Article Text
Abstract
Objective: To better understand the impact of gout on functional status, health-related quality of life (HRQoL), mortality and healthcare utilisation in US veterans.
Methods: All veterans seen in Veterans Integrated Service Network-13 from 1 October 1996 to 31 March 1998 received mailed surveys asking about demographic characteristics; performance of activities of daily living and HRQoL by Short Form-36 (SF-36) for Veterans. Administrative data included demographics; inpatient/outpatient healthcare utilisation; ICD-9 codes for gout, medical comorbidities and arthritis excluding gout—“arthritic comorbidity” and 1-year mortality. Multivariable estimates compared results between veterans with/without gout using least means squared.
Results: Subjects with gout were significantly older, retired, not married, current non-smokers, with more comorbidities. Multivariable-adjusted bodily pain was somewhat worse (49.7 vs 47.1, p<0.01) and mental health (66.7 vs 68.6, p<0.01) domain scores somewhat better in patients with gout, both differences significant but not clinically meaningful (less than threshold of 5–10 points); other SF-36 domain and summary scores and functional limitations were similar. Medical or arthritic comorbidities predicted clinically/statistically lower adjusted scores in all SF-36 domains and physical domains (physical component summary), respectively. Patients with gout had significantly more annual primary care visits (3.5 vs 2.7, p<0.001) and admissions to hospital (18.3% vs 15.1%, p<0.01), fewer mental health visits (10.1% vs 13.7%, p<0.01) and similar mortality (2.6% vs 2.2%, p = 0.23).
Conclusions: Gout is independently associated with higher medical and arthritic comorbidity, primary care and inpatient utilisation. Poorer HRQoL, functional limitation and higher mortality noted in univariate analyses in patients with gout were attributable to higher comorbidity and sociodemographic characteristics.
Statistics from Altmetric.com
Arthritis is the fifth most prevalent condition for users of the Veterans Affairs (VA) healthcare system, affecting 16%.1 Gout is the commonest inflammatory arthritis in men1–4 with an increasing prevalence, partly due to increased obesity, diuretic use and longevity.5 6 In 2002, gout accounted for 1.4 million outpatient visits in the USA, indicating it to be a significant public health problem.7 Despite being a chronic arthritic condition, effective and affordable treatments are available for gout that may have an impact and improve outcomes.
Current evidence indicates that gout is underdiagnosed and undertreated—care is suboptimal.8–10 Ascertaining the impact of gout on subjects’ functional status, health-related quality of life (HRQoL) and associated healthcare burden may improve care patterns through an increased awareness of disease outcomes. Only two small studies11 12 and an abstract13 have evaluated functional limitation patterns in patients with gout, indicating that 10–42% of patients reported limitations in activities11 12 and 4.6 more days of absence from work/year.13 Most studies of the effect of gout on health status were performed more than two decades ago14 15 and with the exception of one recent publication16 reported the frequency of recurrent acute gouty arthritis. There has been limited recognition that health status/HRQoL in gout may be related not only to recurrent flares/chronic joint disease, but more importantly to associated medical comorbidities, including hypertension, cardiac and renal disease.17 18 More effective control of gout is cost effective,19 and it is likely that better management of associated comorbidities will significantly improve patients’ lives.
Much debate in recent decades has focused on whether hyperuricaemia is an independent risk factor for cardiovascular events and cardiac or all-cause mortality.20 In a recent analysis of the Multiple Risk Factor Intervention Trial (MRFIT), an independent association was found between gout and acute myocardial infarction, but the risk of mortality secondary to acute myocardial infarction was not increased in patients with gout.21 Higher mortality was noted in one treatment arm (4/507 vs 0/253) in a recent randomised trial comparing anti-hyperuricaemic drugs in patients with gout.22 To our knowledge, no population-based studies (prospective or retrospective) have examined whether all-cause mortality is increased in gout cohorts, and whether this association is independent of differences in age, gender and comorbidity between populations with and without gout.
