- Email: [email protected]
Combined effect of individual and neighborhood socioeconomic status on mortality in patients with newly diagnosed dyslipidemia: A nationwide Korean cohort study from 2002 to 2013
Nutrition, Metabolism & Cardiovascular Diseases (2016) 26, 207e215
Available online at www.sciencedirect.com
Nutrition, Metabolism & Cardiovascular Diseases journal homepage: www.elsevier.com/locate/nmcd
Combined effect of individual and neighborhood socioeconomic status on mortality in patients with newly diagnosed dyslipidemia: A nationwide Korean cohort study from 2002 to 2013 J. Shin b,c,1, K.H. Cho a,b,1, Y. Choi a,b, S.G. Lee b,d, E.-C. Park b,c, S.-I. Jang b,c,* a
Department of Public Health, Graduate School, Yonsei University, South Korea Institute of Health Services Research, College of Medicine, Yonsei University, South Korea c Department of Preventive Medicine, College of Medicine, Yonsei University, South Korea d Department of Hospital Management, Graduate School of Public Health, Yonsei University, South Korea b
Received 3 July 2015; received in revised form 9 November 2015; accepted 14 December 2015 Available online 23 December 2015
KEYWORDS Socioeconomic status; Dyslipidemia; Hypercholesterolemia; Hypertriglyceridemia; Neighborhood; Mortality
Abstract Background and Aim: The study aims to determine whether dyslipidemia patients living in less affluent neighborhood are at a higher risk of mortality compared to those living in more affluent neighborhoods. Methods and results: A population-based cohort study was conducted using a stratified representative sampling from the National Health Insurance claim data from 2002 to 2013. The target subjects comprise patients newly diagnosed with dyslipidemia receiving medication. We performed a survival analysis using the Cox proportional hazard model. Of 11,946 patients with dyslipidemia, 1053 (8.8%) subjects died during the follow-up period. Of the dyslipidemia patients earning a middle-class income, the adjusted HR in less affluent neighborhoods was higher than that in the more affluent neighborhoods compared to the reference category of high individual SES in more affluent neighborhoods (less affluent; hazard ratio (HR) Z 1.64, 95% confidence interval (CI): 1.35e1.99 vs. more affluent; HR Z 1.48, 95% CI: 1.20e1.81, respectively). We obtained consistent results in patients with lower income, wherein the adjusted HR in less affluent neighborhoods was higher than that in more affluent neighborhoods (less affluent; HR Z 1.52, 95% CI: 1.16e1.97 vs. more affluent; HR Z 1.41, 95% CI: 1.04e1.92, respectively). Conclusion: Living in a less affluent neighborhood contributes to higher mortality among dyslipidemia patients. The individual- and neighborhood-level variables cumulatively affect individuals such that the most at-risk individuals include those having both individual- and neighborhood-level risk factors. These findings raise important clinical and public health concerns and indicate that neighborhood SES approaches should be essentially considered in health-care policies similar to individual SES. ª 2015 The Italian Society of Diabetology, the Italian Society for the Study of Atherosclerosis, the Italian Society of Human Nutrition, and the Department of Clinical Medicine and Surgery, Federico II University. Published by Elsevier B.V. All rights reserved.
* Corresponding author. Department of Preventive Medicine and Institute of Health Services Research, Yonsei University College of Medicine, 50, Yonsei-ro, Seodaemun-gu, Seoul, 120-752, South Korea. Tel.: þ82 2 2228 1863; fax: þ82 2 392 8133. E-mail address: [email protected] (S.-I. Jang). 1 J. Shin and K.H. Cho contributed equally to the manuscript. http://dx.doi.org/10.1016/j.numecd.2015.12.007 0939-4753/ª 2015 The Italian Society of Diabetology, the Italian Society for the Study of Atherosclerosis, the Italian Society of Human Nutrition, and the Department of Clinical Medicine and Surgery, Federico II University. Published by Elsevier B.V. All rights reserved.
