Health-related quality of life predicts future health care utilization and mortality in veterans with self-reported physician-diagnosed arthritis: The veterans arthritis quality of life study

Health-Related Quality of Life Predicts Future Health Care Utilization and Mortality in Veterans with Self-Reported Physician-Diagnosed Arthritis: The Veterans Arthritis Quality of Life Study Jasvinder A. Singh, MD, MPH,* David B. Nelson, PhD,† Howard A. Fink, MD, MPH,‡ Kristin L. Nichol, MD, MPH, MBA§ OBJECTIVE To investigate whether health-related quality of life (HRQOL) measures predict health care utilization and mortality in a cohort of veterans with self-reported physiciandiagnosed arthritis. METHODS A cohort of veterans from the Upper Midwest Veterans Integrated Service Network (VISN) was mailed a self-administered questionnaire that was composed of the SF-36V (modified from SF-36 for use in veterans) and questions regarding demographics, current smoking status, limitation of activities of daily living (ADLs), and preexisting physiciandiagnosed medical conditions, including arthritis. Within subjects reporting physiciandiagnosed arthritis, we analyzed the associations between the SF-36V component summary scales (physical and mental component summary, PCS and MCS, respectively) and the occurrence of any hospitalization, number of hospitalizations, number of outpatient visits, and mortality, for the year after survey administration, using multivariable regression analyses. RESULTS Of 34,440 survey responders who answered a question regarding arthritis, 18,464 (58%) subjects reported physician-diagnosed arthritis. Arthritic patients in the lowest tertile of PCS scores had significantly higher odds of any hospitalization (Odds ratio (OR) 1.49, 95% confidence interval (CI) [1.25-1.76]) and mortality (OR 1.69, 95% CI [1.18-2.42]), and a significantly higher number of hospitalizations/year (Rate ratio (RR) 1.09, 95% CI [1.05-1.13]) and outpatient visits/year (RR 1.07, 95% CI [1.03-1.11]). Arthritic patients in the lowest tertile of MCS scores had significantly higher odds of any hospitalization (OR 1.20, 95% CI [1.02-1.41]), mortality (OR 2.14, 95% CI [1.56-2.94]), and a significantly higher number of hospitalizations/year (RR 1.05, 95% CI [1.02-1.09]) and outpatient visits/year (RR 1.07, 95% CI [1.03-1.11]). CONCLUSIONS HRQOL, as assessed by the SF-36V, predicts future inpatient and outpatient health care utilization and mortality in veterans with self-report of physician-diagnosed arthritis. Semin Arthritis Rheum 34:755-765 © 2004 Elsevier Inc. All rights reserved. KEYWORDS arthritis, HRQOL, SF-36, health care utilization, mortality

*Rheumatology Section, Medicine Service, VA Medical Center, Minneapolis, MN and Division of Rheumatology, Department of Medicine, University of Minnesota, Minneapolis, MN. †Center for Chronic Disease Outcomes Research, VA Medical Center, Minneapolis, MN. ‡Medicine Service, Geriatric Research Education & Clinical Center and Chronic Disease Outcomes Research, VA Medical Center, Minn., MN; Department of Medicine, Univ. of Minnesota, Minneapolis, MN. §Medicine Service, Center for Chronic Disease Outcomes Research, VA Medical Center, Minn., MN. Conflict of Interest: None. Grant Support: VA Upper Midwest Veterans Network-VISN13 Grant. Address reprints requests to Jasvinder A Singh, MD, MPH, Minneapolis VA Medical Center (111R), One Veteran’s Drive, Minneapolis, MN 55417. E-mail: [email protected]

0049-0172/05/$-see front matter © 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.semarthrit.2004.08.001

755

J.A. Singh et al.

756

A

rthritis and other rheumatic conditions are among the commonest illnesses in the United States (U.S.). These diseases affected an estimated 70 million people in 2001 (1), cause significant physical and psychological morbidity (2,3), and are the leading cause of disability in the U.S. (4). They also constitute a significant health and financial burden on the U.S. health care system. These conditions led to 744,000 hospitalizations and 44 million ambulatory-care visits in 1997 (5) and cost $149 billion in direct and indirect costs (2.5% of the Gross National Product) in 1992 in the U.S. (6). Due to a continuing increase in health care costs, an increasing number of studies have assessed various risk factors for high health care utilization, including health-related quality of life (HRQOL). HRQOL measures are associated with hospitalization in patients with chronic diseases like asthma, chronic obstructive lung disease, congestive heart failure, and inflammatory bowel disease (7-11). Very few studies have evaluated the relationship of HRQOL and health care utilization in arthritis patients, and all studies found a positive association between poor HRQOL and higher health care utilization (12-16). However, these studies were small with sample sizes of less than 1000 patients (12-16), and all but 1 study (12) selected patients from rheumatology clinics rather than including patients from all clinics. Only 2 studies (12,14) included patients with osteoarthritis, and only 1 study (12) adjusted for prior health care utilization, which is a strong determinant of future health care utilization (12). The Veterans Health Administration (VHA) is the largest integrated health system in the United States. The VHA provides health care to U.S. war veterans, who are predominantly elderly, poor, and medically underserved men (1719). The VHA provided health care to more than 3 million veterans with a medical care budget of $17.9 billion in the fiscal year 1998 (20,21). Arthritis was 1 of the top 5 most common chronic conditions in a sample of the 3.4 million

veterans who received health care in the veteran affairs (VA) system in fiscal year 1999 (22). None of the previous studies in arthritic subjects included veterans, who are known to be sicker (23) and have poorer HRQOL (24) than non-VA patients. We sought to identify whether the physical component summary and mental component summary (PCS and MCS, respectively) of the SF-36, and other demographic and clinical factors, prospectively predicted the risk of increased health care utilization and mortality in a large cohort study of veterans with self-reported physician-diagnosed arthritis who receive health care at VA hospitals and clinics.

Methods Participants Subjects were considered eligible for the present analyses if they were veterans who had completed a mailed survey questionnaire in which they self-reported a physician-diagnosis of arthritis. The self-administered questionnaire was mailed to all veterans in the Upper Midwest Veterans Integrated Service Network (VISN 13) who had at least 1 outpatient encounter or inpatient stay between 10/1/97 and 3/31/98 at a VISN 13 facility and a valid mailing address—VISN 13 Veterans Quality of Life Study (VISN 13 Vet-QOL Study) (25). VISN 13 consisted of a regional network that provided health care to veterans from all of Minnesota, North Dakota, and South Dakota, and selected counties in Iowa, Nebraska, Wisconsin, and Wyoming. An initial mailing was sent in August, 1998, with a second mailing to nonresponders 10 weeks later. The Veterans Arthritis Quality of Life Study (VAQS) was comprised of subjects who reported the presence of physiciandiagnosed arthritis. VISN 13 no longer exists, having merged with another network directly to its south subsequent to the completion of our study, and is now a part of VISN 23 (which includes Iowa, Minnesota, Nebraska, North Dakota, South Dakota, and portions of northern Kansas, Missouri, western Illinois, Wisconsin, and eastern Wyoming).