This project was designed to ascertain functional status, HRQoL, healthcare utilisation and mortality patterns in veterans with and without gout. The veteran population was selected for several reasons: (a) it offers a large sample size for a disease predominant in men—in 2003 the Department of Veterans Affairs (VA) served approximately 4.9 million subjects with a budget of $25 billion23; (b) other arthritic conditions are very common in this population1 and (c) state of the art validated VA databases capture demographics, diagnoses, healthcare utilisation and mortality.24 25 Objectives were to (a) compare patient-reported functional status, HRQoL, healthcare utilisation patterns and mortality in veterans with and without gout; (b) determine if these were independently associated with a diagnosis of gout and (c) determine whether and in what aspects medical or other arthritic (non-gout) comorbidities affect HRQoL in veterans with gout.
METHODS
We identified a cohort of veterans who used Upper Midwest Veterans Integrated Service Network (VISN-13: Minnesota, North Dakota and South Dakota, and selected counties in Iowa, Nebraska, Wisconsin and Wyoming) for healthcare services between 1 October1996 and 31 March 1998 and had a valid mailing addresses (n = 70 334). We then conducted a cross-sectional mailed survey in August 1998 querying demographics (education level, race/ethnicity), performance of six activities of daily living (ADLs) including bathing, dressing, eating, walking, transferring and using the toilet26 and HRQoL using the Short Form-36 for veterans (SF-36V) (Details previously published as VISN-13 Vet-QOL Study27 28.) Non-responders received the same survey 10 weeks later.
SF-36 is a generic measure of HRQoL that is valid, reliable and responsive to change in patients with osteoarthritis, rheumatoid and psoriatic arthritis.29–32 SF-36V is a modified version, validated in veterans,33 where role physical and role emotional domains were changed from dichotomous to five-choice ordinal scales to avoid floor and ceiling effects; it is similar to SF-36, version 2.0.34 We calculated eight SF-36V subscales (range from 0 (worst) to 100 (best)), and two summary scores, the physical and mental component summary scores (PCS and MCS; normative distributions with mean of 50 and standard deviations of 10 for the US population) in a standard fashion.35 36 Survey data were supplemented by demographics, International Classification of Diseases (ICD)-9 codes and healthcare utilisation data from the national VA databases.
Predictor variables
The main predictor of functional status, HRQoL, healthcare utilisation and mortality was presence or absence of ICD-9 diagnosis of gout. The multivariable analyses adjusted for the following covariates and potential confounders:
Sociodemographic: (a) age (in years), employment status (employed, unemployed, retired, unknown), marital status (married, not married; database-derived); (b) gender (male, female), race (white, non-white), education level (less than 8th grade, some high school, high school graduate, college and beyond; survey-derived).
Comorbidity: (a) Current smoking status: current smoker versus non-smoker (survey-derived); (b) presence or absence of ICD-9 diagnoses of five common medical comorbidities in veterans1: asthma/chronic obstructive pulmonary disease, diabetes, depression, hypertension, heart disease (database-derived); (c) presence or absence of arthritic comorbidities in nine categories: rheumatoid arthritis, osteoarthritis, spondyloarthropathies, crystal-associated arthritis except gout, unspecified arthritis, infectious arthritis, arthritis associated with haemachromatosis, arthritis associated with endocrine or metabolic diseases, spondylosis (database-derived).
Database-derived variables were available for most of the sample (∼70 000) and survey-derived variables for ∼35 000–40 000 responders to the questionnaire. Medical and arthritic (non-gout) comorbidities were categorised into three categories (none, one or two or more), as only 7% and 1%, respectively, had three or more comorbidities.