208
Introduction Dyslipidemia is a common noncommunicable disease in Korea; in the Korean society, the incidence and prevalence of the disease increase with increasing age [1,2]. According to the Korean National Health and Nutritional Examination Survey (KHANES) in 2013, the prevalence of hypercholesterolemia and hypertriglyceridemia was 14.9% and 17.1%, respectively, among the respondents of age >30 years [3]. Upon comparing the current statistics of prevalence to those reported earlier (KHANES, 1998: 10.0% of subjects had hypercholesterolemia and 10.2% hypertriglyceridemia), we found the need for prevention of this continuous steep increase in the number of patients with dyslipidemia. Dyslipidemia can result in other noncommunicable diseases such as metabolic syndrome, cardiovascular diseases, and stroke [4e6]. Dyslipidemia might impose a large burden on human health in a developed society; however, it can be prevented and controlled by early screening and treatment with appropriate dietary strategies at both individual and population levels. According to the U.S. Preventive Services Task Force in 2008, several guidelines have been suggested for lipid disorder in adults [7]. However, whether the screening for lipid disorder is cost-effective is still debatable. In addition, the risk factors for dyslipidemia are well known [8,9]. Among them, individual socioeconomic status (SES) and neighborhood SES are associated with prevalence [10,11]. Many other diseases associated with dyslipidemia such as cardiovascular disease, stroke, and metabolic syndrome have been investigated with regard to the cross-level effects between individual and neighborhood SES. However, few studies have related these factors to dyslipidemia especially in Asians. Therefore, in this study, we first examined the effects of individual and neighborhood SES on the overall mortality rate from 2003. Then, we further investigated the crosslevel effects between individual and neighborhood SES.
J. Shin et al.
medication to investigate the association between combined individual and neighborhood SES and mortality. Ethical approval for this study was granted by the Institutional Review Board of the Graduate School of Public Health at Yonsei University. The requirement for informed consent was waived because the study was based on routinely collected administrative and claim data. Study sample selection
Methods
Among the enrolled subjects, 18,219 with newly diagnosed dyslipidemia (E78; International Classification of Disease (ICD), 10th edition) between 2003 and 2006 were selected. We confirmed that diagnoses were new by verifying a lack of dyslipidemia claims between 2002 and 2005, followed by an initial dyslipidemia claim between 2003 and 2006 and an absence of dyslipidemia in the health history prior to the year of diagnosis. Through this verification process, we excluded 3567 patients who did not have a first claim during 2003e2006. Besides, 2706 patients were excluded due to lack of medication compliance, while 370 patients of age <20 years were excluded. We intended to select true dyslipidemia patients among all the enrollees. Generally, diagnosis of dyslipidemia is made on the basis of clinical guidelines involving various laboratory tests, physical examinations, and history taking including cigarette smoking and family history of premature coronary heart disease [12,13]. However, we could not obtain this concrete information in claim data. Although the concordance rate of claim data to real medical record is high in Korea [14], whether selection of the target population using the main diagnosis code of International classification of Diseases (ICD)-10 codes in claim data is appropriate is still debatable. Thus, we only selected dyslipidemia patients with oral medications (Appendix 1) in order to increase the accuracy of ICD-10 code in the National Health Insurance claim data and to increase the homogeneity of the target population with risk adjustment. The final study sample was 11,946 patients with dyslipidemia (Fig. 1). We observed subjects for a minimum of 8 years from 2006 and a maximum of 11 years from 2003.
Data source
Covariates
This study used data from the Korean National Health Insurance (KNHI) claims database from 2002 to 2013 and the 2005 Korean Census. The National Health Insurance Corporation collects cohort data that are representative of the country’s population. These data include information from 1,025,340 subjects who represent a stratified random sample selected based on age, sex, region, health insurance type, income quintile, and individual total medical costs in 2002. The database includes information on reimbursement for each medical service including basic patient information, identifier for the clinic or hospital, disease code, costs incurred, results of health screening, individual/ family health history, health behaviors, and information related to cause of death. We conducted a cohort study of newly diagnosed dyslipidemia participants with
Age (20e49, 50e59, 60e69, or 70 years), sex, residential area (metropolitan, urban, or rural), Charlson comorbidity index (CCI; 0, 1, 2, or 3) [15], medical history of hypertension and diabetes, disability (normal, mild, and severe), and number of health screenings during the follow-up period (1, 2, 3, or 4) were determined in all patients. Only the comorbidity component of the CCI was calculated, and all diagnostic information was collected from inpatient and outpatient billing data within the diagnosis year. Individual SES We used average monthly insurance premium for household income as a proxy variable. In Korea, the type of
Combined effect of individual and neighborhood socioeconomic status on mortality in patients with newly diagnosed dyslipidemia
Figure 1
209
Flow chart for sampling the target population.