Survey Questionnaire Abbreviations ADL BIRLS CI HMO HRQOL MCS OPC OR PCS PTF QOL SD SF-36V VA VHA VISN

Activity of Daily Living Beneficiaries Identification Record Locator System Confidence Interval Health maintenance organization health-related quality of life mental component summary Outpatient Clinic Tables Odds Ratio physical component summary Patient Treatment File quality of life Standard Deviation Short-Form 36 modified for veterans Veterans Affairs Veterans Health Administration Veterans Integrated Service Network

The self-administered survey questionnaire consisted of 3 components. The first component comprised questions assessing gender, education level, race/ethnicity, current use of cigarettes, and whether the patient had ever been told by a physician that he/she had any of the following conditions: arthritis, chronic obstructive lung disease (COPD)/asthma, heart disease, hypertension, diabetes, and depression. The second component comprised the SF36-V, a version of the short form-36 (SF-36) health-related quality of life (QOL) questionnaire adapted for use in the VA outpatient population (24,26,27). The SF-36 is a generic measure of HRQOL that has been found to be valid, reliable and responsive to clinical change in patients with arthritis (28-33). The ShortForm 36 modified for veterans (SF-36V) consists of 8 multiitem subscales namely, physical functioning, role physical (role limitations due to physical problems), bodily pain, general health, energy/vitality, social functioning, role emotional (role limitations due to emotional problems), and mental

Health care utilization in veterans with arthritis health, similar to SF-36. The only difference is that the scales for role limitation due to physical and emotional problems were changed from a dichotomous response in SF-36 to a 5-choice ordinal scale in SF-36V (24,34). PCS scores and MCS scores were generated from these 8 subscales. These 2-component summary scores were standardized to the U.S. population and norm-based with a scoring range of 0 to 100, a mean of 50, and a standard deviation of 10. Lower scores indicate poorer health functioning. The third component of the survey questionnaire consisted of items assessing respondents’ difficulty with 6 basic activities of daily living (ADL) (bathing, dressing, eating, getting in or out of chairs, walking, and toileting) (35).

757

Explanatory/Predictor Variables

Participant self-reported questionnaire data were supplemented with administrative data from the national VA database in Austin, Texas, which included patient age, marital status, and employment status. Health care utilization data were collected for the year before the survey, including the number of inpatient hospitalizations, the number of outpatient encounters in primary care, medicine subspecialty care, surgical care, and mental health.

The SF36-V PCS and MCS scores were the 2 primary explanatory variables of interest. Additional covariates assessed included: (1) Demographic variables (age categorized as ⬍50, 50-64, or ⱖ65 years, (the latter category representing Medicare-eligible veterans); gender; race categorized as white or non-white; marital status categorized as married or not married unmarried or divorced; employment status categorized as employed, unemployed, or retired; and education level categorized as ⬍8th grade, some high school, high school graduate, or beyond high school); (2) Health care utilization for the year before the survey (inpatient use defined as any hospitalization or none; and number of outpatient visits defined as an aggregate of all primary care, medicine subspecialty care, surgical care, and mental health visits); (3) Presence of any of 5 self-reported physician diagnosed conditions, ie, COPD/asthma, diabetes, depression, hypertension, and heart disease (excludes arthritis, since a diagnosis of arthritis was an inclusion criterion for our present analyses); (4) ADL limitation (no limitation, limitation of 1-3 ADLs, and limitation of 4-6 ADLs); (5) Current smoking status (yes/no); and (6) Medical center site used (multiple site versus individual medical facilities).

Outcome Measures

Data Analysis

Outcome measures included in the current study were mortality, and inpatient hospitalization and outpatient visits at any of the VISN 13 medical facilities. Inpatient hospitalization was assessed both as a dichotomous (any/ none) and a continuous variable (number of hospitalizations) and outpatient use was assessed as a continuous variable (total number of outpatient visits as a sum total of primary care, medicine subspecialty, surgical, and mental health visits). These data were obtained from the computerized VA administrative databases, including patient treatment file (PTF) and outpatient clinic (OPC) tables, which have been found to be reliable for demographics and most common diagnoses (36), and valid for specific diagnoses (37,38). Mortality was designated as a dichotomous variable (death: yes/no) and these data were obtained from 3 sources: the Beneficiaries Identification Record Locator System (BIRLS) death file, which obtains death notices when a patient’s survivor applies for benefits; through PTF when a veteran dies at a VA facility; and through the Social Security Administration. BIRLS is comparable with the National Death Index regarding mortality ascertainment (39); PTF is complementary to BIRLS providing additional mortality data (39); and Social Security Administration has been used as a source of vital statistics for many years (40). The total number of VA outpatient visits was determined for an aggregate of all primary care, medicine subspecialty care, surgical care, and mental health visits. In a subsequent analysis of the number of outpatient visits, we separately analyzed the medical-surgical visits (combination of primary care, medicine subspecialty, and surgical visits) and the mental health visits during the year after the survey.

We used t-tests and Pearson chi-square tests to compare the characteristics of responders and nonresponders to the survey, between responders and nonresponders to the “arthritis” question (among survey responders), and between those with and without a self-report of physician-diagnosed arthritis (among responders to the “arthritis” question). The primary objective of our study (VAQS) was to evaluate the ability of PCS and MCS scores to predict mortality and health utilization in veterans who self-reported a physiciandiagnosis of arthritis. We used PCS and MCS scores as measures of QOL rather than individual subscales to reduce the number of statistical comparisons (41). PCS and MCS have proved useful in most studies and better than the best SF-36 subscale (41). The association between the PCS and MCS scores with the occurrence of any hospitalization, any outpatient visit in a VA facility, and mortality was examined using separate series of logistic regression models and results were expressed as odds ratios (OR). The models incorporated the respective SF36-V summary score, age, and covariates that either altered the point estimate for the effect of the SF36-V summary scale by at least 10% (confounders) or were significant explanatory measures in a univariate analysis with a P value ⬍ 0.05. Two-way interactions between all independent variables (including those with age) were assessed and were included in subsequent models if the P value for the interaction was less than 0.10. Potential models, identified using various model selection procedures, were compared in an iterative process and a final model was selected based on principles of improvement of fit, parsimony, and clinical relevance. To evaluate the sensitivity of the estimates for the association of the SF36-V summary scores and the outcome derived from this final model, the results from this model