Study outcomes
Outcomes of this current study included: (a) percentage of subjects with limitations in each ADL and number of ADLs with limitations for each subject; (b) SF-36V domain, PCS and MCS scores as continuous variables; (c) healthcare utilisation 1 year after the survey: outpatient primary care, medical subspecialty, surgery clinic visits as continuous variables and inpatient and mental health visits as categorical variables (none vs one or more), owing to a right skewed distribution and (d) mortality 1 year after the survey. Health care utilisation and mortality data were obtained for 1 year after the index survey date and ICD-9 codes before the survey, from inpatient and outpatient VA datasets. These datasets are reliable for demographics and valid for most common,25 and specific diagnoses.37 38 In a random sample of patients with an ICD-9 code for gout in the Minneapolis VA database, we found that this diagnosis was supported by medical chart documentation in 78%.39
Definitions of minimal clinically important differences were based on changes of 5–10 points in SF-36 domain and 2.5–5 points in summary component scores in published randomised controlled trials; consistently similar in various rheumatic conditions.40–43 Although considered to represent the minimum improvement “perceptible” to patients when enrolled in therapeutic trials, minimal clinically important differences offer a reasonable “benchmark” for comparisons across patient and disease populations.
Statistical analyses
Continuous and categorical variables between subjects with and without gout were compared between survey responders and non-responders using the Student t test and χ2 testing. We used multiple logistic regression models to compare the proportion of patients with and without gout with: limitation of each ADL; any hospitalisation, any mental health clinic visit or mortality 1 year after the survey. Multiple linear regression analyses modelled the number of primary care, specialty and surgical care visits over the year after the survey and the SF-36V domain and summary scores as outcomes. These regression model-estimated probabilities were calculated using average values of the additional covariates (as described above). All analyses were performed using SPSS, version 11.5 and SAS, version 9.0. Owing to multiple comparisons performed, p values <0.01 were considered significant.
RESULTS
Demographic and comorbidity characteristics
Overall 58% responded to the survey (40 508 of the 70 334 eligible). Of the 70 334 (demographic and utilisation variables were available for most), 1581 (2.2%) had an ICD-9 diagnosis for gout; of the 40 508 survey respondents (HRQoL and functional limitation data were available for this group only), 1090 (2.7%) had an ICD-9 diagnosis for gout. Veterans with gout were significantly older, male, not married, retired and current non-smokers (table 1).
Non-responders to the survey were significantly younger (56.3 vs 64.5 years), less likely to be married (47% vs 64.9%), less likely to be retired (26.7% vs 43.9%) and had smaller number of outpatient visits (eight vs nine visits/year) but had similar inpatient utilisation rates (13.1% with hospitalisation vs 13%), as compared with responders.
The higher prevalence of unadjusted medical and arthritic comorbidity in patients with gout as described in table 1 persisted after adjustment for age, sex and race (1.4 vs 0.8 and 0.5 vs 0.2, respectively; p<0.001 for both).
Functional limitations and decrements in HRQoL in patients with gout
The absolute unadjusted differences of 9% and 6.5% limitation in transferring and walking between patients with and patients without gout were not significant after multivariable adjustment (fig 1A). Veterans with gout reported significant, marginally higher unadjusted mean number of limitations in ADLs than those without (1.7 vs 1.6, p = 0.016); no longer significant after multivariable analyses (1.5 vs 1.6, p = 0.88).
In unadjusted analyses, subjects with gout reported statistically significant and clinically meaningful lower physical HRQoL: decrements of 5.8–7 points in physical functioning, bodily pain, role physical domains and three points in PCS scores (fig 1B). After multivariable adjustment, none of these domain and PCS score differences were statistically or clinically significant except that bodily pain remained slightly lower in those with gout by 2.6 points, but this did not meet the clinically meaningful difference of 5–10 points. Statistically significant differences (that were small and not clinically meaningful) were noted in mental/emotional and general health HRQoL domain scores in unadjusted analyses; not evident after multivariable adjustment (fig 1C). MCS domain scores were not statistically (or clinically) different in unadjusted or adjusted analyses.