health insurance is classified as national health insurance or medical aid. People can qualify for medical aid if their single-family household income is <$600 per month; otherwise, they have national health insurance. People who have national health insurance based on employment pay a monthly insurance premium according to annual salary, and people who are self-employed pay for their premium based on the value of their property. People qualified for national health insurance were distributed from 1 to 100 percentile, and people who had medical aid were classified into the 0 percentile. We categorized individual household income into three groups (low, 0e20 percentile; middle, 21e80 percentile; and high, 81e100 percentile).
calculated at Si (city), Gun (county), and Gu (borough) level by merging these four basic indicators, similar to the calculation of the Carstairs index. Si, Gun, and Gu are Korean geographical units covering all small areas in Korea. We calculated z-scores at Si, Gun, and Gu levels using the mean and standard deviation of the four indicators. The z-score was calculated by subtracting the mean from the observed value for each indicator and divided by standard deviation; then, the four standardized z-scores were summed. “Less affluent” and “more affluent” neighborhoods were distinguished based on median neighborhood deprivation index.
Neighborhood SES
The outcome variable was survival time, starting from the date of diagnosis until either the date of death or the end of the study, and we defined mortality as death by any cause, as identified by death certificate data in the national death registry.
We used a modified Carstairs index [16] as a measurement for neighborhood SES using census data from 2005. When originally calculating the Carstairs index, four variables from census data were used based on previous studies: 1) residents in households headed by an unskilled worker, 2) unemployed males, 3) residents overcrowded, and 4) residents without a car. However, as we could not obtain data from the census with regard to whether or not residents owned cars, similar to the case in Lee’s study [17], we replaced “residents without car” with “residence not owner-occupied.” Neighborhood deprivation index was
Outcome variable
Statistical analysis Descriptive statistics were computed for all variables as frequencies and percentages for categorical variables using a Chi-squared test. Survival probability for all-cause mortality was estimated by the KaplaneMeier product limit method with log-rank tests to stratify SES. In order to
210
J. Shin et al.
investigate the association between combined individual and neighborhood SES and mortality, we performed survival analysis using a Cox proportional hazard model. However, with the structure of SES being hierarchical, we also considered using a frailty model that included random effects to deal with covariates of a hierarchy and denoted and tested variance and p-value for mortality among neighborhoods. The variance was 0.036 and p-value 0.19. Therefore, the frailty model was not considered ultimately. The proportional hazard assumptions were tested using the scaled Schoenfeld residuals, and no violation was
found. All statistical analyses were performed using the SAS 9.3 software. Results Demographic characteristics Of the 11,946 eligible subjects, 1053 (8.8%) died during the study period (Table 1). Statistical differences were observed between the two groups for all of the individual patient characteristics including age, sex, health insurance
Table 1 Baseline characteristics of newly diagnosed dyslipidemia patients. (N, (%))
Characteristics
Total
Alive
N Z 11,946
N Z 10,893
Age, n (%) 20e49 3614 3548 50e59 3476 3331 60e69 3206 2899 70 1650 1115 Sex, n (%) Male 5453 4921 Female 6493 5972 Health insurance type, n (%) National health insurance 11,874 10,836 Medical aid 72 57 Income, n (%) 20 percentile 1499 1347 21e80 percentile 6207 5663 81 percentile 4240 3883 Carstairs index, n (%) Less affluent 6670 6056 More affluent 5276 4837 Individual household income level: neighborhood deprivation, n (%) HigheMore affluent 2009 1849 Highe Less affluent 2231 2034 MiddleeMore affluent 2679 2457 MiddleeLess affluent 3528 3206 LoweMore affluent 588 531 LoweLess affluent 911 816 Residential area, n (%) Metropolitan 5833 5326 Urban 4925 4509 Rural 1188 1058 Charlson comorbidity index, n (%) 0e1 7248 6858 2 2593 2312 3 1189 1028 4 916 695 Diabetes, n (%) Yes 4675 4165 No 3873 3647 Hypertension, n (%) Yes 8073 7246 No 3873 3647 Disability, n (%) Normal 11,117 10,236 Mild disability 663 553 Severe disability 166 104 Number of health screenings during the follow-up period, n (%) 1 4809 4074 2 2041 1912 3 5096 4907