Administrative Data

758 were compared with those from models derived utilizing a stepwise algorithm for covariate inclusion, the model incorporating all 18 variables, and the model incorporating only the SF36-V score and age. Poisson regression techniques were utilized to examine the relationship between the PCS and MCS scores and the number of outpatient visits and the number of hospitalizations, with results expressed as the percent increase in the annual rate of utilization (rate ratio, RR). The distribution of the number of outpatient visits in the year after the survey was highly skewed in that a small portion of the patients frequently visited psychiatry day-care units, sometimes as many as 330 visits per year. In the implementation of these regression techniques the outcome measures are often truncated and over-dispersion parameters added to the regression model (42). Truncation is used to address issues with tails for the distributions longer than possessed by Poisson distributions and overdispersion parameters are incorporated in the models to address variances for the outcome distributions greater than possessed by Poisson distributions. In our analyses, we truncated the number of outpatient visits at 20 visits (92% of subjects had 20 or fewer visits) and incorporated an over-dispersion parameter in the model. Methods for deriving multivariable adjusted models, and for the selection and testing of the final model, were analogous to those described above for the dichotomous outcome measures. The Poisson regression based analyses for both medicalsurgical clinic and mental health clinic visits included all 18 variables. Since age-adjusted estimates were not substantially different from the unadjusted estimates, and those from the full model (included all 18 variables) were not substantially different from the multivariable model (included significant covariates and confounders in addition to PCS and MCS) derived estimates, we present the unadjusted and multivariable-adjusted estimates only. We used generalized linear models to obtain the mean PCS and MCS scores (least-squares means) for subjects with and without arthritis, which are adjusted for the following variables: Demographics (age, sex, race, education level, employment, and marital status), smoking status, limitation of ADLs, utilization variables (medical center site use, prior hospitalization, number of prior outpatient clinic visits), and 5 comorbidities (COPD/asthma, depression, diabetes, hypertension, heart disease). Analyses were done using SPSS 11.5 (Chicago, IL) and SAS 8.0 (Cary, NC). For all analyses, a P value of ⬍0.05 was considered statistically significant.

Results Demographic and Clinical Characteristics of the Entire Cohort Compared with subjects who reported no arthritis, those who reported arthritis were significantly older, less educated, less likely to be employed and more likely to be male,

J.A. Singh et al. married and retired (Table 1). 40,508 of 70,334 eligible veterans responded to the survey for an overall response rate of 58%. Nonresponders were significantly younger (56.3 versus 64.5 years), less likely to be married (47 versus 65%), less likely to be retired (27 versus 44%), and had lower number of future outpatient (8 versus 9 visits/year) but had the same future inpatient utilization rates (13% with 1 or more hospitalization versus 13%), as compared with responders. Among the 40,508 responders to the survey, 34,440 veterans (85%) answered the arthritis question and 18,464 veterans (54%) reported physician-diagnosed arthritis. Fifteen percent of survey respondents (6068/ 40,508) did not answer the “arthritis question.” These subjects were almost a decade older (72.5 versus 63 years, P ⬍ 0.001); less likely to have been educated beyond high school (20 versus 37%, P ⬍ 0.001), to be females (3 versus 5%, P ⬍ 0.001), and currently employed (19 versus 34%, P ⬍ 0.001); had higher future hospitalization rates (54 versus 46%, P ⬍ 0.001), but were as likely to be married (61 versus 61%) and had similar number of future outpatient visits (8 versus 8.3 visits/year), when compared with those who responded to the question. Compared with subjects who reported no arthritis, those who reported arthritis were significantly older, less educated, and more likely to be male, married, and either unemployed or retired (Table 1). Patients who self-reported a physician-diagnosis of arthritis were also sicker than subjects who self-reported no physician-diagnosis of arthritis, with a significantly greater comorbidity; greater likelihood of any inpatient hospitalization; and a significantly higher number of hospitalizations, primary care, medicine subspecialty, and surgery clinic visits (Table 2). Patients with self-reported physician-diagnosed arthritis had significantly fewer mental health clinic visits but were not different for postsurvey mortality as compared with patients without self-reported physician-diagnosed arthritis (Table 2).

Determinants of Inpatient Health Care Utilization in Veterans with Self-Reported Physician-Diagnosed Arthritis In subjects with self-reported physician-diagnosed arthritis, lower PCS and MCS scores were associated with significantly higher inpatient health care utilization. In unadjusted analyses, patients in the lowest and middle tertiles of PCS had 2.4 times and 1.5 times greater odds, and those in the lowest and middle tertiles of MCS had 1.7- and 1.3-fold greater odds of being admitted to the hospital in the year following the survey than those in the respective highest tertiles (Table 3). These differences persisted after adjusting for age without any substantial change in the odds ratios. In the final multivariable model, the odds of hospitalization for patients in only the lowest tertiles of PCS and MCS remained significantly higher than for those in the highest tertile (odds ratios, 1.5 and 1.2, respectively; Table 3). The results were similar in a full

Health care utilization in veterans with arthritis

759

Table 1 Demographic Characteristics of the Entire Surveyed Cohort of VISN 13 HRQOL Study that Completed the Arthritis Questionnaire

Age (Mean years ⴞ SD)† Sex (% male)† Race (% white) Marital status (% married)† Education† % <8th grade Some HS HS graduate Beyond HS Employment status† % Employed Not employed Retired Unknown

All subjects (N ⴝ 34,440)

Subjects without Self-Reported Physician-Diagnosed Arthritis (N ⴝ 15,976)

Subjects with Self-Reported Physician-Diagnosed Arthritis (N ⴝ 18,464)

64.4 ⴞ 13.7 96% 90% 61%

60.3 ⴞ 14.6 95 91 55

65.4 ⴞ 12.5 96 90 65

19% 11% 35% 35%

13 9 36 42

20 12 35 33

32% 18% 44% 7%

39 19 34 8

30 18 46 6

HS, high school; All percentages rounded to the nearest digit, so the total may not exactly be 100%. †Differences between subjects with and without self-report of physician-diagnosed arthritis statistically significant with a P value of <0.001.

multivariable model and in the model derived using stepwise regression (data not shown). Patients in the lowest and middle tertiles of PCS had 134 and 50% higher number, and those in the lowest and middle tertiles of MCS had 64 and 21% higher number, of hospitalizations/year, respectively, as compared with patients in the respective highest tertiles (Table 3). After adjusting for other significant covariates and confounders in the multivariable model, those in the lowest and middle tertiles of PCS had 9 and 6% higher, respectively, and those in the lowest tertiles of MCS a 5% higher number of hospitalizations (all significant with P ⬍ 0.01).

Determinants of Outpatient Health Care Utilization in Veterans with Self-Reported Physician-Diagnosed Arthritis In unadjusted analyses, compared with arthritis patients in the respective highest tertile, a significantly higher number of outpatient clinic visits/year were observed in subjects in the lowest and middle tertiles for PCS (40 and 24%, respectively) and MCS scores (37 and 14%, respectively) (Table 3). Results were largely unchanged by adjustment for age and were attenuated in multivariable-adjusted models. In the multivariable-adjusted models, significantly more frequent outpatient visits occurred in veterans in the lowest and middle tertiles of PCS (7 and 5% higher, respectively) and the lowest MCS tertile (7% higher) (Table 3). In a further analysis of the relationship of PCS and MCS with different types of outpatient visits, we found that patients in the lowest and middle tertiles of PCS scores had a 12 and 6% higher number of medical-surgical visits per year compared with the highest tertile, while the patients in the lowest and middle tertiles of MCS scores had a 74 and 25% higher number of mental health visits per year, respectively,

as compared with the highest tertile (all statistically significant, P ⬍ 0.001 for each; data not shown).

