The Veterans Short Form 36
Questionnaire Is Predictive of Mortality
and Health-Care Utilization in a
Population of Veterans With a SelfReported Diagnosis of Asthma or
COPD*
Mark D. Sprenkle, MD, MS; Dennis E. Niewoehner, MD; David B. Nelson, PhD;
and Kristin L. Nichol, MD, MPH, MBA
Study objective: Measures of health-related quality of life (HRQL) correlate with disease stage in
persons with COPD. However, as their predictive capacity for mortality or medical utilization is
less well defined, we sought to examine the relationship of a general measure of HRQL and
outcomes in persons with obstructive lung disease.
Design: Prospective cohort study.
Setting: Upper Midwest Veterans Integrated Service Network (VISN)-13.
Participants: All veterans in VISN-13 (n ⴝ 70,017) were surveyed with the Veterans Short Form
36 (SF-36V). Persons with reported asthma or COPD who completed the SF-36V formed the study
cohort (n ⴝ 8,354).
Measurements and results: For purposes of analysis, individuals were divided into quartiles of
HRQL according to their physical component summary (PCS) and mental component summary
(MCS), values derived from the SF-36V. Outcomes of mortality, hospitalization, and outpatient
visits were recorded for 12 months after the survey. Outpatient utilization was dichotomized into
high vs low use, with high use being defined as the upper quartile of visits in the 12 months prior
to survey mailing. The study cohort had a mean age of 65 years and was largely male (95%), both
consistent with a veteran population. After correcting for potential confounding factors through
multivariable regression, the PCS was independently predictive of death, hospitalization, and
high outpatient utilization. When using the first quartile of PCS as the reference population,
those in the fourth quartile of PCS had a hazard ratio for death of 5.47 (95% confidence interval
[CI], 3.63 to 8.26). Similarly, the odds ratios for hospitalization, high primary care visits, and high
specialty medicine visits in the fourth quartile of PCS were 1.82 (95% CI, 1.51 to 2.19), 1.54 (95%
CI, 1.26 to 1.87), and 1.46 (95% CI, 1.21 to 1.78), respectively. The MCS, through multivariable
regression, was predictive of death but unassociated with subsequent hospitalization or high
outpatient utilization.
Conclusion: HRQL, as assessed by the SF-36V, is an independent predictor of mortality,
hospitalization, and outpatient utilization in persons with self-reported obstructive lung disease.
(CHEST 2004; 126:81– 89)
Key words: COPD; health status; hospitalization; mortality
Abbreviations: CI ⫽ confidence interval; HRQL ⫽ health-related quality of life; MCS ⫽ mental component summary;
PCS ⫽ physical component summary; SF-36V ⫽ Veterans Short Form 36; VA ⫽ Veterans Affairs; VISN-13 ⫽ Upper
Midwest Veterans’ Integrated Service Network
in which patients perceive, manage,
T heandmanner
adapt to chronic respiratory diseases varies
significantly. An individual’s personal experience
with a disease is modified by many factors, including
disease severity, emotional coping skills, social support systems, comorbid disease, and global functional capacity.1–3 Numerous disease-specific4,5 and
general measures6 –9 of health-related quality of life
(HRQL) have been developed to quantify the experience individuals have with their health. Indeed, the
past 2 decades have seen a growing body of literature
on the relationship between HRQL and obstructive
lung disease.
Given the prognostic importance of lung function
www.chestjournal.org
Downloaded From: http://publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/22011/ on 06/16/2017
CHEST / 126 / 1 / JULY, 2004
81
in obstructive lung disease, many studies3,10 –12 examining HRQL in obstructive lung disease have
focused on their relationship to physiologic measures
of pulmonary or exercise capacity. In general, these
studies have shown a statistically significant correlation. However, typically this correlation is not strong,
suggesting that these physiologic tests and measures
of HRQL may quantify different aspects of disease.
While studies13–15 examining impaired functional
status of persons with COPD have found an association with mortality, fewer studies16,17 have attempted to validate HRQL measures regarding clinically important outcomes such as mortality or
medical utilization.
We postulated that a general measure of HRQL,
the short form 36-item questionnaire,18 would be
independently predictive of mortality and healthcare utilization. To evaluate this hypothesis, we
utilized existing data from a large population-based
survey performed through a Veterans Affairs (VA)
medical network. After identifying individuals with a
self-reported diagnosis of obstructive lung disease
and quantifying their HRQL, we prospectively examined outcomes of mortality, hospitalization, and
outpatient visits.
Materials and Methods
This study was approved by the institutional review board of
the Minneapolis Veterans Affairs Medical Center.
Study Population
The study population was drawn from veterans who received
care through the Upper Midwest Veterans’ Integrated Service
Network (VISN)-13. VISN-13 was a network of VA hospitals in
North Dakota, South Dakota, and Minnesota. VISN-13 subsequently merged with VISN-12, containing care centers in Iowa,
to form VISN-23. Through VISN-13, veterans could receive care
at five medical centers: Minneapolis, MN; St. Cloud, MN; Fargo,
ND; Sioux Falls, SD; and Ft. Meade, SD. All veterans in
VISN-13 with a medical encounter at a VA facility between
October 1, 1997, and March 31, 1998, and valid mailing address
were sent a survey containing the short form-36 and a checklist of
*From the Division of Pulmonary and Critical Care Medicine
(Dr. Sprenkle), Hennepin County Medical Center; and the
Department of Medicine (Drs. Nichol and Nelson) and Division
of Pulmonary and Critical Care (Dr. Niewoehner), Veterans
Affairs Medical Center, Minneapolis, MN.