Effect of medical versus arthritic comorbidity on HRQoL in patients with gout
After multivariable adjustment, compared with those without gout, patients with gout with one or more medical comorbidities reported significantly lower scores in all SF-36V domains except bodily pain and vitality (7.1–11 point decrements) and PCS and MCS scores (3.1–4.2 point decrements), all of which are clinically meaningful (fig 2A). Those with one or more arthritic comorbidities reported statistically significant and clinically meaningful lower scores in physical HRQoL domains—that is, physical functioning, role physical, bodily pain domains (7.3–10 point decrements) and PCS (5.8 point decrement) (fig 2B). General health, vitality and social functioning domain scores were also statistically significantly and clinically meaningfully lower, by a smaller magnitude (4.5–6.3 point decrements), but no significant differences were noted in role emotional and mental health domain and MCS scores.
Healthcare utilisation and mortality in subjects with and without gout
Multivariable adjustments attenuated unadjusted differences in specialty and surgical care visits (1.66 vs 1.52, p = 0.35 and 2.07 vs 1.89, p = 0.06, respectively, 9.5% higher in patients with gout), but primary care visits and inpatient admissions remained 31% and 21% higher in those with gout (p<0.001) (table 2). Multivariable-adjusted visits to mental healthcare clinics were lower in patients with gout versus patients without gout (10.1% vs 13.7%, 26% lower; p<0.01). Details of the regression models are given in Appendix 1. Unadjusted 1-year mortality was higher in veterans with gout than in those without (5.8% vs 3.9%, 50% higher; p<0.01); after multivariable adjustments, it was similar in both groups (2.6 vs 2.2%, 18% higher, respectively; p>0.05).
DISCUSSION
The two most important findings in this study included (a) significantly more primary care outpatient clinic and inpatient healthcare use by veterans with gout; largely attributable to gout and partially to comorbidities and sociodemographic characteristics; and (b) poorer functional ability and HRQoL, increased specialty medicine and surgical clinic use and higher mortality, primarily attributable to sociodemographic and (medical and arthritic) comorbidities. Since increased primary care and inpatient utilisation appear to be related to gout, further studies should examine which factor/factors are most predictive: disease severity; patient (knowledge about the disease and its treatment, medication compliance); doctor (knowledge, management of arthritic symptoms); treatment (adverse effects) and system factors (quality of care for gout). Gaps in doctors’ and patients’ knowledge44 and suboptimal care for gout have been well described,8 10 therefore future studies should explore whether this is associated with increased health services utilisation or poorer outcomes, or both. A better understanding of why patients use more health services will not only allow us to reduce gout-related healthcare utilisation but also enable us to improve treatment and patient outcomes.
To our knowledge, this is the first population-based assessment of HRQoL, utilisation and functional ability in a large cohort with gout. A recently published UK study reported lower physical and overall HRQoL in patients with gout than in patients without gout in univariate analyses, but only lower physical HRQoL after multivariable adjustment.16 Our study confirms and extends these HRQoL findings to a much larger population-based US cohort using SF-36V as the HRQoL measure, but differs in important aspects from the previous study in that (a) no differences were noted in physical HRQoL after adjustment for sociodemographics and comorbidity as opposed to the earlier study; (b) response was higher, 58% versus 23%; (c) the study sample was larger, 1500 versus 137 patients with gout; (d) this was a population-based survey of 70 000 adults in a three states of the USA compared with patients >30 years from two general practices in the UK; (e) different cohorts were studied, 98% men with mean age of 68 years versus 81% men with mean age of 64 years; (f) comorbidity assessment was made by ICD-9 codes rather than by self-reporting and (g) different HRQoL assessments were made using SF-36 compared with WHO-QoL-Bref.
