses, 5 physical examination
findings, and 7 laboratory and radiographic findings available at presentation.
This rule classifies patients into 5 risk classes, with the 30-day mortality ranging
from 0.1% for those in class I to 31.1% for those in class V.
ASSESSMENT OF MORTALITY AND THE
CAUSE OF DEATH
Mortality was assessed at 90 days after initial enrollment in the study. For
all patients who died during the follow-up period, death summaries were
prepared by study research nurses using salient information obtained from the
medical record, family or caregiver interviews, and autopsy reports (when
available).
Each death summary was independently reviewed by
2 study investigators who were part of a 5-member clinical review panel
(C.M.C., D.E.S., T.J.M., W.N.K., and M.J.F.). Four members of the clinical
review panel were general internists (C.M.C., D.E.S., W.N.K., and M.J.F.) and 1
was an infectious disease specialist (T.J.M.); all reviewers had extensive
clinical and research experience regarding patients with community-acquired
pneumonia. The reviewers were asked to assign the underlying and immediate
causes of death based on World Health Organization criteria,11 and to assess the
role that community-acquired pneumonia played in the patient's death. The
underlying cause of death was defined as the disease or injury that initiated
the cascade of morbid events leading directly to death. The immediate cause of
death was defined as the disease process, injury, or complication immediately
preceding death. If community-acquired pneumonia was not considered to be the
underlying or immediate cause of death, then each reviewer was asked to
determine whether community-acquired pneumonia played a major or a minor role
in the patient's death. Pneumonia was judged as playing a major role if death
would not have occurred if the patient did not have pneumonia but another
condition was present that also contributed. Pneumonia was defined as playing a
minor role if community-acquired pneumonia was not essential to explain the
patient's death but played some role in the patient's death.
After the causes of death and the role of
pneumonia in causing death were independently assigned by 2 reviewers, each
case was presented to the 5-member clinical review panel. Final assignments of
the underlying and immediate cause of death and the role of pneumonia in
causing death were based on the full consensus of this panel.
Mortality was classified as pneumonia related if
pneumonia was an immediate or underlying cause of death or if it played a major
role in the patient's death. Mortality was defined as pneumonia unrelated if
pneumonia was neither an immediate nor an underlying cause of death, and played
only a minor role, no role, or an unknown role in the cause of death.
STATISTICAL ANALYSES
Univariate statistics were used to compare differences in sociodemographic and
clinical characteristics in patients with pneumonia-related and
pneumonia-unrelated mortality. Causes of death as a function of pneumonia
severity risk class and timing of death were analyzed using simple descriptive
techniques. Categorical variables were analyzed using the 
2 test, and continuous variables were analyzed
using the t test. To analyze time
to death for patients with pneumonia-related and pneumonia-unrelated mortality,
Kaplan-Meier estimated probabilities were computed. Statistical significance
was assessed using the summary log-rank test. Statistical significance was
defined as P.05 (2-tailed) for all univariate and multivariate analyses.
To evaluate risk factors for pneumonia-related,
pneumonia-unrelated, and all-cause mortality, baseline patient sociodemographic
and clinical characteristics were used as independent variables in 3 Cox
proportional hazards regression models, using the 3 mortality outcomes as the
respective dependent measures. The baseline variables included all factors
composing the Pneumonia Patient Outcomes Research Team severity model, in
addition to others that were postulated to have an association with 90-day
mortality.10 Site of care,
severity risk class, intensive care unit status, do not resuscitate status, and
symptoms were omitted as potential predictors. All baseline variables that were
statistically significant in any of the 3 Cox proportional hazards regression
models were then used in a competing-risk Cox proportional hazards regression
model with pneumonia-related mortality, pneumonia-unrelated mortality, and
survival as the respective dependent measures.12 The
Kolmogorov-Smirnov test was used to test the statistical significance of the
survival curves for pneumonia-related and pneumonia-unrelated mortality in the
competing-risk analysis.13
RESULTS
Of the 2287 patients enrolled in the Pneumonia
Patient Outcomes Research Team cohort study, 208 (9%) died within 90 days.
Overall, 194 (14%) of the 1343 inpatients and 14 (1%) of the 944 outpatients
died within this follow-up period.
CAUSES OF DEATH
As shown in Table 1,
respiratory failure (38%), sepsis or bacteremia (7%), and cardiac arrhythmia
(7%) were the 3 most frequent immediate causes of death. Neurological
conditions (29%), lung cancer (13%), and cardiac ischemia (13%) were the 3 most
frequent underlying causes of death.