Presented, in part, in abstract form at the American Thoracic
Society International Meeting in Seattle, WA, May 18, 2003.
Funding was provided by the Veterans Affairs Upper-Midwest
Veterans Service Network.
Manuscript received September 18, 2003; revision accepted
February 10, 2004.
Reproduction of this article is prohibited without written permission from the American College of Chest Physicians (e-mail:
[email protected]).
Correspondence to: Mark D. Sprenkle, MD, MS, Pulm Med 865B,
Hennepin County Medical Center, 701 Park Ave, Minneapolis,
MN 55415; e-mail [email protected]
six common medical conditions. The initial mailing took place on
August 21, 1998. Those persons who had not responded to the
initial mailing by October 1, 1998, were sent a second survey on
October 30, 1998. Those persons who returned a completed
survey with a self-reported diagnosis of obstructive lung disease
formed our study cohort.
Questionnaire
HRQL was assessed by an adapted form of the short form-36,
the Veterans Short Form 36 (SF-36V).19 This questionnaire was
specifically adapted for use in veterans receiving care in an
ambulatory setting. The SF-36V measures eight dimensions of
health status including physical functioning, role limitations due
to physical problems, bodily pain, general health, energy/vitality,
social functioning, role limitations due to emotional problems,
and mental health. These eight dimensions can be summarized
numerically into the physical component summary (PCS) and the
mental component summary (MCS). The PCS and MCS are
standardized to the US population and range between 0 and 100,
with higher scores indicated better HRQL. In addition, the
questionnaire included a list of six medical conditions: (1) asthma,
emphysema, or chronic bronchitis; (2) hypertension; (3) diabetes
mellitus; (4) arthritis; (5) myocardial infarction or history of
angina; and (6) depression. Certain demographic characteristics,
including age, educational level, race, and smoking status, were
also ascertained through the survey.
Data Collection and Outcome Variables
In addition to the mailed survey, certain demographic variables
were ascertained through the VISN-13 patient database (ie,
gender, employment status, marital status, and percentage of
service connection). Outcomes were assessed for a 1-year period
after the survey date. Medical utilization was measured as the
number of hospitalizations, primary care visits, and specialty
medicine visits. Death from any cause was determined from the
Beneficiary Identification and Record Locator System for the
year following the date of survey. Beneficiary Identification and
Record Locator System data have been shown to capture 95 to
98% of deaths.20,21
Statistical Analysis
All statistical analyses were performed using a statistical software package (SAS, version 7.0; SAS Institute; Cary, NC). For
purposes of analysis, the study cohort was divided into quartiles
of PCS and MCS. In simple univariate analyses, Pearson ␹2 tests
and a one-way analyses of variance were used to assess the ability
of the PCS and MCS to predict subsequent categorical and
continuous outcomes, respectively, where the population was
partitioned into groups according to the PCS and MCS quartiles.
Those persons with the best HRQL (ie, first quartile) served as
the reference category. The Mantel test for trend was used to
assess for a linear component to the trend across quartiles of PCS
and MCS in the categorical outcome variable of any prior
hospitalization in the year prior to survey mailing. Assuming the
results of the F test in the one-way analysis were statistically
significant, multiple comparisons were performed against the
reference population using the Bonferoni multiple comparisons
correction method.
In order to assess the independent predictive capacity of
HRQL on mortality and medical utilization, multivariable regression analyses were performed for each outcome measure where
the regression models included the PCS and MCS quartile
groups and demographic variables comprising age (⬍ 60 years,
82
Downloaded From: http://publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/22011/ on 06/16/2017
Clinical Investigations
60 to 70 years, 70 to 80 years, ⬎80 years), gender, education
level, employment status, marital status, percentage of service
connection, and smoking status as explanatory variables. Comorbidity was modeled both as a count of the number of selfreported diseases, as well as by including individually the three
disease states found to be associated with an increased risk of
death and hospitalization by univariate logistic regression (ie,
diabetes mellitus, depression, myocardial infarction, or history of
angina). As both representations of comorbidity resulted in
substantively similar results in the multivariable models, we
elected to use the count of comorbidities in the final models. In
addition, hospitalization, number of primary care visits, and
number of specialty medicine visits in the year prior to the survey
mailing were included in order to account for a potential
association with prior utilization.
A Cox proportional-hazards survival analysis was used to model
mortality and to assess the independent predictive capacity of
PCS and MCS quartile for mortality. In an examination of a
potential interaction between follow-up time and PCS/MCS
quartile, we found no evidence against the assumption of proportional hazards. Logistic regressions were constructed to model
hospitalization and outpatient visits. The Hosmer-Lemeshow
statistic was used with the logistic regression models to assess
model fit. Primary care and specialty medicine utilization was
dichotomized into high vs low use, with high use defined as
utilization surpassing the upper quartile of visits in the 12 months
prior to the survey mailing. In the Cox proportional hazards and
logistic regression analyses, backward elimination model selection procedures were implemented to generate parsimonious
predictive models. Covariates were eliminated at a p value of 0.05
when using backward elimination. Given the established prognostic importance of age and gender in obstructive lung disease,
these covariates were maintained in all multivariable regression
models.