Another significant finding in our study is the 50% higher unadjusted 1-year mortality rate in veterans with gout, largely explained by higher comorbidity and sociodemographic differences. This finding emphasises that gout is a chronic disease with significant associated comorbidities (heart disease, diabetes, hypertension, etc), and that mortality is attributable to these comorbid conditions. This important observation in a population-based study in addition to the recent report of similar myocardial infarction-related mortality in gout versus non-gout from the MRFIT data,21 indicates that increased mortality in patients with gout may be related to other comorbidities and sociodemographics.
That a greater proportion of patients with gout reported limitations in transferring and walking in unadjusted analyses is consistent with predominantly monarticular and/or lower extremity involvement in this disease. However, we were surprised to find that in multivariable analyses, these ADL limitations were attributable to differences in age, sociodemographics and comorbidities, implying that variables other than gout conferred difficulty in walking/transferring. Such limitations in younger employed men with gout may have an impact on work productivity and social functioning, as reported in two recent abstracts.13 45
Compared with normative US data matched for age and gender, domain scores (except role emotional, mental health and MCS) were 15–25 points lower across all gout populations—with and without medical and/or arthritic comorbidities; again reflecting the broad impact of this disease upon HRQoL, another significant finding. SF-36V bodily pain domain scores in patients with gout showed larger decrements with arthritic than with medical comorbidities (9 vs 2.7 point decrement), indicating that the focus of care in subjects with gout and more medical comorbidities should be to improve both physical and mental/emotional HRQoL, whereas it should be to improve physical function and pain management in those with gout and other arthritic conditions.
This study has several limitations: results may not be generalisable to the general US population owing to non-response bias and male predominance, or to other VA networks owing to regional variation. However, the prevalence of gout in this cohort of 2.2% is similar to the 2.7% in a National probability sample.46 The analyses were unable to control for use of drugs and non-VA health services in the sample—both may have had an impact on outcomes. Although data were collected for most common comorbidities, information about renal failure, obesity, metabolic syndrome and alcohol use were not specifically queried owing to limited resources; these may be more common in patients with gout and may be responsible for residual confounding in the adjusted analyses. However, adjustment for hypertension, diabetes and cardiac disease should have partially accounted for these factors. Ascertainment error in diagnosis of gout may be present as reported in a recent study in a Health Maintenance Organisation population47 and our observation of 22% inaccuracy of ICD-9 codes for gout,39 which probably biased our study towards null and may have masked some differences. Cross-sectional measurement of HRQoL is likely to have underestimated the effect of acute versus intercritical gout on HRQoL, which can be estimated in a cohort study.
This population-based cohort with a large sample size has numerous strengths. Healthcare utilisation, functional limitations and HRQoL were examined and analyses controlled for clinical and demographic differences including socioeconomic status. Utilisation patterns for veterans reported here may provide meaningful information for policy makers; many veterans eligible for Medicare use dual services.48
In summary, this study documents higher primary care utilisation and inpatient utilisation by veterans with gout, which can only partially be explained by differences in comorbidities and sociodemographic characteristics. Higher mortality, poorer HRQoL and functional limitation in patients with gout appears not to be due to gout itself, but to comorbidity and sociodemographic differences, suggesting that gout is a marker for higher medical comorbidity load, which indicates poor prognosis and higher all-cause mortality. Therefore a diagnosis of gout can be used to identify a high-risk group. Medical and non-gout arthritic comorbidity has a different impact on HRQoL in patients with gout. These findings suggest that more intensive management of comorbidities and of arthritis associated with gout will probably improve outcomes in patients with gout.
Acknowledgments
We thank David Nelson and Alisha Baines of the Center for Chronic Diseases Outcomes and Research, a centre for excellence at the Minneapolis VA Medical Center, for their help in statistical analyses and production of graphs.
Appendix 1
REFERENCES
Footnotes
Funding: Supported by a VA Scholar Grant from the Center for Epidemiological and Clinical Research, Minneapolis VA Medical Center, Minneapolis, Minnesota, USA
Competing interests: None.
Ethics approval: Ethics committee approval obtained.
The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs.