Archives of Physical Medicine and Rehabilitation journal homepage: www.archives-pmr.org Archives of Physical Medicine and Rehabilitation 2015;-:-------
ORIGINAL ARTICLE
Smoking and Physical Activity: Examining Health Behaviors and 15-Year Mortality Among Individuals With Multiple Sclerosis Aaron P. Turner, PhD,a,b,c,d Narineh Hartoonian, PhD,a,b,d Charles Maynard, PhD,a Steven L. Leipertz, PhD,a,b Jodie K. Haselkorn, MD, MPHa,b,d,e From the aVA Puget Sound Health Care System, Seattle, WA; bVA Multiple Sclerosis Center of Excellence West, Seattle, WA; cVA Center of Excellence in Substance Abuse Treatment and Education, Seattle, WA; dDepartment of Rehabilitation Medicine, University of Washington School of Medicine, Seattle, WA; and eDepartment of Epidemiology, University of Washington, Seattle, WA.
Abstract Objectives: To examine 2 modifiable health behaviorsdsmoking and physical activitydand their relationship to mortality among individuals with multiple sclerosis (MS). Design: Secondary analysis of Large Health Survey. Setting: Data were obtained from a linkage of the Veterans Affairs (VA) MS National Data Repository, containing information on service provision to all individuals with MS receiving health services within the U.S. Department of Veterans Affairs; the VA 1999 Large Health Survey, containing information on smoking and physical activity; and the VA Vital Status File. All-cause mortality was examined for the 15-year period from 1999 through 2013. Participants: Participants (NZ2994) with MS who completed the Large Health Survey containing information on smoking and physical activity. Interventions: Not applicable. Main Outcome Measure: Survival. Results: There were 1500 deaths (50.1%) during the study period. Cox proportional hazard analyses were conducted to examine the association between smoking and physical activity and 15-year mortality. After adjusting for demographic factors, physical functioning, mental health, and comorbid medical conditions, baseline smoking was associated with greater mortality (hazard ratio [HR]Z1.38; 95% confidence interval [CI], 1.184e1.60). Higher levels of baseline physical activity were associated with lower mortality (activity 1e2 times/wk: HRZ.64; 95% CI, .518e .798; activity 3 times/wk: HRZ.53; 95% CI, .388e.715). Conclusions: Results suggest that modifiable health behaviors represent a promising opportunity for intervention to improve the lives of individuals with MS. Archives of Physical Medicine and Rehabilitation 2015;-:------ª 2015 by the American Congress of Rehabilitation Medicine
Multiple sclerosis (MS) is an inflammatory and degenerative disease of the central nervous system that is estimated to affect as many as 1 million people worldwide.1 It is associated with a Presented in part to the American Board of Rehabilitation Psychology and the Division of Rehabilitation Psychology (Division 22) of the American Psychological Association, March 1, 2014, San Antonio, TX. Supported by a Veterans Affairs (VA) Rehabilitation Research and Development Service Career Development Award (grant no. B3319VA), the VA Center of Excellence in Substance Abuse Treatment and Education, the VA Multiple Sclerosis Center of Excellence West, VA Office of Quality and Performance, and National Multiple Sclerosis Society (grant no. MB0026). Disclosures: none.
variety of unpredictable yet often chronic impairments that may include motor, sensory, and cognitive deficits, fatigue, pain, and depression.2-4 As a result, there is considerable interest in understanding factors contributing to morbidity and mortality in this disease. Information about risk factors contributing to mortality among individuals with MS is limited. Available evidence suggests increased rates of mortality associated with disease-related variables such as primary progressive course and later age of onset.5 Additional risk of mortality is also associated with comorbid
0003-9993/14/$36 - see front matter ª 2015 by the American Congress of Rehabilitation Medicine http://dx.doi.org/10.1016/j.apmr.2014.10.014
2
A.P. Turner et al
medical conditions such as infectious and respiratory disease, cardiovascular disease, and to a lesser degree, depression.5-7 Less is known about the contribution of health behaviors to mortality in individuals with MS. Identifying behavioral risk factors is particularly promising because they are often modifiable, and positive changes in behavior frequently result in significant improvements in health and longevity. In the general population, smoking has been linked to greater mortality from a variety of causes,8 and smoking cessation is associated with significant reduction in that risk.9 Similarly, low levels of physical activity are also associated with higher rates of mortality,10 and greater levels of physical activity bring a lower risk of death.11 The current study was conducted to better understand the relationship between modifiable health behaviors and mortality among individuals with MS. Specifically, it examined physical activity and smoking and their association with 15-year all-cause mortality in a large national sample of veterans with MS. We hypothesized that after adjusting for demographics, physical and mental health functioning, and medical comorbidity, smoking and lower levels of physical activity would independently and additionally be associated with mortality. The end goal was to better quantify health risks specific to MS that could be used in health care encounters to provide education and promote behavior change in this population.