Results
Survey Response and Demographic Characteristics
The initial mailing to patients with a recent medical encounter at one of the VISN-13 medical centers
encompassed a group of 70,674 veterans (Fig 1). A
total of 40,605 individuals responded to the survey, a
response rate of 58.0% after excluding those who
had died or no longer had a valid mailing address. Of
this population, 10,135 persons had a self-reported
diagnosis of asthma, COPD, or chronic bronchitis. In
this group of respondents, 1,781 persons did not
complete the SF-36V questionnaire adequately to
compute a PCS or MCS. These individuals were
excluded, forming the final study cohort of 8,354
persons.
The demographic characteristics of the study population are consistent with a veteran population from
the northern Midwest (Table 1). The mean age of
the study cohort was 65.0 years with a large preponderance of men. Individuals reported a high number
of comorbidities, with a mean value (including obstructive lung disease) of 2.79. The majority of our
study population was white, educated to high school
or beyond, and retired or not employed.
Figure 1. Flowsheet of survey response and formation of the
study cohort.
Univariate Analysis
In univariate analysis, there were statistically significant associations between in almost all demographic characteristics and the PCS or MCS measures (Tables 2, 3). The age and number of
comorbidities of the study population increased with
worsening HRQL. In addition, a worse PCS or MCS
was a marker for those who are not employed or have
an educational level below graduation from high
school. Prior health-care utilization was strongly
associated with HRQL. Worsening quartile of PCS
and MCS was associated, in a stepwise fashion, with
a higher rate of hospitalization in the year prior to
study entry. Similar associations of prior increased
outpatient medical utilization with worse HRQL
were present by quartile of both PCS and MCS.
www.chestjournal.org
Downloaded From: http://publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/22011/ on 06/16/2017
CHEST / 126 / 1 / JULY, 2004
83
Table 1—Demographics of the COPD Cohort*
Variables
Mean age, yr
Age groups, yr
⬍ 60
60–70
70–80
⬎ 80
Gender
Male
Female
Ethnic group
Black
White
Education
⬍ 12th grade
High school or beyond
Employment status
Employed
Retired or not employed
Current marital status
Married
Not married
Comorbidities
Arthritis
Depression
Diabetes
Hypertension
Ischemic heart disease
PCS
MCS
Hospitalizations, No.†
Primary care visits, median (IQR)‡
Specialty medicine visits,‡ median (IQR)‡
Data
65.0 ⫾ 12.6
2,537 (30.5)
2,250 (27.1)
2,953 (35.5)
577 (6.9)
7,840 (95.6)
362 (4.4)
790 (9.5)
7,564 (90.5)
2,750 (32.9)
5,604 (67.1)
2,273 (27.2)
6,081 (72.8)
5,221 (62.5)
3,133 (37.5)
2.79 ⫾ 1.03
4,987 (63.2)
2,769 (34.6)
1,553 (19.5)
3,298 (42.0)
3,567 (45.5)
30.5 ⫾ 10.3
43.9 ⫾ 12.6
0.16 ⫾ 0.37
3.0 (4.0)
0.0 (2.0)
*Data are presented as mean ⫾ SD or No. (%) unless otherwise
indicated; IQR ⫽ interquartile range.
†Mean number of hospitalizations in the 12 months prior to survey
mailing date.
‡Median No. (interquartile range) of outpatient visits in the 12
months prior to survey mailing.
During the period of follow-up, there were 429
deaths (5.1%) in the study cohort. Hospitalization
occurred at least once in 1,407 of the entire population (16.8%). In unadjusted models, a statistically
significant association was found between the quartile of PCS and all outcome measures (Tables 4, 5).
The hazard ratio for mortality was 8.72 (95% confidence interval [CI], 5.81 to 13.1) in the lowest
quartile of PCS when compared to those persons in
the highest quartile. The risk of any hospitalization
was also more likely with worsening PCS (odds ratio,
2.62; 95% CI, 2.21 to 3.11 when comparing fourth to
first quartiles of PCS). Similarly, the odds ratios for
high outpatient primary care and specialty medicine
utilization increased in a stepwise fashion with worsening quartile of PCS. Persons with the worst mental
health status, as assessed by quartile of MCS, had
higher mortality, rate of hospitalization, and primary
care utilization by univariate analyses (Tables 4, 5).
When comparing the lowest quartile of MCS to the
highest, the hazard ratio for mortality was 2.50 (95%
CI, 1.88 to 3.33). Hospitalization occurred more
frequently in those with lower MCS (odds ratio, 1.63;
95% CI, 1.38 to 1.92 when comparing fourth to first
quartiles of MCS). While a statistically significant
association was noted between the MCS and high
primary care utilization, no association was found
between MCS and high specialty medicine visits in
univariate analyses.
Multivariable Analysis
Analysis by a Cox proportional hazards model
revealed the PCS and MCS to be independent
predictors of mortality (Table 4). With PCS modeled
as a continuous variable, a 5-point decrease in the
PCS resulted in a 39% increased risk of death
(p ⬍ 0.0001). When analyzed by PCS quartile, with
the first quartile of PCS as the reference population,
a hazard ratio of 5.47 (95% CI, 3.63 to 8.26) was
found for the fourth quartile. Similarly, the hazard
ratio for mortality increased with worsening quartile
of MCS.
An independent relationship between HRQL, as
assessed by PCS, and health-care utilization was
noted in multivariable logistic regression (Table 5).
The risk of hospitalization was significantly increased
for all quartiles of PCS in comparison to the reference population. When examining PCS as a continuous variable, a 5-point decrease results in a 13%
increased risk of hospitalization. The risk of high
outpatient utilization was, in general, statistically
significant. With the first quartile serving as the
reference population, those individuals in the fourth
quartile of PCS had odds ratios of 1.54 (95% CI, 1.26
to 1.87) and 1.46 (95% CI, 1.21 to 1.78) for high
primary care and specialty medicine utilization, respectively. We found no association between MCS
and health-care utilization after correcting for potential confounding factors (Table 5).