Determinants of Death in Veterans with Self-Reported Physician-Diagnosed Arthritis Veterans in the lower PCS and MCS tertiles were at significantly higher risk of death. Patients in the lowest tertiles of PCS and MCS had 3.8 and 2.5 times higher odds, and those in the middle tertiles of PCS and MCS had 1.3 and 1.6 times higher odds of death, respectively, as compared with those in highest tertiles (Table 3, all results statistically significant). The differences persisted after age-adjustment. After adjusting for various confounders and significant covariates, patients in the lowest tertiles of PCS and MCS scores, but not in middle tertiles, were still at significantly higher odds of death, the odds ratios being 1.7 and 2.1, respectively (Table 3).

Comparison of Health-Related Quality of Life between Subjects with and without Self-Reported Physician-Diagnosed Arthritis The following adjusted SF-36V subscale scores (mean ⫾ standard error of the mean) were significantly lower, both clinically and statistically, in subjects with arthritis as compared with those without arthritis, respectively: Bodily pain (36.7 ⫾ 0.5 versus 49.5 ⫾ 0.5, P ⬍ 0.001), role physical (37.4 ⫾ 0.2 versus 41.2 ⫾ 0.5, P ⬍ 0.001), physical functioning (37.9 ⫾ 0.5 versus 41.6 ⫾ 0.5, P ⬍ 0.001), and PCS (29.8 ⫾ 0.2 versus 33.1 ⫾ 0.2, P ⬍ 0.001). MCS (40.5 ⫾ 0.1 versus 41.1 ⫾ 0.2, P ⬍ 0.001), general health (40 ⫾ 0.4 versus 37.7 ⫾ 0.4, P ⬍ 0.001), and vitality scores (33.4 ⫾ 0.4 versus 35.7 ⫾ 0.5, P ⬍ 0.001) were statistically lower (due to the large sample size), but not clinically different in subjects with arthritis when compared with those without arthritis.

J.A. Singh et al.

760

Table 2 Comparison of Comorbid Medical Conditions, and Health Care Utilization and Mortality for 1 Year Postsurvey, in Subjects with and without Self-report of Physician-Diagnosed Arthritis

Comorbid diseases COPD/asthma Depression Diabetes Hypertension Heart problems Current tobacco use Number of additional comorbid conditions† None One Two Three Four or more % With >1 hospital admissions␨ % Subjects died␨ Number of hospital admissions␨ Number of outpatient visits␨ Primary care visits Surgery visits Medicine subspecialty visits Mental health visits

Subjects without Self-Report of Physician-Diagnosed Arthritis % or mean ⴞ SD

Subjects with Self-Report of Physician-Diagnosed Arthritis % or mean ⴞ SD

20% 22% 15% 28% 26% 26%

30 31 19 40 39 22

1.8 (1.7–1.8)† 1.6 (1.5–1.7)† 1.4 (1.3–1.5)† 1.7 (1.6–1.8)† 1.9 (1.8–2.0)† 0.8 (0.8–0.9)†

36% 34% 19% 8% 3% 10%

21 32 25 14 8 14

1.4 (1.3–1.5)†

3% 0.16 ⴞ 0.6

3 0.23 ⴞ 0.7†

7.6 ⴞ 22.5 2.0 ⴞ 2.8 1.4 ⴞ 2.7 1.3 ⴞ 4.2 2.9 ⴞ 21.3

9 ⴞ 19.6† 2.9 ⴞ 3.2† 2.1 ⴞ 3.4† 1.5 ⴞ 4.4† 2.4 ⴞ 17.7*

Odds Ratio (OR) with (95% CI)¶

1.1 (1.0–1.2) NS NA NA NA NA NA NA

All percentages rounded to the nearest digit, so the total may not exactly be 100%. SD, standard deviation; OR, Odds ratio; CI, confidence interval; COPD, chronic obstructive pulmonary disease; NA, not applicable. †Chi-square P value < 0.001; *P value < 0.05; NS, not statistically significant (P > 0.05). ¶Odds ratios were calculated with subjects without disease conditions in the reference category. ␨Health care utilization and mortality outcomes were assessed for 1-year postsurvey.

There were no statistically significant differences in social functioning, role emotional, and mental health subscale scores between the 2 groups.

Association of Other Covariates with Outcomes (Appendix 1) We found that prior hospitalization significantly increased the odds of any future hospitalization, number of hospitalizations and outpatient visits, and mortality (Appendix 1). More frequent prior outpatient visits significantly increased the odds of any future hospitalization, number of hospitalizations and outpatient visits, but not mortality. Additional factors that predicted higher health care utilization included older age, employment status, and presence of comorbidities. Older age, employment status, current smoking and presence of COPD, heart disease, and diabetes increased the risk of death (Appendix 1).

Discussion The data from this study demonstrated that both worse physical and mental health related quality of life are important

independent determinants of higher future health care utilization and mortality in veterans with self-reported physiciandiagnosed arthritis. This association remained significant even after adjusting for demographic characteristics, comorbidities, functional status, and prior health care utilization. PCS scores predicted future medical-surgical outpatient visits, while MCS scores predicted future mental health outpatient visits. Both prior inpatient and outpatient use predicted future inpatient and outpatient utilization, and prior inpatient utilization predicted mortality. Veterans with self-reported physician-diagnosed arthritis had worse HRQOL, poorer functional status as assessed by ADLs, and higher comorbidity than veterans without arthritis. Our study is the first to explore the effect of HRQOL, as assessed by SF-36, on health care utilization in such a large cohort of veterans with self-reported physician-diagnosed arthritis. Most previous studies of health care utilization in patients with arthritis were clinic- and not population-based, and did not include prior health care utilization as a predictor (13-16). In the only population-based study, involving 361 health maintenance organization (HMO) members older than 60 years with osteoarthritis, prior health care use, age,

Health care utilization in veterans with arthritis

761

Table 3 Odds and Rate Ratios (with 95% Confidence Intervals) of the Relationship of PCS and MCS Tertiles with Health Outcomes in Veterans with Self-Report of Physician-Diagnosed Arthritis PCS Tertiles Lowest

MCS Tertiles Middle

Lowest

Middle

1.74 (1.55–1.95)† 1.20 (1.02–1.41)*

1.31 (1.16–1.47)† 1.09 (0.94–1.25)

1.64 (1.52–1.78)‡ 1.05 (1.02–1.09)§

1.21 (1.11–1.31)‡ 1.01 (0.98–1.04)

1.37 (1.32–1.42)‡ 1.07 (1.03–1.11)†

1.14 (1.10–1.18)‡ 1.02 (0.99–1.05)