Death was defined as pneumonia related in 110
(53%) of the 208 deaths. Of the pneumonia-related deaths, pneumonia was the
underlying cause of death in 20 patients, the immediate cause of death in 9,
and a major contributor to death in 81. Of the pneumonia-unrelated deaths,
pneumonia played a minor role in 34 patients, no role in 52, and an unknown
role in 12.
There were distinct differences between the
immediate and underlying causes of death for pneumonia-related and
pneumonia-unrelated mortality. The most frequent immediate causes of death for
pneumonia-related mortality were respiratory failure (50%), pneumonia (8%),
multisystem organ failure (6%), and sepsis (6%). In comparison, respiratory
failure (26%), sepsis or bacteremia (9%), cardiac arrhythmia (8%), and
congestive heart failure (7%) were the leading immediate causes of death for
pneumonia-unrelated mortality. The most frequent underlying causes of death for
pneumonia-related mortality were neurological conditions (22%), pneumonia
(18%), and cerebrovascular accident (13%), compared with lung cancer (19%),
other malignancies (17%), and cardiac ischemia (17%) for those with
pneumonia-unrelated mortality.
FACTORS ASSOCIATED WITH
MORTALITY
The demographic and clinical factors with significant univariate associations
with all-cause 90-day mortality are shown in Table 2.
Overall, 85% of all deaths occurred among patients in the 2 highest risk
classes; a greater proportion of pneumonia-related deaths also occurred within
risk classes IV and V.
Survival plots and frequency distributions of
death over time of pneumonia-related and pneumonia-unrelated mortality are
shown in Figure 1
and Figure 2.
For the 110 pneumonia-related deaths, 45% occurred within 2 weeks and 76%
occurred within 30 days of presentation, compared with 8% and 30%,
respectively, of the pneumonia-unrelated deaths (P<.001 for both comparisons). The odds of a
pneumonia-related death occurring within 30 days of presentation was 7.7 that
of a pneumonia-unrelated death. The Kolmogorov-Smirnov test confirmed
significantly different patterns in the time to death for those with
pneumonia-related and pneumonia-unrelated mortality (P.001).
As shown in Table 3,
6 factors were independently associated with pneumonia-related mortality only:
hypothermia, altered mental status, elevated serum urea nitrogen level, chronic
liver disease, white blood cell count less than 4000/µL, and hypoxemia. In
addition, 6 factors were associated with pneumonia-unrelated mortality only:
dementia, immunosuppression, active cancer, systolic hypotension, male sex, and
multilobar infiltrates. Two variables, increasing age and evidence of
aspiration, were independently associated with pneumonia-related and
pneumonia-unrelated mortality. The magnitude of association for the factors
independently associated with pneumonia-related mortality only ranged from a
hazard ratio of 1.90 for temperature lower than 36.0°C to 3.88 for chronic
liver disease. The magnitude of association for the factors independently
associated with pneumonia-unrelated mortality only ranged from 1.59 for male sex
to 2.82 for dementia.
COMMENT
This detailed study of mortality in patients
with community-acquired pneumonia demonstrates substantial differences in the causes,
timing, and risk factors for pneumonia-related and pneumonia-unrelated deaths.
The causes of death for patients in this study were similar to the most common
causes of death for adults in the United States: coronary artery disease,
malignancies, stroke, and chronic obstructive pulmonary disease.14 The most frequent
immediate causes of death in this study were respiratory failure and cardiac
disease, while malignancies and neurological disorders were the most frequent
underlying causes of death. However, several causes of death that many would
associate with community-acquired pneumonia, including sepsis, bacteremia, and
multisystem organ failure, were infrequent causes of death in this cohort. In
addition, when the cause of death was stratified by the role of pneumonia,
there were distinct differences between the 2 types of mortality. For patients
with pneumonia-related mortality, the most frequent causes of death were respiratory
failure and neurological disease, while for patients with pneumonia-unrelated
mortality, the most frequent causes of death were malignancy and cardiac
disease.
In this study, slightly more than half of the
deaths were classified as pneumonia related, and more than 75% of the
pneumonia-related deaths occurred within the first 30 days after presentation.
After 30 days, the number of pneumonia-related deaths diminished rapidly, with
less than 15% of all pneumonia-related deaths occurring after 45 days. In
contrast, most pneumonia-unrelated deaths occurred between 30 and 90 days after
presentation, with only 10% occurring within the first 2 weeks of presentation.