Nonresponders
Of the entire eligible population of veterans
mailed both surveys, a total of 29,412 persons did not
respond. Demographic characteristics of the entire
population of responders and nonresponders are
presented in Table 6. Nonresponders, in general,
were younger and less likely to be married. In the
year prior to survey mailing, nonresponders had
fewer outpatient visits but more hospitalizations than
respondents. In addition, mortality during the year
following survey mailing was higher in nonresponders.
84
Downloaded From: http://publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/22011/ on 06/16/2017
Clinical Investigations
Table 2—Univariate Analysis of Demographic Characteristics and Previous Health-Care Utilization
by PCS Quartile*
Variables
Quartile 1: ⬎ 37
Quartile 2: 29–37
Quartile 3: 23–29
Quartile 4: ⬍ 23
Mean age, yr
Gender
Male
Female
Ethnicity
Black
White
Comorbidities
Education
⬍ 12th grade
High school or greater
Employment status
Employed
Retired or not employed
Marital status
Married
Not married
Hospitalization㛳
Primary care visits㛳
Specialty medicine visits㛳
60.8 ⫾ 14.0
64.9 ⫾ 12.5
66.4 ⫾ 11.5
67.9 ⫾ 11.1
⬍ 0.0001†
1,907 (24.3)
141 (39.0)
1,960 (25.0)
91 (25.1)
1,990 (25.4)
64 (17.7)
1,983 (25.3)
66 (18.2)
⬍ 0.0001‡
174 (22.0)
1,913 (25.3)
2.3 ⫾ 1.0
206 (26.1)
1,885 (24.9)
2.8 ⫾ 1.0
213 (27.0)
1,876 (24.8)
3.0 ⫾ 1.0
197 (24.9)
1,890 (25.0)
3.1 ⫾ 1.0
0.19‡
⬍ 0.0001†
467 (17.0)
1,620 (28.9)
647 (23.5)
1,444 (25.8)
782 (28.4)
1,307 (23.3)
854 (31.1)
1,233 (22.0)
⬍ 0.0001‡
873 (38.4)
1,214 (20.0)
577 (25.4)
1,514 (24.9)
445 (19.6)
1,644 (27.0)
378 (16.6)
1,709 (28.1)
⬍ 0.0001‡
1,161 (22.2)
926 (29.5)
201 (9.6)
2.6 ⫾ 4.1
1.0 ⫾ 2.9
1,305 (25.0)
786 (25.1)
285 (13.6)
3.2 ⫾ 3.3
1.5 ⫾ 3.5
1,366 (26.2)
723 (23.1)
421 (20.2)
4.1 ⫾ 4.3
1.8 ⫾ 3.2
1,389 (26.6)
698 (22.3)
462 (22.1)
4.1 ⫾ 4.3
1.9 ⫾ 3.8
⬍ 0.0001‡
⬍ 0.0001§
⬍ 0.0001†
⬍ 0.0001†
p Value
*Data are presented as mean ⫾ SD or No. (%).
†One-way analysis of variance used to compare PCS quartiles of continuous variables.
‡␹2 test used to compare PCS quartiles of categorical variables.
§Mantel’s extended test for trend used to assess linear trend between PCS quartile and hospitalization.
㛳Hospitalization and number of primary care/specialty visits assessed in the 1-year period prior to survey mailing.
Discussion
Our results indicate that HRQL is a useful predictor of outcomes in persons with a self-reported
diagnosis of obstructive lung disease. Our study
cohort included an elderly, predominantly male population of 8,354 veterans. In this study population,
we found a statistically significant, stepwise association between the quartile of PCS and all outcome
measures. In multivariable analysis, HRQL was
found to be an independent predictor of mortality
and medical utilization. Of the outcome measures we
analyzed, we found mortality to have the strongest
association with HRQL (odds ratio, 5.47, fourth
quartile vs first quartile). However, a significant
association was also present between risk of hospitalization and HRQL, with 82% greater odds of
hospitalization in the fourth quartile of PCS compared to the reference population after adjustment
for confounding factors. Somewhat smaller, though
significantly increased, risks were present when examining high outpatient primary care and specialty
medicine utilization. Interestingly, while both the
PCS and MCS were independently predictive of
mortality, the MCS did not prove to be associated
with increased risk of hospitalization or high outpatient utilization.
There is little information in the medical literature
on HRQL and clinical outcomes in obstructive lung
disease. There have been a few investigations examining the predictive capacity of functional status
measures in persons with COPD.13–15 These studies
all found that functional status was independently
predictive of clinically important end points such as
mortality or weaning from mechanical ventilation.
However, tools to assess functional status and measures of HRQL assess different aspects of daily living
as it relates to health.1 Functional capacity measures
are focused on the ability of an individual to perform
activities of daily living, while measures of HRQL are
more subjective measures of the multiple ways (ie,
physical and emotional) that health impacts daily life.
While one study16 did find an increased risk of
mortality in those with poor quality of life, this study
measured quality of life using a global rating scale of
“excellent” to “poor.”
Fan et al17 performed the only previous study to
examine a validated, systematic measure of HRQL to
assess the risk mortality and medical utilization in
persons with obstructive lung disease. They examined a population of persons enrolled at general
internal medicine clinics at seven VA hospitals geographically distributed across the United States.