2.52 (2.00–3.18)† 2.14 (1.56–2.94)†

1.58 (1.22–2.03)† 1.27 (0.93–1.72)

Any hospitalization Unadjusted OR Multivariable adjusted OR1

2.40 (2.14–2.70)† 1.49 (1.25–1.76)†

1.49 (1.32–1.67)† 1.11 (0.94–1.30)

No. of hospitalizations Unadjusted RR Multivariable adjusted RR2

2.34 (2.15–2.54)‡ 1.09 (1.05–1.13)†

1.50 (1.37–1.64)‡ 1.06 (1.02–1.10)§

No. of outpatient visits Unadjusted RR Multivariable adjusted RR3

1.40 (1.35–1.45)‡ 1.07 (1.03–1.11)†

1.24 (1.19–1.28)‡ 1.05 (1.01–1.08)§ Death

Unadjusted OR Multivariable adjusted OR4

3.78 (2.95–4.84)† 1.69 (1.18–2.42)§

1.30 (0.98–1.73) 0.83 (0.58–1.19)

OR, odds ratio; RR, rate ratio; Age-adjusted estimates of OR and RR were not much different from the unadjusted estimates. Full model derived estimates of OR and RR were similar to the multivariable adjusted estimates. Multivariable models contained PCS and MCS, covariates independently associated with the outcome, and confounders of association of PCS and MCS scores with the outcome. Significant independent covariates for each outcome are listed in Appendix 1. ‡p < 0.0001, †p < 0.001, §p < 0.01, *p < 0.05. 1Multivariable model for any hospitalization included covariates independently associated with the outcome (Appendix 1) and the following confounders: Both PCS and MCS: employment status, prior inpatient hospitalization, prior outpatient use, and heart disease; PCS alone: limitation of ADLs; MCS alone: depression. This model explained 11.5% of variation in hospitalization and the c-statistic of the model was 0.78. 2Multivariable model for number of hospitalizations included covariates independently associated with the outcome (Appendix 1) and the following confounders: Both PCS and MCS: limitation of ADLs, prior inpatient hospitalization, prior outpatient use, and heart disease; PCS alone: diabetes; MCS alone: current smoking status, employment status, asthma, depression, hypertension. 3Multivariable model for number of outpatient visits included covariates independently associated with the outcome (Appendix 1) and the following confounders: Both PCS and MCS: employment status, prior inpatient hospitalization and prior outpatient use; PCS alone: limitation of ADLs and diabetes; MCS alone: depression. 4Multivariable model for mortality included covariates independently associated with the outcome (Appendix 1) and the following confounders: Both PCS and MCS: limitation of ADLs and COPD/asthma; PCS alone: age category, employment status, educational status, prior inpatient hospitalization, presence of heart disease, diabetes, and current smoking status; MCS alone: age category (negative). This multivariable model explained 6.3% of the variation in mortality, and the c-statistic of the model was 0.81.

HRQOL, physical impairment, and pain were significant predictors of future health care use (12). This study did not control for comorbidities, which we found to be not only predictors of future utilization and mortality but also confounders of the association of PCS and MCS with these outcomes in our study. Additional limitations of this previous study include a small sample size and enrollment of only older subjects, which limit its generalizability. Further, our study included all patients with any form of arthritis irrespective of whether they were seen in a subspecialty clinic. This is in contrast to previous studies that selected patients attending a specific/subspecialty clinic or suffering from only specific types of arthritis (rheumatoid arthritis or osteoarthritis or other arthritides). There are numerous reasons for studying health care utilization in veterans with arthritis: (1) Arthritis, although more common in women, affects 28.9 million men (28%) in the U.S. according to a recent survey (1); (2) The VA veteran

population may be representative of older, disadvantaged subgroups of the general population, who are in dire need of our attention due to a substantial difficulty in access to medical care; (3) Arthritis is among the top 5 most common chronic diseases in veterans (22) and contributes significantly to health care costs at the VA (22); and (4) With the aging of the “baby boomers” generation, in the near future more veterans ⱖ 65 years of age will have access to Medicare and our study findings may help improve understanding of the determinants of Medicare resource utilization by veterans with arthritis. We found that both PCS and MCS predicted future health care utilization and mortality and that PCS was a slightly stronger predictor of inpatient utilization and MCS a stronger predictor of mortality in multivariable adjusted analyses. A practical use of our model is that PCS and MCS scores can be used to assess risk of resource utilization and mortality in patients, since both PCS and MCS scores were independently

J.A. Singh et al.

762 associated with the outcomes and added predictive value to prior utilization, comorbidities, and socio-demographic variables. We observed an impressive 20 to 50% increase in odds of any hospitalization and 69 to 114% increased odds of 1-year mortality among veterans with arthritis in the lowest tertiles of MCS or PCS. There was a modest, but significant, increase of 7% in the number of outpatient visits and 5 to 9% increase in the number of hospitalizations in veterans with arthritis in the lowest MCS or PCS tertiles. These observations indicate significant poor health outcomes and higher resource utilization in this group of veterans with arthritis, a finding that has implications for clinical patient care. We hypothesize that the poor health outcomes in arthritis patients with lower PCS and MCS scores in our study may be due to the presence of unmeasured additional comorbidity in these subjects. Our observation of a higher prevalence of depression with a lower number of mental health visits and a higher prevalence of all medical comorbidities (hypertension, COPD, heart disease, diabetes) among those who selfreported a physician-diagnosis of arthritis, compared with those without arthritis, suggests an unmet need for medical and psychiatric evaluation in these patients. Programs that focus on early screening and treatment of psychosocial and physical morbidity to improve the physical and mental health in arthritic veterans with lower PCS and MCS scores may lead to a decrease in health care utilization and mortality. Since it is easier to measure HRQOL on a populationbasis rather than collecting complete comorbidity data with disease severity and treatment, we believe that PCS and MCS scores may be able to predict health outcomes in a relatively inexpensive manner even in the absence of comorbidity data. Based on our findings, randomized studies of the effect of these interventions on health outcomes in arthritis patients stratified by PCS and MCS scores are desirable. The fact that PCS and in some cases MCS was a stronger predictor of mortality and utilization than individual comorbidities in our study emphasizes the impact of poor PCS and MCS scores on outcomes when compared with having these additional comorbidities. A few observations in our much larger sample are similar to and confirm the findings from previous studies. These include (1) worse HRQOL (43-45) and higher health care utilization (46,47) in arthritic patients as compared with those without arthritis; (2) greater magnitude of association of PCS and MCS scores with each of the outcomes than for any of the medical comorbidities alone in arthritis patients (48); (3) strong association of HRQOL measures with mortality in arthritis patients (49-55); and (4) finding that prior health care utilization was a strong predictor of future utilization in patients with arthritis (12). The strengths of our study include the following: the study had a large sample size; the study participants included a large percentage (58%) of all veterans receiving health care in the