These findings suggest that community-acquired pneumonia has a stronger
association with mortality within 45 days of presentation and that prognosis
beyond this point is more heavily influenced by the patient's age, sex, and
other significant comorbid conditions.
We also found that the independent predictors of
pneumonia-related and pneumonia-unrelated mortality were quite different. For
pneumonia-unrelated mortality, comorbid conditions such as malignancy,
immunosuppression, and dementia were independently associated with mortality.
In contrast, chronic liver disease, a relatively rare condition, was the only
comorbid condition independently associated with pneumonia-related mortality.
In addition, for pneumonia-related mortality, acute physiologic or laboratory
derangements, such as hypothermia, decreased white blood cell count, elevated
serum urea nitrogen level, and hypoxemia, were independent predictors of
mortality. For pneumonia-unrelated mortality, systolic hypotension was the only
acute physiologic derangement associated with mortality. Increasing age and
evidence of aspiration were the only risk factors associated with
pneumonia-related and pneumonia-unrelated mortality. Increasing age is a
significant risk factor for mortality, after community-acquired pneumonia,
according to previous studies3, 15 of pneumonia
prognosis. Aspiration events are related to multiple contributing factors that
could affect prognosis, including neurological problems, malnutrition, and
altered mental status.16-18
There are several limitations of this work that
should be acknowledged. First, approximately 130 patients who met study
eligibility were not enrolled because of death before study enrollment.
Therefore, this study may not reflect the full spectrum of patients who died
within 90 days of community-acquired pneumonia. Second, the validity of using a
clinical review committee to determine the cause of death for patients with
community-acquired pneumonia has not been previously established. Although
determining the cause of death by autopsy results represents the reference
standard, autopsies were performed on only 22 of the patients who died, which
limited our ability to assess the accuracy of the assignments of cause of death
by the clinical committee. Nevertheless, this method was chosen because it was
the most practical in nature and likely to provide more reliable data than
death certificate reports. Similar clinical consensus methods have been used to
classify mortality for many other conditions, such as cancer- and
cardiac-related mortality.19 Third, the accuracy
of the case summaries was not independently confirmed by the physician
investigations, which may have affected the assignments of the cause of death.
Fourth, many outpatients had missing data for physical signs and laboratory
values, which may have affected our analyses to determine factors associated
with pneumonia-related and pneumonia-unrelated mortality. However, our
assumption that missing values were normal has been used in our prior validated
models of pneumonia severity. Finally, the moderate number of deaths in this
study may have limited the ability to detect clinical predictors of mortality
and our ability to distinguish differences in the magnitude of effect for
pneumonia-related and pneumonia-unrelated mortality.
In conclusion, this study demonstrates that
there are significant differences between pneumonia-related and
pneumonia-unrelated mortality, including the underlying and immediate causes of
death, the timing of death, and the clinical predictors of death. These
findings suggest that researchers, and those interested in evaluating the
quality of pneumonia care, should use a strategy to differentiate between
pneumonia-related and pneumonia-unrelated mortality. Possible strategies
include using a shorter follow-up (30 days)
or using a clinical review committee to assign the role of community-acquired
pneumonia in the processes leading to death.
Author/Article Information
From the Division of General Internal Medicine, Department of Medicine, and the
Center for Research on Health Care, University of Pittsburgh (Drs Mortensen,
Kapoor, and Fine and Mr Obrosky), and the Center for the Study of Health
Disparities, VA Pittsburgh Healthcare System (Dr Fine), Pittsburgh, Pa; the
General Medicine Unit, Department of Medicine, Massachusetts General Hospital
and Harvard Medical School, Boston (Drs Coley and Singer); and the Division of
Infectious Disease, Department of Medicine, University of Alberta, Edmonton (Dr
Marrie).
Corresponding author and reprints: Michael J. Fine, MD, MSc, Center for the
Study of Health Disparities, VA Pittsburgh Healthcare Systems (Mail Stop
130-U), University Drive C, Location 11E127, Pittsburgh, PA 15240-1001 (e-mail:
[log in to unmask]).
Accepted for publication October 2, 2001.
This study was supported by grant R01 HS06468
from the Agency for Healthcare Research and Quality, Rockville, Md; and grant
F32 HS00135 from the Agency for Healthcare Research and Quality National
Research Service Award (Dr Mortensen).