Their primary analysis involved the predictive capacity of a disease-specific HRQL measure, the Seattle
Obstructive Lung Disease Questionnaire, in determining subsequent mortality or hospitalization over a
www.chestjournal.org
Downloaded From: http://publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/22011/ on 06/16/2017
CHEST / 126 / 1 / JULY, 2004
85
Table 3—Univariate Analysis of Demographic Characteristics and Previous Health-Care Utilization by
MCS Quartile*
Variables
Quartile 1: ⬎ 54
Quartile 2: 44–54
Quartile 3: 35–43
Quartile 4: ⬍ 35
Mean age, yr
Gender
Male
Female
Ethnicity
Black
White
Comorbidities
Education
⬍ 12th grade
High school or greater
Employment status
Employed
Retired or not employed
Marital status
Married
Not married
Hospitalization㛳
Primary care visits㛳
Specialty medicine visits㛳
61.7 ⫾ 13.2
66.7 ⫾ 12.1
65.6 ⫾ 12.5
65.9 ⫾ 12.1
⬍ 0.0001†
1,955 (24.9)
101 (27.9)
1,951 (24.9)
96 (26.5)
1,971 (25.1)
72 (19.9)
1,963 (25.1)
93 (25.7)
0.14‡
152 (19.2)
1,938 (25.6)
2.3 ⫾ 1.0
173 (21.9)
1,913 (25.3)
2.8 ⫾ 1.0
203 (25.7)
1,888 (25.0)
3.0 ⫾ 1.0
262 (33.2)
1,825 (24.1)
3.1 ⫾ 1.0
⬍ 0.0001‡
⬍ 0.0001†
577 (21.0)
1,513 (27.0)
692 (25.2)
1,394 (24.9)
796 (28.9)
1,295 (23.1)
685 (24.9)
1,402 (25.0)
⬍ 0.0001‡
662 (29.1)
1,428 (23.5)
649 (28.6)
1,437 (23.6)
475 (20.9)
1,616 (26.6)
487 (21.4)
1,600 (26.3)
⬍ 0.0001‡
1,389 (26.6)
701 (22.4)
254 (12.2)
3.1 ⫾ 3.3
1.5 ⫾ 3.5
1,353 (25.9)
733 (23.4)
283 (13.6)
3.5 ⫾ 3.9
1.4 ⫾ 2.8
1,308 (25.1)
783 (25.0)
363 (17.4)
3.6 ⫾ 3.8
1.6 ⫾ 3.4
1,171 (22.4)
916 (29.2)
469 (22.5)
3.9 ⫾ 5.1
1.7 ⫾ 3.7
⬍ 0.0001‡
⬍ 0.0001§
⬍ 0.0001†
0.02†
p Value
*Data are presented as mean ⫾ SD or No. (%).
†One-way analysis of variance used to compare MCS quartiles of continuous variables.
‡␹2 test used to compare MCS quartiles of categorical variables.
§Mantel extended test for trend used to assess linear trend between MCS quartile and hospitalization.
㛳Hospitalization and number of primary care/specialty visits assessed in the 1-year period prior following survey mailing.
1-year period of follow-up. Additionally analyses
were performed based on participant response to the
SF-36V.
The present study both supports and expands on
the study performed by Fan et al.17 Both studies
found a statistically significant, independent association between HRQL and mortality or hospitalization.
Though both studies enrolled a population of veter-
ans, some differences in the study populations exist.
Though our study was geographically limited to
Northern-Midwest states, we surveyed a broader
sample of veterans, enrolling all persons seen recently throughout a Veterans’ Integrated Service
Network. This difference in survey administration
allowed for inclusion of persons outside of general
internal medicine clinics. In addition, we were able
Table 4 —Cox Proportional Hazards Model of the Relationship Between the PCS and MCS and Mortality
Mortality
Unadjusted model (univariate)
PCS, continuous
(5-point decrease)
PCS, 1st quartile
PCS, 2nd quartile
PCS, 3rd quartile
PCS, 4th quartile
Adjusted model
(multivariable)†‡
PCS, continuous
(5-point decrease)
PCS, 1st quartile
PCS, 2nd quartile
PCS, 3rd quartile
PCS, 4th quartile
Mortality
Hazard Ratio
95% CI
p Value*
Hazard Ratio
95% CI
p Value*
1.49
1.40–1.58
⬍ 0.0001
1.14
1.09–1.18
⬍ 0.0001
Reference
0.78–1.53
1.50–2.71
1.88–3.33
Reference
0.59
⬍ 0.0001
⬍ 0.0001
1.11–1.21
⬍ 0.0001
Reference
0.75–1.48
1.30–2.36
1.92–3.44
Reference
0.76
0.0002
⬍ 0.0001
Reference
3.23
4.08
8.72
1.39
Reference
2.51
2.79
5.47
Reference
2.08–5.01
2.65–6.27
5.81–13.1
MCS, continuous
(5-point decrease)
Reference MCS, 1st quartile
⬍ 0.0001 MCS, 2nd quartile
⬍ 0.0001 MCS, 3rd quartile
⬍ 0.0001 MCS, 4th quartile
1.30–1.48
⬍ 0.0001
Reference
1.61–3.91
1.81–4.31
3.63–8.26
MCS, continuous
(5-point decrease)
Reference MCS, 1st quartile
⬍ 0.0001 MCS, 2nd quartile
⬍ 0.0001 MCS, 3rd quartile
⬍ 0.0001 MCS, 4th quartile
Reference
1.10
2.02
2.50
1.16
Reference
1.05
1.75
2.57
*p Value represents the statistical comparison to the reference population when comparing quartiles of PCS or MCS.
†Covariates included in the adjusted model (PCS): prior hospitalization, age, gender, employment status.
‡Covariates included in the adjusted model (MCS): prior hospitalization, prior specialty medicine visit, age, gender, employment status.
86
Downloaded From: http://publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/22011/ on 06/16/2017
Clinical Investigations
Table 5—Univariate and Multivariate Logistic Regression Models of the Relationship Between the PCS and MCS
and the Risk of Hospitalization or Primary Care/Specialty Medicine Visits
Hospitalization†‡
Variables
Unadjusted model
(univariate)
PCS, continuous
(5-point decrease)
PCS, 1st quartile
PCS, 2nd quartile
PCS, 3rd quartile
PCS, 4th quartile
Adjusted model
(multivariable)
PCS, continuous
(5-point decrease)
PCS, 1st quartile
PCS, 2nd quartile
PCS, 3rd quartile
PCS, 4th quartile
Unadjusted model
(univariate)
MCS, continuous
(5-point decrease)
MCS, 1st quartile
MCS, 2nd quartile
MCS, 3rd quartile
MCS, 4th quartile
Adjusted model
(multivariable)
MCS, continuous
(5-point decrease)
MCS, 1st quartile
MCS, 2nd quartile
MCS, 3rd quartile
MCS, 4th quartile
Primary Care Visits§㛳
Specialty Medicine Visits¶#
Risk Ratio
95% CI
p Value*
Risk Ratio
95% CI
p Value*
Risk Ratio
95% CI
p Value*
1.22
1.18–1.25
⬍ 0.0001
1.21
1.17–1.24
⬍ 0.0001
1.18
1.14–1.21
⬍ 0.0001
Reference
1.43
1.86
2.62
Reference
1.19–1.72
1.56–2.22
2.21–3.11
Reference Reference
⬍ 0.0001
1.48
⬍ 0.0001
2.26
⬍ 0.0001
2.57
Reference
1.23–1.78
1.90–2.69
2.16–3.05
Reference Reference
⬍ 0.0001
1.69
⬍ 0.0001
2.16
⬍ 0.0001
2.32
Reference
1.43–2.00
1.83–2.54
1.97–2.73
Reference
⬍ 0.0001
⬍ 0.0001
⬍ 0.0001
1.13
1.09–1.17
⬍ 0.0001
1.05–1.13
⬍ 0.0001
1.03–1.11
0.0001
Reference
1.22
1.29
1.82
Reference
1.00–1.48
1.06–1.56
1.51–2.19
Reference Reference
0.050
1.14
0.010
1.39
⬍ 0.0001
1.54
Reference
0.93–1.40
1.14–1.69
1.26–1.87
Reference Reference
0.195
1.41
0.0011
1.43
⬍ 0.0001
1.46
Reference
1.16–1.71
1.17–1.73
1.21–1.78
Reference
0.0006
0.0004
0.0001
1.08
1.06–1.11
⬍ 0.0001
1.03–1.07
⬍ 0.0001
0.99–1.04
Reference
1.02
1.60
1.63
Reference
0.85–1.22
1.35–1.88
1.38–1.92
Reference Reference
0.840
1.07
⬍ 0.0001
1.30
⬍ 0.0001
1.37
Reference
0.91–1.26
1.11–1.53
1.17–1.61
Reference Reference
0.43
0.92
0.0013
1.11
⬍ 0.0001
1.08
1.03
1.00–1.06
Reference
0.90
1.28
1.12
Reference
0.74–1.09
1.07–1.54
0.93–1.37
0.047
1.09
1.05
1.00
0.97–1.02
Reference Reference
0.27
0.93
0.009
1.01
0.24
0.92
Reference
0.77–1.13
0.83–1.22
0.76–1.12
0.75
1.07
1.02
0.98
Reference Reference
0.47
0.86
0.96
0.94
0.42
0.86
Reference
0.79–1.08
0.96–1.29
0.93–1.26
0.95–1.01
Reference
0.71–1.03
0.78–1.13
0.71–1.05
0.14
Reference
0.30
0.16
0.31
0.11
Reference
0.10
0.52
0.13
*Represents the statistical comparison to the reference population when comparing quartiles of PCS or MCS.
†Adjusted model covariates, PCS: prior hospitalization, prior primary care, prior specialty medicine, age, gender, employment status, marital
status.
‡Adjusted model covariates, MCS: prior hospitalization, prior primary care, prior specialty medicine, age, gender, employment status, marital
status, comorbidity, smoking status.
§Adjusted model covariates, PCS: prior hospitalization, prior primary care, age, gender, employment status, comorbidity.
㛳Adjusted model covariates, MCS: prior hospitalization, prior primary care, age, gender, employment status, comorbidity, smoking status.
¶Adjusted model covariates, PCS: prior hospitalization, prior specialty medicine, age, gender, comorbidity, percent of service connection.
#Adjusted model covariates, MCS: prior hospitalization, prior specialty medicine, age, gender, comorbidity, smoking status, percent of service
connection.
to survey a larger number of persons, resulting in a
larger number of persons available for follow-up
(n ⫽ 3,282 vs n ⫽ 8,354). Finally, we were able to
examine not only mortality and hospitalization, but
also outpatient primary and specialty medicine care
as outcome variables.