Upper Midwest Veterans Network, which is probably representative of veterans seeking health care at other VA networks across the country and thus generalizable to them; we included prior health care utilization rates as predictors in addition to clinical, functional status, and demographic variables; and the findings were robust across various regression models assessed. There are several limitations to this study. Although the study is probably representative of veterans receiving health care at VA facilities in the U.S., the results may not be generalizable to veterans receiving health care outside the VA network (56) or to other populations. We did not measure non-VA health care use by our subjects and a proportion of veterans receive health care outside of the VA (57). The use of self-report for a diagnosis is debated (58) with studies showing a wide range of agreement with medical record data for a diagnosis of arthritis/musculoskeletal disease ranging from moderate agreement (kappa ⫽ 0.48) (59) to poor agreement (kappa ⫽ 0.27 and 0.38, respectively) (60,61). Since 16 to 22% of patients with arthritis or chronic joint symptoms may not consult a physician for their condition (62,63), self-reported data may be more sensitive in estimating the prevalence of arthritis in the population than medical encounter-based data. We did not collect data about disease severity, the presence of a number of other chronic illnesses including stroke and cancer, and other variables, which may be associated with the outcomes. Inclusion of additional comorbidities may attenuate the association of HRQOL with the outcomes, and thus we may have over-estimated the strength of these associations. The striking association of self-reported physician-diagnosed arthritis with heart disease, hypertension, and diabetes may be due to a higher body mass index in these subjects, which was not measured in our study and thus may have confounded some of the results. Although this study focused on veterans with arthritis, the presence of other comorbidities (that were not included in our survey) in veterans with and without arthritis may also affect health care utilization and outcomes. A study of association of baseline HRQOL with health outcomes and mortality at 5- or 10-year or longer time periods in arthritis patients may be of considerable interest, but limited resources prevented us from pursuing this. In summary, HRQOL summary measures are important independent determinants of health care utilization and mortality in veterans with self-reported physician-diagnosed arthritis. These summary measures could be used to identify veterans most at risk for poor health outcomes and preventive interventions targeted at these groups of arthritis patients could improve health outcomes and decrease health care expenditures.

Acknowledgments We thank Dr. Maren Mahowald, University of Minnesota, for critical review of the manuscript. We thank Dr. Petzel for providing the vision and the funding support for the VISN 13 Veterans Quality of Life (VISN 13 Vet-QOL) Study.

Health care utilization in veterans with arthritis

763

Appendix 1 Factors Other than PCS and MCS Summary Scores that Were Significantly Associated with Health Outcomes in Veterans with Arthritis (VAQS)

Predictors Prior hospitalization Prior outpatient visits Lowest tertile Middle tertile Highest tertile Limitation of ADLs No limitation 1–3 ADL limitation 4–6 ADL limitation Currently smoke COPD/asthma Heart disease Hypertension Diabetes Depression Age category <50 years 50–64 year 65 and older Gender Female Male Employment status Employed Not employed Retired Marital status Married Unmarried/divorced Site Multi-site Minneapolis St. Cloud Sioux Falls Blackhills Fargo Education level <8th grade Some HS HS graduate Beyond HS

Number of Outpatient Visits RR (95% CI)

Any Hospitalization OR (95% CI)

Number of Hospitalizations RR (95% CI)

2.75 (2.41–3.15) †

2.23 (2.07–2.41) †

1.10 (1.07–1.14) †

2.09 (1.62–2.71) †

Reference 2.05 (1.72–2.45) † 3.61 (3.03–4.30)

1.80 (1.66–1.96) † 3.42 (3.04–3.84)

1.73 (1.71–1.76) † 4.23 (4.06–4.40)

0.81 (0.60–1.09) NS 1.02 (0.76–1.38)

Reference 1.22 (1.03–1.44) 1.21 (0.99–1.48) 1.34 (1.16–1.53) 1.19 (1.05–1.34) 1.22 (1.08–1.37) 0.93 (0.82–1.04) 1.13 (0.98–1.29) 1.06 (0.92–1.22)

1.19 (1.07–1.32) 1.15 (1.01–1.30) 0.87 (0.80–0.95) 1.24 (1.15–1.33) 1.28 (1.19–1.38) 0.92 (0.85–0.99) 1.14 (1.05–1.23) 1.11 (1.02–1.21)

1.06 (1.02–1.10) 1.00 (0.96–1.04) 0.99 (0.96–1.02) 1.03 (1.01–1.06) 1.03 (1.01–1.06) 1.02 (0.99–1.05) 1.10 (1.06–1.13) 1.08 (1.05–1.12)

1.22 (0.84–1.78) 1.88 (1.24–2.86) 1.32 (1.01–1.75) 1.25 (1.01–1.56) 1.62 (1.29–2.04) 1.36 (0.80–2.32) 1.31 (1.03–1.67) 1.02 (0.79–1.31)

NS † § § NS NS NS

§ § † † * § *

Death OR (95% CI)

† NS * * NS † †

§ * * † NS * NS

Reference 1.36 (1.11–1.66) § 1.41 (1.13–1.76)

1.18 (1.04–1.34) § 1.26 (1.09–1.44)

1.06 (1.02–1.10) * 1.03 (0.98–1.08)

Reference 1.01 (0.81–1.41) NS

0.75 (0.61–0.91) §

1.03 (0.97–1.10) NS

1.79 (0.77–4.13) NS

Reference 1.40 (1.18–1.67) † 1.57 (1.34–1.85)

1.35 (1.21–1.50) † 1.51 (1.36–1.67)

1.10 (1.06–1.14) § 1.12 (1.08–1.16)

1.86 (1.20–2.87) † 2.28 (1.58–3.29)

Reference 1.29 (1.15–1.46) †

0.79 (0.74–0.85) †

1.01 (0.99–1.04) NS

1.18 (0.93–1.48) NS

Reference 0.92 (0.78–1.09) 0.61 (0.44–0.86) 0.96 (0.77–1.20) 1.02 (0.83–1.25) 0.98 (0.79–1.23)

NS

0.85 (0.77–0.94) 0.72 (0.59–0.88) 0.88 (0.78–1.00) 0.90 (0.81–1.02) 0.93 (0.82–1.06)

§

1.01 (0.97–1.05) 0.84 (0.78–0.90) 0.90 (0.86–0.94) 0.90 (0.86–0.94) 0.83 (0.80–0.88)

†

1.03 (0.73–1.46) 0.95 (0.52–1.73) 1.17 (0.77–1.77) 1.05 (0.69–1.59) 0.99 (0.64–1.53)

NS

NS

0.96 (0.85–1.08) 0.96 (0.87–1.05) 0.98 (0.89–1.08)

NS

1.00 (0.96–1.05) 1.01 (0.98–1.05) 1.06 (1.02–1.10)

§

0.95 (0.66–1.35) 0.99 (0.75–1.30) 0.98 (0.72–1.33)

NS

Reference 0.95 (0.78–1.17) 0.92 (0.78–1.08) 0.89 (0.75–1.06)

2.40 (1.20–4.83) † 7.70 (3.88–15.25)

*P < 0.05; §P < 0.01; †P < 0.001; NS, nonsignificant. Race was an insignificant factor for all the outcomes. OR, odds ratio; RR, rate ratio; CI, confidence interval; HS, high school.