Methods Participants Participants were drawn from the U.S. Department of Veterans Affairs (VA) MS National Data Repository that contains information on all veterans who received MS-related health care services within the Veterans Health Administration. The repository is updated continually, and it contained 32,009 unique cases at the time of the data extraction for this study. To reduce the likelihood individuals were erroneously included in the sample because of medical coding errors, participants included in a final target population of all veterans with MS had to meet 1 or more of the following 4 criteria: (1) inpatient hospitalization for MS (International Classification of Diseases, Ninth Revision diagnostic code 340 for MS); (2) prescription of 1 of the 3 disease-modifying agents (interferon 1a, interferon 1b, or glatiramer acetate) used for treatment of MS at that time; (3) VA serviceeconnected disability for MS (the diagnosis had been confirmed through a medical review process); or (4) at least 1 outpatient encounter for which the primary International Classification of Diseases diagnosis code was 340 for each year in which VA care was received. Our method for identifying veterans with MS has been shown to be an effective means of identifying individuals who do not have MS12 and has been used in previous investigations of health behavior in MS.13,14 A total of
List of abbreviations: HR LVHS MCS MS PCS SF-36v
hazard ratio Large Veteran Health Survey Medical Component Summary multiple sclerosis Physical Component Summary Medical Outcomes Study 36-Item Short-Form Health SurveyeVeterans version SIC Seattle Index of Comorbidity VA Veterans Affairs
17,470 veterans with MS were included in the final target population covering the period 1999 through 2013. Data from the target population were then linked to the VA 1999 Large Veteran Health Survey (LVHS),15 a nationally representative sample of veterans receiving care across the VA health care system. Of the 1.4 million enrollees who were mailed surveys, the overall LVHS was returned by 877,775 (63.1% response rate). Similarly, 2994 of 4685 individuals with a diagnosis of MS returned the mailed surveys (63.9% response rate). All procedures were approved by the local institutional review board.
Measures Demographic information Sex, race (non-Hispanic whites vs other), education level (less than high school vs high school graduate plus some college vs college graduate or graduate school), and marital status (currently married vs separated/divorced/widowed vs never married) were obtained from the LVHS. Age in years at the time of the survey was obtained from the VA MS National Data Repository. Missing values in the LVHS race variable were replaced with values from the repository race variable where available. Smoking Cigarette use was measured with a question adapted from the Behavioral Risk Factors Surveillance System.16 Participants were asked, “Do you now smoke cigarettes every day, some days, or not at all.” Responses were dichotomized to reflect the presence or absence of current smoking. Physical activity Physical activity was measured with the question “How often do you engage in regular activities (eg, brisk walking, jogging, bicycling) long enough to work up a sweat?” Categorical responses ranged from never to more than 5 times per week. This single-item indicator of physical activity has been widely used in epidemiologic research and has demonstrated good construct validity, correlating with laboratory-based measures of physical fitness including maximal oxygen uptake and maximal treadmill performance,17,18 as well as convergent validity, correlating with other longer questionnaire assessments of physical activity.19 Lower levels of physical activity reported on this single item have been associated with an increased risk for cancer, diabetes, and stroke.2022 Responses were grouped into 3 categories indicating engagement in no activity, activity 1 or 2 times per week, or 3 times per week. Physical disability Physical disability was measured using the Physical Component Summary (PCS) score of the Medical Outcomes Study 36-Item Short-Form Health SurveyeVeterans version (SF-36v), a version of the Medical Outcomes Study 36-Item Short-Form Health Survey adapted for use with veterans. The SF-36v has been shown to be a reliable and valid measure of health status relevant to the assessment of disability, disease burden, and morbidity.23 The responses from this survey can be used to calculate 2 global ratings of physical functioning and mental health. The Medical Outcomes Study 36-Item Short-Form Health Survey has been shown to be sensitive to changes in physical abilities among individuals with MS, predict subsequent disability,24 and be correlated with disease-specific instruments such as the Expanded Disability Status Scale.25 Higher PCS scores reflect better physical functioning. www.archives-pmr.org
Health behavior and mortality in multiple sclerosis Mental health Mental health was similarly measured using the Mental Component Summary (MCS) of the SF-36v. Higher MCS scores reflect better mental health. Medical comorbidity Medical comorbidity was measured using the Seattle Index of Comorbidity (SIC).26 The SIC is a weighted composite of selfreported medical conditions combined with age and current smoking status into a single score reflecting total medical comorbidity. The SIC score is predictive of both mortality and hospitalization.26 Self-reported medical conditions were obtained from Large Health Survey data. For the present study, both age and smoking were independent variables of interest, so these 2 variables were not included in the comorbidity index. Mortality Veterans who died between January 1, 1999, and December 31, 2013, were identified using the VA Vital Status File. The VA Vital Status File is an aggregate of VA administrative death data, the Medicare Vital Status File, and the Social Security Administration Death Master File. It has been shown to have greater than 98% sensitivity and 99% specificity relative to the National Death Index in general, and 100% sensitivity among users of VA inpatient and outpatient services (similar to the current study population).27 The primary outcome was all-cause mortality, with survival defined as the time to death or censoring in years.
Data analytic strategy Sample selection and representativeness To determine the extent to which the final study sample was representative of the larger population of veterans with MS, we compared demographic characteristics such as age, sex, and race. We also examined selection bias (whether an individual was or was not sent a survey) and response bias (whether an individual did or did not return a survey) for these 3 variables. The prevalence of current smoking and physical activity was estimated using simple proportions. Data were inspected to ensure that underlying assumptions of the statistical methods were not violated. An evaluation of descriptive statistics and histograms was performed to assess univariate normality and linearity. In addition, Cook’s distance was used to screen for multivariate outliers. No extreme cases were identified. Proportional hazards assumptions were tested using cumulative martingale residual plots for continuous covariates and the log-log transformation of the estimated probability of survival from the Kaplan-Meier analysis for categorical covariates. Data indicated acceptable fit for all continuous and categorical covariates. Health behavior and mortality To study the adjusted effect of health behaviors on mortality, we used multivariate Cox proportional hazards regression analysis. More specifically, we examined whether smoking and engaging in physical activity were associated with 15-year all-cause mortality in veterans with MS after adjusting for demographics (age, sex, race, education, marital status) and medical comorbidities associated with mortality.26 To adjust for the effects of physical disability and mental health, the PCS score and MCS score were also included. www.archives-pmr.org
3 All predictors were entered simultaneously, and the significance of the overall model was evaluated with the chi-square statistic. The significance of the individual predictors was established through evaluation of the Wald statistic, and hazard ratios (HRs) and 95% confidence intervals were assessed to determine the effect of a 1-unit increase in a predictor variable on the risk of 15-year mortality. No time-varying covariates were used in the analysis. Survival curves were constructed to illustrate the association between the 3 categories of physical activity (no activity in a given week, activity 1e2 times/wk, or 3 times/wk), smoking (current smoker or not), and mortality. Of the 2994 veterans who returned the mailed Large Health Survey (sample size used in the analysis), 18% had missing response on the race variable, 10.1% had missing data on the SF36v, 6.9% on education, 6% on sex, 6% on physical activity, 6.2% on smoking, 4% on the comorbidities, and 3.7% on marital status. The expectation-maximization algorithm through missing value analysis was used to detect randomly missing data. Findings demonstrate that the missing data deviate significantly from randomness using Little’s missing completely at random test (c29Z53.34, P
ORIGINAL ARTICLE
Smoking and Physical Activity: Examining Health Behaviors and 15-Year Mortality Among Individuals With Multiple Sclerosis Aaron P. Turner, PhD,a,b,c,d Narineh Hartoonian, PhD,a,b,d Charles Maynard, PhD,a Steven L. Leipertz, PhD,a,b Jodie K. Haselkorn, MD, MPHa,b,d,e From the aVA Puget Sound Health Care System, Seattle, WA; bVA Multiple Sclerosis Center of Excellence West, Seattle, WA; cVA Center of Excellence in Substance Abuse Treatment and Education, Seattle, WA; dDepartment of Rehabilitation Medicine, University of Washington School of Medicine, Seattle, WA; and eDepartment of Epidemiology, University of Washington, Seattle, WA.
Abstract Objectives: To examine 2 modifiable health behaviorsdsmoking and physical activitydand their relationship to mortality among individuals with multiple sclerosis (MS). Design: Secondary analysis of Large Health Survey. Setting: Data were obtained from a linkage of the Veterans Affairs (VA) MS National Data Repository, containing information on service provision to all individuals with MS receiving health services within the U.S. Department of Veterans Affairs; the VA 1999 Large Health Survey, containing information on smoking and physical activity; and the VA Vital Status File. All-cause mortality was examined for the 15-year period from 1999 through 2013. Participants: Participants (NZ2994) with MS who completed the Large Health Survey containing information on smoking and physical activity. Interventions: Not applicable. Main Outcome Measure: Survival. Results: There were 1500 deaths (50.1%) during the study period. Cox proportional hazard analyses were conducted to examine the association between smoking and physical activity and 15-year mortality. After adjusting for demographic factors, physical functioning, mental health, and comorbid medical conditions, baseline smoking was associated with greater mortality (hazard ratio [HR]Z1.38; 95% confidence interval [CI], 1.184e1.60). Higher levels of baseline physical activity were associated with lower mortality (activity 1e2 times/wk: HRZ.64; 95% CI, .518e .798; activity 3 times/wk: HRZ.53; 95% CI, .388e.715). Conclusions: Results suggest that modifiable health behaviors represent a promising opportunity for intervention to improve the lives of individuals with MS. Archives of Physical Medicine and Rehabilitation 2015;-:------ª 2015 by the American Congress of Rehabilitation Medicine
Multiple sclerosis (MS) is an inflammatory and degenerative disease of the central nervous system that is estimated to affect as many as 1 million people worldwide.1 It is associated with a Presented in part to the American Board of Rehabilitation Psychology and the Division of Rehabilitation Psychology (Division 22) of the American Psychological Association, March 1, 2014, San Antonio, TX. Supported by a Veterans Affairs (VA) Rehabilitation Research and Development Service Career Development Award (grant no. B3319VA), the VA Center of Excellence in Substance Abuse Treatment and Education, the VA Multiple Sclerosis Center of Excellence West, VA Office of Quality and Performance, and National Multiple Sclerosis Society (grant no. MB0026). Disclosures: none.
variety of unpredictable yet often chronic impairments that may include motor, sensory, and cognitive deficits, fatigue, pain, and depression.2-4 As a result, there is considerable interest in understanding factors contributing to morbidity and mortality in this disease. Information about risk factors contributing to mortality among individuals with MS is limited. Available evidence suggests increased rates of mortality associated with disease-related variables such as primary progressive course and later age of onset.5 Additional risk of mortality is also associated with comorbid
0003-9993/14/$36 - see front matter ª 2015 by the American Congress of Rehabilitation Medicine http://dx.doi.org/10.1016/j.apmr.2014.10.014
2
A.P. Turner et al
medical conditions such as infectious and respiratory disease, cardiovascular disease, and to a lesser degree, depression.5-7 Less is known about the contribution of health behaviors to mortality in individuals with MS. Identifying behavioral risk factors is particularly promising because they are often modifiable, and positive changes in behavior frequently result in significant improvements in health and longevity. In the general population, smoking has been linked to greater mortality from a variety of causes,8 and smoking cessation is associated with significant reduction in that risk.9 Similarly, low levels of physical activity are also associated with higher rates of mortality,10 and greater levels of physical activity bring a lower risk of death.11 The current study was conducted to better understand the relationship between modifiable health behaviors and mortality among individuals with MS. Specifically, it examined physical activity and smoking and their association with 15-year all-cause mortality in a large national sample of veterans with MS. We hypothesized that after adjusting for demographics, physical and mental health functioning, and medical comorbidity, smoking and lower levels of physical activity would independently and additionally be associated with mortality. The end goal was to better quantify health risks specific to MS that could be used in health care encounters to provide education and promote behavior change in this population.
Methods Participants Participants were drawn from the U.S. Department of Veterans Affairs (VA) MS National Data Repository that contains information on all veterans who received MS-related health care services within the Veterans Health Administration. The repository is updated continually, and it contained 32,009 unique cases at the time of the data extraction for this study. To reduce the likelihood individuals were erroneously included in the sample because of medical coding errors, participants included in a final target population of all veterans with MS had to meet 1 or more of the following 4 criteria: (1) inpatient hospitalization for MS (International Classification of Diseases, Ninth Revision diagnostic code 340 for MS); (2) prescription of 1 of the 3 disease-modifying agents (interferon 1a, interferon 1b, or glatiramer acetate) used for treatment of MS at that time; (3) VA serviceeconnected disability for MS (the diagnosis had been confirmed through a medical review process); or (4) at least 1 outpatient encounter for which the primary International Classification of Diseases diagnosis code was 340 for each year in which VA care was received. Our method for identifying veterans with MS has been shown to be an effective means of identifying individuals who do not have MS12 and has been used in previous investigations of health behavior in MS.13,14 A total of
List of abbreviations: HR LVHS MCS MS PCS SF-36v
hazard ratio Large Veteran Health Survey Medical Component Summary multiple sclerosis Physical Component Summary Medical Outcomes Study 36-Item Short-Form Health SurveyeVeterans version SIC Seattle Index of Comorbidity VA Veterans Affairs
17,470 veterans with MS were included in the final target population covering the period 1999 through 2013. Data from the target population were then linked to the VA 1999 Large Veteran Health Survey (LVHS),15 a nationally representative sample of veterans receiving care across the VA health care system. Of the 1.4 million enrollees who were mailed surveys, the overall LVHS was returned by 877,775 (63.1% response rate). Similarly, 2994 of 4685 individuals with a diagnosis of MS returned the mailed surveys (63.9% response rate). All procedures were approved by the local institutional review board.
Measures Demographic information Sex, race (non-Hispanic whites vs other), education level (less than high school vs high school graduate plus some college vs college graduate or graduate school), and marital status (currently married vs separated/divorced/widowed vs never married) were obtained from the LVHS. Age in years at the time of the survey was obtained from the VA MS National Data Repository. Missing values in the LVHS race variable were replaced with values from the repository race variable where available. Smoking Cigarette use was measured with a question adapted from the Behavioral Risk Factors Surveillance System.16 Participants were asked, “Do you now smoke cigarettes every day, some days, or not at all.” Responses were dichotomized to reflect the presence or absence of current smoking. Physical activity Physical activity was measured with the question “How often do you engage in regular activities (eg, brisk walking, jogging, bicycling) long enough to work up a sweat?” Categorical responses ranged from never to more than 5 times per week. This single-item indicator of physical activity has been widely used in epidemiologic research and has demonstrated good construct validity, correlating with laboratory-based measures of physical fitness including maximal oxygen uptake and maximal treadmill performance,17,18 as well as convergent validity, correlating with other longer questionnaire assessments of physical activity.19 Lower levels of physical activity reported on this single item have been associated with an increased risk for cancer, diabetes, and stroke.2022 Responses were grouped into 3 categories indicating engagement in no activity, activity 1 or 2 times per week, or 3 times per week. Physical disability Physical disability was measured using the Physical Component Summary (PCS) score of the Medical Outcomes Study 36-Item Short-Form Health SurveyeVeterans version (SF-36v), a version of the Medical Outcomes Study 36-Item Short-Form Health Survey adapted for use with veterans. The SF-36v has been shown to be a reliable and valid measure of health status relevant to the assessment of disability, disease burden, and morbidity.23 The responses from this survey can be used to calculate 2 global ratings of physical functioning and mental health. The Medical Outcomes Study 36-Item Short-Form Health Survey has been shown to be sensitive to changes in physical abilities among individuals with MS, predict subsequent disability,24 and be correlated with disease-specific instruments such as the Expanded Disability Status Scale.25 Higher PCS scores reflect better physical functioning. www.archives-pmr.org