We thank Karen Lahive, MD, for coordinating
study activities at the Harvard Community Health Plan–Kenmore Center; Terry
Sefcik, MS, for data management; and the following clinical research assistants
for cohort study patient enrollment and data collection: Mary Walsh, RN, Donna
Polenik, RN, MPH, and Kathryn Fine, RN, in Pittsburgh; Mary Ungaro, RN, Leila
Haddad, AB, and Marian Hendershot, RN, in Boston; and Rhonda Grandy, RN, Jackie
Cunning, RN, Dawn Menon, GN, Linda Kraft, RN, and Maxine Young, RN, in Halifax.
REFERENCES
1.
Adams P, Hendershot G, Marano M.
Current estimates from the National Health Interview Survey, 1996.
Vital Health Stat 10.
1999;No. 200.
2.
Pneumonia and influenza death rates: United States, 1979-1994.
MMWR Morb Mortal Wkly Rep.
1995;44:535-537.
MEDLINE
3.
Fine MJ, Smith MA, Carson CA, et al.
Prognosis and outcomes of patients with community-acquired pneumonia: a
meta-analysis.
JAMA.
1996;275:134-141.
MEDLINE
4.
Marrie TJ, Durant H, Yates L.
Community-acquired pneumonia requiring hospitalization: 5-year prospective
study.
Rev Infect Dis.
1989;11:586-599.
MEDLINE
5.
Davis RB, Iezzoni LI, Phillips RS, Reiley P, Coffman GA, Safran C.
Predicting in-hospital mortality: the importance of functional status
information.
Med Care.
1995;33:906-921.
MEDLINE
6.
Hasley PB, Albaum MN, Li YH, et al.
Do pulmonary radiographic findings at presentation predict mortality in
patients with community-acquired pneumonia?
Arch Intern Med.
1996;156:2206-2212.
MEDLINE
7.
Fine MJ, Stone RA, Singer DE, et al.
Processes and outcomes of care for patients with community-acquired pneumonia:
results from the Pneumonia Patient Outcomes Research Team (PORT) cohort study.
Arch Intern Med.
1999;159:970-980.
ABSTRACT
| FULL TEXT
| PDF
| MEDLINE
8.
Albaum MN, Hill LC, Murphy M, et al, for the PORT Investigators.
Interobserver reliability of the chest radiograph in community-acquired
pneumonia.
Chest.
1996;110:343-350.
MEDLINE
9.
Groskin SA.
Heitzman's the Lung: Radiologic-Pathologic
Correlations.
Vol 3. St Louis, Mo: Mosby–Year Book Inc; 1993.
10.
Fine MJ, Auble TE, Yealy DM, et al.
A prediction rule to identify low-risk patients with community-acquired
pneumonia.
N Engl J Med.
1997;336:243-250.
MEDLINE
11.
Manual of the International Statistical
Classification of Diseases, Injuries, and Causes of Death.
Geneva, Switzerland: World Health Organization; 1977.
12.
Kalbefleisch J, Prentice R.
The Statistical Analysis of Failure Time
Data.
New York, NY: John Wiley & Sons Inc; 1980.
13.
Conover W.
Practical Nonparametric Statistics.
New York, NY: John Wiley & Sons Inc; 1980.
14.
Hoyert DL, Kochanek KD, Murphy SL.
Deaths: final data for 1997.
Natl Vital Stat Rep.
1999;47:1-104.
MEDLINE
15.
Fine MJ, Hanusa BH, Lave JR, et al.
Comparison of a disease-specific and a generic severity of illness measure for
patients with community-acquired pneumonia.
J Gen Intern Med.
1995;10:359-368.
MEDLINE
16.
Horner J, Alberts MJ, Dawson DV, Cook GM.
Swallowing in Alzheimer's disease.
Alzheimer Dis Assoc Disord.
1994;8:177-189.
MEDLINE
17.
Horner J, Massey EW, Brazer SR.
Aspiration in bilateral stroke patients.
Neurology.
1990;40:1686-1688.
MEDLINE
18.
McDonald AM, Dietsche L, Litsche M, et al.
A retrospective study of nosocomial pneumonia at a long-term care facility.
Am J Infect Control.
1992;20:234-238.
MEDLINE
19.
Julian DG, Camm AJ, Frangin G, et al, for the European Myocardial Infarct
Amiodarone Trial Investigators.
Randomised trial of effect of amiodarone on mortality in patients with
left-ventricular dysfunction after recent myocardial infarction: EMIAT.
Lancet.
1997;349:667-674.
MEDLINE
Edward E.
Rylander, M.D.
Diplomat American
Board of Family Practice.
Diplomat American
Board of Palliative Medicine.