References 1. CDC. Prevalence of self-reported arthritis or chronic joint symptoms among adults—United States, 2001. MMWR Morb Mortal Wkly Rep 2002;51(42):948-50. 2. Deyo RA, Inui TS, Leininger J, Overman S. Physical and psychosocial function in rheumatoid arthritis. Clinical use of a self-administered health status instrument. Arch Intern Med 1982;142(5): 879-82. 3. Achterberg-Lawlis J. The psychological dimensions of arthritis. J Consult Clin Psychol 1982;50(6):984-92.

4. CDC. Prevalence of disabilities and associated health conditions among adults in United States, 1999. MMWR 2001;50(07):120-5. 5. CDC. Impact of arthritis and other rheumatic conditions on the healthcare system—United States, 1997. MMWR Morb Mortal Wkly Rep 1999;48:349-53. 6. Yelin E, Callahan LF. The economic cost and social and psychological impact of musculoskeletal conditions. National Arthritis Data Work Groups. Arthritis Rheum 1995;38(10):1351-62. 7. Eisner MD, Ackerson LM, Chi F, Kalkbrenner A, Buchner D, Mendoza G, et al. Health-related quality of life and future health care utilization for asthma. Ann Allergy Asthma Immunol 2002;89(1):46-55.

764 8. Vollmer WM, Markson LE, O’Connor E, Sanocki LL, Fitterman L, Berger M, et al. Association of asthma control with health care utilization and quality of life. Am J Respir Crit Care Med 1999;160(5 Pt. 1):1647-52. 9. Drossman DA, Leserman J, Mitchell CM, Li ZM, Zagami EA, Patrick DL. Health status and health care use in persons with inflammatory bowel disease. A national sample. Dig Dis Sci 1991;36(12):1746-55. 10. Traver GA. Measures of symptoms and life quality to predict emergent use of institutional health care resources in chronic obstructive airways disease. Heart Lung 1988;17(6 Pt. 1):689-97. 11. Konstam V, Salem D, Pouleur H, Kostis J, Gorkin L, Shumaker S, et al. Baseline quality of life as a predictor of mortality and hospitalization in 5,025 patients with congestive heart failure. SOLVD Investigations. Studies of Left Ventricular Dysfunction Investigators. Am J Cardiol 1996;78(8):890-5. 12. Cronan TA, Shaw WS, Gallagher RA, Weisman M. Predicting health care use among older osteoarthritis patients in an HMO. Arthritis Care Res 1995;8(2):66-72. 13. Berkanovic E, Hurwicz ML. Physician visits by rheumatoid arthritis patients: a prospective analysis. Arthritis Care Res 1995;8(2):73-9. 14. Ethgen O, Kahler KH, Kong SX, Reginster JY, Wolfe F. The effect of health related quality of life on reported use of health care resources in patients with osteoarthritis and rheumatoid arthritis: a longitudinal analysis. J Rheumatol 2002;29(6):1147-55. 15. Hawley DJ, Wolfe F. Anxiety and depression in patients with rheumatoid arthritis: a prospective study of 400 patients. J Rheumatol 1988; 15(6):932-41. 16. Katz PP, Yelin EH. Prevalence and correlates of depressive symptoms among persons with rheumatoid arthritis. J Rheumatol 1993;20(5): 790-6. 17. Randall M, Kilpatrick KE, Pendergast JF, Jones KR, Vogel WB. Differences in patient characteristics between Veterans Administration and community hospitals. Implications for VA planning. Med Care 1987; 25(11):1099-104. 18. Wolinsky FD, Coe RM, Mosely 2nd, RR Homan SM. Veterans’ and nonveterans’ use of health services. A comparative analysis. Med Care 1985;23(12):1358-71. 19. Wilson NJ, Kizer KW. The VA health care system: an unrecognized national safety net. Health Aff (Millwood) 1997;16(4):200-4. 20. Petersen LA, Normand SL, Daley J, McNeil BJ. Outcome of myocardial infarction in Veterans Health Administration patients as compared with medicare patients. N Engl J Med 2000;343(26):1934-41. 21. Iglehart JK. Reform of the Veterans Affairs heath care system. N Engl J Med 1996;335(18):1407-11. 22. Yu W, Ravelo A, Wagner T, Phibbs C, Bhandari A, Chen S, et al. Prevalence and costs of chronic conditions in the VA health care system. Med Care 2003;60(3):146S-67S. 23. Agha Z, Lofgren RP, VanRuiswyk JV, Layde PM. Are patients at Veterans Affairs medical centers sicker? A comparative analysis of health status and medical resource use. Arch Intern Med 2000;160(21):3252-7. 24. Kazis LE, Miller DR, Clark J, Skinner K, Lee A, Rogers W, et al. Healthrelated quality of life in patients served by the Department of Veterans Affairs: results from the Veterans Health Study. Arch Intern Med 1998; 158(6):626-32. 25. Singh JA, Borowsky SJ, Nugent S, Murdoch M, Zhao Y, Nelson DB, et al. Health-related quality of life, functional impairment and health care utilization in veterans: Veterans’ Quality of Life Study (Vet-QOL Study). J Am Geriatrics Soc 2005;53(1):108-13. 26. Ware Jr, JE Sherbourne CD. The MOS 36-item short-form health survey (SF-36). I. Conceptual framework and item selection Med Care 1992;30(6):473-83. 27. Kazis LE, Ren XS, Lee A, Skinner K, Rogers W, Clark J, et al. Health status in VA patients: results from the Veterans Health Study. Am J Med Qual 1999;14(1):28-38. 28. Tuttleman M, Pillemer SR, Tilley BC, Fowler SE, Buckley LM, Alarcon GS, et al. A cross sectional assessment of health status instruments in patients with rheumatoid arthritis participating in a clinical trial. Minocycline in Rheumatoid Arthritis Trial Group. J Rheumatol 1997; 24(10):1910-5.

J.A. Singh et al. 29. Kvien TK, Kaasa S, Smedstad LM. Performance of the Norwegian SF-36 Health Survey in patients with rheumatoid arthritis. II. A comparison of the SF-36 with disease-specific measures J Clin Epidemiol 1998;51(11):1077-86. 30. Husted JA, Gladman DD, Farewell VT, Long JA, Cook RJ. Validating the SF-36 health survey questionnaire in patients with psoriatic arthritis. J Rheumatol 1997;24(3):511-7. 31. Kosinski M, Keller SD, Ware Jr., JE, Hatoum HT, Kong SX. The SF-36 Health Survey as a generic outcome measure in clinical trials of patients with osteoarthritis and rheumatoid arthritis: relative validity of scales in relation to clinical measures of arthritis severity. Med Care 1999;37(5 Suppl.):MS23-39. 32. Gill TM, Feinstein AR. A critical appraisal of the quality of quality-oflife measurements. JAMA 1994;272(8):619-26. 33. McHorney CA. Health status assessment methods for adults: past accomplishments and future challenges. Annu Rev Public Health 1999; 20:):309-35. 34. Kazis L, Skinner K, Rogers W, Lee A, Ren X, Miller D, et al. Health status of veterans: physical and mental component summary scores (SF-36V). 1998 National Survey of Ambulatory Care Patients. Executive Report: Department of Veterans Affairs, Veterans Health Administration, Office of Performance and Quality, Washington D.C. 35. Katz SFA, Moskowitz RW, Jackson BA, Jaffe MW. Studies of illness in aged: The index of ADL: a standard measure of biological and psychological function. JAMA 1963;185(Sept. 21):9):914-9. 36. Kashner TM. Agreement between administrative files and written medical records: a case of the Department of Veterans Affairs. Med Care 1998;36(9):1324-36. 37. Szeto HC, Coleman RK, Gholami P, Hoffman BB, Goldstein MK. Accuracy of computerized outpatient diagnoses in a Veterans Affairs general medicine clinic. Am J Manag Care 2002;8(1):37-43. 38. Petersen LA, Wright S, Normand SL, Daley J. Positive predictive value of the diagnosis of acute myocardial infarction in an administrative database. J Gen Intern Med 1999;14(9):555-8. 39. Fisher SG, Weber L, Goldberg J, Davis F. Mortality ascertainment in the veteran population: alternatives to the National Death Index. Am J Epidemiol 1995;141(3):242-50. 40. Curb JD, Ford CE, Pressel S, Palmer M, Babcock C, Hawkins CM. Ascertainment of vital status through the National Death Index and the Social Security Administration. Am J Epidemiol 1985;121(5):754-66. 41. Ware Jr., JE, Kosinski M, Bayliss MS, McHorney CA, Rogers WH, Raczek A. Comparison of methods for the scoring and statistical analysis of SF-36 health profile and summary measures: summary of results from the Medical Outcomes Study. Med Care 1995;33(4 Suppl.): AS264-79. 42. McCullagh P, Nelder JA. Generalized Linear Models. Second ed. London: Chapman and Hall/CRC, 1989. 43. Hill CL, Parsons J, Taylor A, Leach G. Health related quality of life in a population sample with arthritis. J Rheumatol 1999;26(9):2029-35. 44. Wiles NJ, Scott DG, Barrett EM, Merry P, Arie E, Gaffney K, et al. Benchmarking: the five year outcome of rheumatoid arthritis assessed using a pain score, the Health Assessment Questionnaire, and the Short Form-36 (SF-36) in a community and a clinic based sample. Ann Rheum Dis 2001;60(10):956-61. 45. CDC. Health related quality of life among adults with arthritis—Behavioral risk factor surveillance system, 11 states, 1986-1998. MMWR 2000;49(17):366-9. 46. Dunlop DD, Manheim LM, Song J, Chang RW. Health care utilization among older adults with arthritis. Arthritis Rheum 2003;49(2):164-71. 47. Yelin E, Herrndorf A, Trupin L, Sonneborn D. A national study of medical care expenditures for musculoskeletal conditions: the impact of health insurance and managed care. Arthritis Rheum 2001;44(5): 1160-9. 48. Winograd CH, Gerety MB, Chung M, Goldstein MK, Dominguez Jr., F Vallone R. Screening for frailty: criteria and predictors of outcomes. J Am Geriatr Soc 1991;39(8):778-84. 49. Leigh JP, Fries JF. Mortality predictors among 263 patients with rheumatoid arthritis. J Rheumatol 1991;18(9):1307-12. 50. Soderlin MK, Nieminen P, Hakala M. Functional status predicts mor-

Health care utilization in veterans with arthritis

51.

52.

53.

54.

55.

56.

tality in a community based rheumatoid arthritis population. J Rheumatol 1998;25(10):1895-9. Pincus T, Brooks RH, Callahan LF. Prediction of long-term mortality in patients with rheumatoid arthritis according to simple questionnaire and joint count measures. Ann Intern Med 1994;120(1):26-34. Wolfe F, Michaud K, Gefeller O, Choi HK. Predicting mortality in patients with rheumatoid arthritis. Arthritis Rheum 2003;48(6):153042. Wolfe F, Mitchell DM, Sibley JT, Fries JF, Bloch DA, Williams CA, et al. The mortality of rheumatoid arthritis. Arthritis Rheum 1994;37(4): 481-94. Callahan LF, Cordray DS, Wells G, Pincus T. Formal education and five-year mortality in rheumatoid arthritis: mediation by helplessness scale score. Arthritis Care Res 1996;9(6):463-72. Callahan LF, Pincus T, Huston 3rd, RH, Brooks EP Nance Jr., JW Kaye JJ. Measures of activity and damage in rheumatoid arthritis: depiction of changes and prediction of mortality over five years. Arthritis Care Res 1997;10(6):381-94. Oddone EZ, Petersen LA, Weinberger M, Freedman J, Kressin NR. Contribution of the Veterans Health Administration in understanding racial disparities in access and utilization of health care: a spirit of inquiry. Med Care 2002;40(1 Suppl.):I3-13.

765 57. Fleming C, Fisher ES, Chang CH, Bubolz TA, Malenka DJ. Studying outcomes and hospital utilization in the elderly. The advantages of a merged data base for Medicare and Veterans Affairs hospitals. Med Care 1992;30(5):377-91. 58. Harlow SD, Linet MS. Agreement between questionnaire data and medical records. The evidence for accuracy of recall. Am J Epidemiol 1989; 129(2):233-48. 59. Haapanen N, Miilunpalo S, Pasanen M, Oja P, Vuori I. Agreement between questionnaire data and medical records of chronic diseases in middle-aged and elderly Finnish men and women. Am J Epidemiol 1997;145(8):762-9. 60. Linet MS, Harlow SD, McLaughlin JK, McCaffrey LD. A comparison of interview data and medical records for previous medical conditions and surgery. J Clin Epidemiol 1989;42(12):1207-13. 61. Heliovaara M, Aromaa A, Klaukka T, Knekt P, Joukamaa M, Impivaara O. Reliability and validity of interview data on chronic diseases. The Mini-Finland Health Survey. J Clin Epidemiol 1993;46(2):181-91. 62. CDC. Adults who have never seen a health-care provider for Chronic joint symptoms—United States, 2001. MMWR Morb Mortal Wkly Rep 2003;52:416-9. 63. Rao JK, Callahan LF, Helmick 3rd. CG Characteristics of persons with self-reported arthritis and other rheumatic conditions who do not see a doctor. J Rheumatol 1997;24(1):169-73.