Dead (91.2)
N Z 1053
P-value (8.8) <0.001
(98.2) (95.8) (90.4) (67.6)
66 145 307 535
(1.8) (4.2) (9.6) (32.4)
(90.2) (92.0)
532 521
(9.8) (8.0)
0.0009
(91.3) (79.2)
1038 15
(8.7) (20.8)
0.0003
(89.9) (91.2) (91.6)
152 544 357
(10.1) (8.8) (8.4)
0.123
(90.8) (91.7)
614 439
(9.2) (8.3)
0.090
(92.0) (91.2) (91.7) (90.9) (90.3) (89.6)
160 197 222 322 57 95
(8.0) (8.8) (8.3) (9.1) (9.7) (10.4)
0.245
(91.3) (91.6) (89.1)
507 416 130
(8.7) (8.5) (10.9)
0.022
(94.6) (89.2) (86.5) (75.9)
390 281 161 221
(5.4) (10.8) (13.5) (24.1)
<0.001
(89.1) (94.2)
510 226
(10.9) (5.8)
<0.001
(89.8) (94.2)
827 226
(10.2) (5.8)
<0.001
(92.1) (83.4) (62.7)
881 110 62
(7.9) (16.6) (37.4)
<0.001
(84.7) (93.7) (96.3)
735 129 189
(15.3) (6.3) (3.7)
<0.001
Combined effect of individual and neighborhood socioeconomic status on mortality in patients with newly diagnosed dyslipidemia
type, residential area, CCI, major comorbidities, disability, and number of health screenings during the follow-up period. In terms of the combined variables of interest, between individual income and neighborhood SES, the reference group of high-income residents from more affluent neighborhoods had 2009 patients (16.8%). Among them, 160 patients (8.8%) died during the follow-up period. By contrast, 95 (10.4%) of 911 low-income patients in less affluent areas died during the study. More than three-quarters of the study subjects had diabetes or hypertension, and 29.6% had both. During the study, 11.9% of the dyslipidemia patients with both noncommunicable diseases died, while only 4.8% of patients with only dyslipidemia died.
Survival analysis KaplaneMeier analysis showed no distinctive feature regarding the cross-level effect between individual and neighborhood SES (Fig. 2; p-value < 0.014 by log-rank test). However, according to the Cox repression analysis for the separate effects of individual and neighborhood SES (Table 2), the adjusted hazard ratios (HRs) for middle and low SES were 1.34 (95% confidence interval (CI): 1.10e1.64 and 1.43 (1.25e1.64), respectively, compared to the high-
211
income reference group. Moreover, the adjusted HR for living in a less affluent neighborhood was 1.13 (95% CI: 1.00e1.29). Other demographic characteristics such as older age category, male sex, more severely disabled status, and higher CCI also had statistically significant higher adjusted HRs. In addition, dyslipidemia patients living in rural areas (HR Z 1.30, 95% CI: 1.07e1.59) had higher adjusted HRs compared to those living in metropolitan areas. By contrast, healthy behaviors such as participation in a national health examination either twice (HR Z 0.43, 95% CI: 0.36e0.52) or thrice (HR Z 0.27, 95% CI: 0.23e0.32) during the follow-up were associated with lower adjusted HR. In terms of the combined variables of interest between individual income and neighborhood deprivations, we set the high-income group from more affluent neighborhoods as the reference category (Table 3). Compared to the reference category, patients with high income living in less affluent neighborhoods had a higher mortality risk (HR Z 1.19, 95% CI: 0.96e1.47) regardless of statistical indifference. Among dyslipidemia patients with middle income, the adjusted HR in less affluent neighborhoods was higher than that of those living in more affluent neighborhoods (less affluent; HR Z 1.64, 95% CI: 1.35e1.99 vs. more affluent; HR Z 1.48, 95% CI: 1.20e1.81, respectively). We obtained consistent results in patients with low income, as the adjusted HR in less affluent neighborhoods was higher than that in more affluent neighborhoods (less
Figure 2 Survival probability for all-cause mortality stratified to individual household income according to less or more affluent neighborhoods.