Why is it important to examine HRQL measures
in obstructive lung disease when measures of pulmonary capacity can already provide useful prognostic
information in this patient population? It appears
that HRQL measures and spirometry measure different characteristics of chronic lung disease. Studies
examining the relationship between FEV1 and gen-
eral measures of HRQL have found only a weak-tomoderate correlation (correlation coefficients, 0.20
to 0.25).3,10 A smaller study by Mahler and Mackowiak12 also found FEV1 and five of nine components
of the short form-36 to be statistically correlated,
though the correlation between a composite score
(ie, PCS) and FEV1 was not presented.12 As HRQL
measures examine other consequences of disease on
daily living besides exertional capacity, it is possible
that these tests could provide significant prognostic
information, independent of individual pulmonary
function measures. Furthermore, it is possible that
HRQL measures and pulmonary function tests could
www.chestjournal.org
Downloaded From: http://publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/22011/ on 06/16/2017
CHEST / 126 / 1 / JULY, 2004
87
Table 6 —Demographic Characteristics of Responders
and Nonresponders*
Characteristics
Age groups, yr
⬍ 50
50–64
ⱖ 65
Gender
Male
Female
Current marital status
Married
Not married
Unknown
No. of clinic visits‡
0
1 to 9
ⱖ 10
Hospitalization‡
Mortality§
Responders
(n ⫽ 40,605)
Nonresponders
(n ⫽ 29,412)
p
Value†
17.2
24.8
58
37.6
25.3
37.1
⬍ 0.0001
95.8
4.2
95.3
4.7
⬍ 0.01
60.8
32.9
6.3
43.2
48.8
8
⬍ 0.0001
19.9
39.3
40.8
12.6
3.9
34.7
34.2
31.1
13.6
4.7
⬍ 0.0001
⬍ 0.0001
⬍ 0.0001
*Data are presented as %.
†␹2 test.
‡Clinic visits or hospitalizations in the year prior to survey mailing.
§Mortality in the year following survey mailing.
provide additive prognostic information in persons
with obstructive lung disease. Unfortunately, given
the size of our study cohort, we were unable to
systematically perform pulmonary function studies
to examine these questions.
Validation of HRQL measures, both general and
disease specific, is a challenging task.1,22,23 As noted
above, validation of these measures in obstructive
lung disease can be performed through their comparison to well-established measures of disease severity, such as spirometry. In addition, validation can
be supported by significant change in measured
HRQL after administration of therapies, such as
inhaled bronchodilators.24 –26 This study provides an
additional evidence, in the form of independent
prognostic capacity, to support the value of HRQL
measures in obstructive lung disease. Whether this
type of patient information could be used to target
at-risk populations for medical interventions including vaccination, smoking cessation, or pulmonary
rehabilitation requires further investigation.
Our study has a number of limitations. First,
diagnostic misclassification may have occurred. We
relied on a self-reported diagnosis of obstructive lung
disease in order to identify individuals for our study
cohort. Other investigators have examined the concordance between physician-diagnosed chronic lung
disease and patient self-report27–29; these studies
have found very good agreement, with ␬ coefficients
ranging from 0.56 to 0.73. While the size of our study
population is a strength of this investigation, it made
definitive diagnosis of their obstructive lung disease
prohibitive. In addition, our study population included both asthmatics and persons with COPD.
Because of concerns of diagnostic misclassification,
study participants were not asked whether they had
asthma or COPD. Instead, they were asked an
inclusive question: “have you been told by a physician that you have asthma, emphysema, or chronic
bronchitis?” It is possible that the relationship of
HRQL to mortality and medical utilization is different in persons with asthma compared with COPD.
Given the elderly nature of our veteran population, it
is likely that the majority of our study population had
COPD. However we would be unable to distinguish
differences in the prognostic capacity of HRQL
depending on the type of obstructive lung disease
present.
We do not have any data regarding study participants receiving medical care outside of the VA
medical centers enrolled in this study. As a consequence, confounding by differential non-VA hospitalization or outpatient utilization among our study
cohort may be present. VA hospitals often do not
provide emergent medical care, largely due to issues
of poor proximity or lack of emergency medical
facilities. As a consequence, veterans can receive
medical care, if necessary, at community hospitals. It
is probable that persons with the worst HRQL would
be at greater risk of emergent hospitalization or
ambulatory medical evaluation than those with better HRQL. If we accept that persons with impaired
HRQL would be more likely to receive care at a
non-VA hospital or clinic, this ascertainment bias
would lessen the statistical association between
HRQL and the risk of hospitalization or outpatient
visit. As death statistics were collected regardless of
the place of death, the relationship between HRQL
and mortality is unaffected by this potential bias.
Next, residual confounding by inadequate assessment of patient comorbidities could be present.
Diagnostic misclassification could have occurred as a
consequence of patient self-report of comorbidities.
In addition, only a limited number of comorbidities
were assessed through the study questionnaire. Finally, the results of these analyses are valid only for
those persons who were willing to participate in this
survey. While our response rate was reasonably high
at 58.0%, there are clear demographic differences
between responders and nonresponders that preclude generalization of these finding to the entire
eligible veteran population.
In conclusion, we have shown that the assessment
of HRQL in persons with obstructive lung disease
can independently identify populations at risk for
death and increased medical resource utilization.
This information provides yet more evidence to
88
Downloaded From: http://publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/22011/ on 06/16/2017
Clinical Investigations
validate the use of general measures of HRQL in
chronic lung disease. While tools to assess HRQL are
being used widely to assess the therapies in COPD,
further research is needed to discover whether impaired HRQL can be used as a marker to identify
individuals who would benefit from medical intervention or pulmonary rehabilitation.
14
15
16
References
1 Curtis RJ, Deyo RA, Hudson LD. Health-related quality of
life among patients with chronic obstructive pulmonary disease. Thorax 1994; 49:162–170
2 Guthrie SJ, Hill KM, Muers MF. Living with severe COPD:
a qualitative exploration of the experience of patients in
Leeds. Respir Med 2001; 95:196 –204
3 Prigatano GP, Wright EC, Levin D. Quality of life and its
predictors in patients with mild hypoxemia and chronic
obstructive pulmonary disease. Arch Intern Med 1984; 144:
1613–1619
4 Jones PW, Quirk FH, Baveystock CM, et al. A self-complete
measure of health status for chronic airflow limitation: the St.
George’s Respiratory Questionnaire. Am Rev Respir Dis
1992; 145:1321–1327
5 Guyatt G, Berman L, Townsend M, et al. A measure of
quality of life for clinical trials in chronic lung disease. Thorax
1987; 42:773–778
6 Ware JE, Sherbourne CD. The MOS 36-item short-form
health survey (SF-36). Med Care 1992; 30:473– 483
7 Kaplan RM, Atkins CJ, Timms R. Validity of a quality of
well-being scale as an outcome measure in chronic obstructive pulmonary disease. J Chronic Dis 1984; 37:85–95
8 Bergner M, Bobbitt RA, Carter WB, et al. The Sickness
Impact Profile: development and final revision of a health
status measure. Med Care 1981; 19:787– 805
9 Hunt SM, McKenna SP, McEwen J, et al. A quantitative
approach to perceived health status: a validation study. J
Epidemiol Community Health 1980; 34:281–286
10 Ferrer M, Alonso J, Morera J, et al. Chronic obstructive
pulmonary disease stage and health-related quality of life.
Ann Intern Med 1997; 127:1072–1079
11 Harper R, Brazier JE, Waterhouse JC, et al. Comparison of
outcome measures for patients with chronic obstructive pulmonary disease (COPD) in an outpatient setting. Thorax
1997; 52:879 – 887
12 Mahler DA, Mackowiak JI. Evaluation of the short-form
36-item questionnaire to measure health-related quality of
life in patients with COPD. Chest 1995; 107:1585–1589
13 Strom K. Survival of patients with chronic obstructive pulmo-
17
18
19
20
21
22
23
24
25
26
27
28
29
nary disease receiving long-term domiciliary oxygen therapy.
Am Rev Respir Dis 1993; 147:585–591
Menzies R, Gibbons W, Goldberg P. Determinants of weaning and survival among patients with COPD who require
mechanical ventilation for acute respiratory failure. Chest
1989; 95:398 – 405
Connors AF, Dawson NV, Thomas C, et al. Outcomes
following acute exacerbation of severe chronic obstructive
lung disease. Am J Respir Crit Care Med 1996; 154:959 –967
Lynn JL, Ely EW, Zhong Z, et al. Living and dying with
chronic obstructive pulmonary disease. J Am Geriatr Soc
2000; 48:S91–S100
Fan VS, Curtis JR, Tu SP, et al. Using quality of life to predict
hospitalization and mortality in patients with obstructive lung
diseases. Chest 2002; 122:429 – 436
Ware JE Jr, Sherbourne CD. The MOS 36-item Short Form
Health Survey (SF-36): I. Conceptual framework and item
selection. Med Care 1992; 30:473– 483
Kazis LE, Ren XS, Lee A, et al. Health status in VA patients:
results from the Veterans Health Study. Am J Med Qual
1999; 14:28 –38
Fisher SG, Weber L, Goldberg J, et al. Mortality ascertainment in the veteran population: alternatives to the National
Death Index. Am J Epidemiol 1995; 141:242–250
Fleming C, Fisher E, Chang C, et al. Studying outcomes and
hospital utilization in the elderly. Med Care 1992; 30:377–391
Mahler DA. How should health-related quality of life be
assessed in patients with COPD? Chest 2000; 117:54S–57S
Mahler DA, Jones PW. Measurement of dyspnea and quality
of life in advanced lung disease. Clin Chest Med 1997;
18:457– 469
Guyatt GH, Townsend M, Pugsley SO. Bronchodilators in
chronic airflow limitation. Am Rev Respir Dis 1987; 135:
1069 –1074
Jones PW, Bosh TK. Quality of life changes in COPD patients
treated with salmeterol. Am J Respir Crit Care Med 1997;
155:1283–1289
Mahler DA, Donohue JF, Barbee RA, et al. Efficacy of
salmeterol xinafoate in the treatment of COPD. Chest 1999;
115:957–965
Nichol KL, Korn JE, Baum P. Estimation of outpatient risk
characteristics and influenza vaccination status: validation of a
self-administered questionnaire. Am J Prev Med 1991;
7:199 –203
Katz JN, Chang LC, Sangha O, et al. Can comorbidity be
measured by questionnaire rather than medical record review? Med Care 1996; 34:73– 84
Kriegsman DM, Penninx BW, van Eijk JT, et al. Self-reports
and general practitioner information on the presence of
chronic diseases in community dwelling elderly. J Clin Epidemiol 1996; 49:1407–1417
www.chestjournal.org
Downloaded From: http://publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/22011/ on 06/16/2017
CHEST / 126 / 1 / JULY, 2004
89