Health behavior and mortality in multiple sclerosis Mental health Mental health was similarly measured using the Mental Component Summary (MCS) of the SF-36v. Higher MCS scores reflect better mental health. Medical comorbidity Medical comorbidity was measured using the Seattle Index of Comorbidity (SIC).26 The SIC is a weighted composite of selfreported medical conditions combined with age and current smoking status into a single score reflecting total medical comorbidity. The SIC score is predictive of both mortality and hospitalization.26 Self-reported medical conditions were obtained from Large Health Survey data. For the present study, both age and smoking were independent variables of interest, so these 2 variables were not included in the comorbidity index. Mortality Veterans who died between January 1, 1999, and December 31, 2013, were identified using the VA Vital Status File. The VA Vital Status File is an aggregate of VA administrative death data, the Medicare Vital Status File, and the Social Security Administration Death Master File. It has been shown to have greater than 98% sensitivity and 99% specificity relative to the National Death Index in general, and 100% sensitivity among users of VA inpatient and outpatient services (similar to the current study population).27 The primary outcome was all-cause mortality, with survival defined as the time to death or censoring in years.
Data analytic strategy Sample selection and representativeness To determine the extent to which the final study sample was representative of the larger population of veterans with MS, we compared demographic characteristics such as age, sex, and race. We also examined selection bias (whether an individual was or was not sent a survey) and response bias (whether an individual did or did not return a survey) for these 3 variables. The prevalence of current smoking and physical activity was estimated using simple proportions. Data were inspected to ensure that underlying assumptions of the statistical methods were not violated. An evaluation of descriptive statistics and histograms was performed to assess univariate normality and linearity. In addition, Cook’s distance was used to screen for multivariate outliers. No extreme cases were identified. Proportional hazards assumptions were tested using cumulative martingale residual plots for continuous covariates and the log-log transformation of the estimated probability of survival from the Kaplan-Meier analysis for categorical covariates. Data indicated acceptable fit for all continuous and categorical covariates. Health behavior and mortality To study the adjusted effect of health behaviors on mortality, we used multivariate Cox proportional hazards regression analysis. More specifically, we examined whether smoking and engaging in physical activity were associated with 15-year all-cause mortality in veterans with MS after adjusting for demographics (age, sex, race, education, marital status) and medical comorbidities associated with mortality.26 To adjust for the effects of physical disability and mental health, the PCS score and MCS score were also included. www.archives-pmr.org
3 All predictors were entered simultaneously, and the significance of the overall model was evaluated with the chi-square statistic. The significance of the individual predictors was established through evaluation of the Wald statistic, and hazard ratios (HRs) and 95% confidence intervals were assessed to determine the effect of a 1-unit increase in a predictor variable on the risk of 15-year mortality. No time-varying covariates were used in the analysis. Survival curves were constructed to illustrate the association between the 3 categories of physical activity (no activity in a given week, activity 1e2 times/wk, or 3 times/wk), smoking (current smoker or not), and mortality. Of the 2994 veterans who returned the mailed Large Health Survey (sample size used in the analysis), 18% had missing response on the race variable, 10.1% had missing data on the SF36v, 6.9% on education, 6% on sex, 6% on physical activity, 6.2% on smoking, 4% on the comorbidities, and 3.7% on marital status. The expectation-maximization algorithm through missing value analysis was used to detect randomly missing data. Findings demonstrate that the missing data deviate significantly from randomness using Little’s missing completely at random test (c29Z53.34, P