212
J. Shin et al.
Table 2 Hazard ratios for all-cause mortality among newly diagnosed dyslipidemia patients. Unadjusted Characteristics
HR
Age (years) 20e49 1.00 50e59 2.34 60e69 5.46 70 21.40 Sex Male 1.24 Female 1.00 Health insurance type National health insurance 1.00 Medical aid 2.40 Individual household income 20 percentile 1.04 21e80 percentile 0.95 81 percentile 1.00 Carstairs index Less affluent 1.11 More affluent 1.00 Residential area Metropolitan 1.00 Urban 0.97 Rural 1.28 Diabetes Yes 1.45 No 1.00 Hypertension Yes 1.75 No 1.00 Charlson comorbidity indexa 0e1 1.00 2 2.08 3 2.65 4 5.06 Disability Normal 1.00 Mild disability 2.29 Severe disability 5.68 Health screenings during the follow-up period 1 1.00 2 0.40 3 0.23 a
Adjusted 95% CI
HR
95% CI
(1.75e3.12) (4.19e7.13) (16.57e27.63)
1.00 2.98 6.40 22.28
(2.22e4.00) (4.86e8.42) (17.41e29.90)
(1.10e1.40)
(1.44e4.00)
2.16 1.00 1.00 1.40
(1.91e2.45)
(0.82e2.39)
(0.82e1.32) (0.81e1.12)
1.43 1.34 1.00
(1.25e1.64) (1.10e1.64)
(0.99e1.26)
1.13 1.00
(1.00e1.29)
(0.85e1.11) (1.05e1.55)
1.00 1.03 1.30
(0.90e1.17) (1.07e1.59)
(1.28e1.63)
1.11 1.00
(0.98e1.25)
(1.51e2.02)
0.81 1.00
(0.69e0.94)
(1.78e2.42) (2.20e3.18) (4.29e5.97)
1.00 1.69 1.72 2.68
(1.45e1.97) (1.43e2.07) (2.26e3.18)
(1.87e2.79) (4.39e7.36)
1.00 1.50 2.39
(1.23e1.84) (1.84e3.11)
(0.33e0.48) (0.19e0.27)
1.00 0.43 0.27
(0.36e0.52) (0.23e0.31)
Calculated by extracting diabetes, hypertension, and hyperlipidemia among comorbidity components.
affluent; HR Z 1.52, 95% CI: 1.16e1.97 vs. more affluent; HR Z 1.41, 95% CI: 1.04e1.92, respectively). We also calculated the HRs for disease-specific mortality including cerebrovascular and ischemic heart disease; however, the observed number of deaths for these specific disease categories was too small to evaluate the HRs with an appropriate statistical significance (Supplementary Table 1).
Discussion This study examined the cross-level effects of individual and neighborhood SES on the risk of all-cause mortality in dyslipidemia patients in a population-based study using data from the national health insurance system in Korea. According to the results, dyslipidemia patients with low
individual income living in less affluent neighborhoods had significantly increased all-cause mortality rates after adjustments were made for a variety of comorbidities. Our findings are consistent with the results of another study that investigated the relationship between individual SES and prevalence of dyslipidemia [18]. Dyslipidemia is considered to be associated with cardio- and cerebrovascular diseases, [19e21] both of which are also influenced by individual and neighborhood SES. For example, according to a recent study in Japan, stroke mortality rate in less affluent neighborhoods is higher than that in more affluent neighborhoods [22]. According to another study in the U.S., a higher risk of incidence and mortality from ischemic stroke was observed in the most less affluent neighborhoods [23,24]. Moreover, metabolic syndrome, with dyslipidemia being one of the diagnostic
Combined effect of individual and neighborhood socioeconomic status on mortality in patients with newly diagnosed dyslipidemia
213
Table 3 HRs of mortality according to individual household income in less and more affluent neighborhoods. Less affluent neighborhoods All-cause mortality
Individual household income High (81 percentile) Middle (21e80 percentile) Low (20 percentile)
More affluent neighborhoods
No. of deaths (deaths per 1000 py)
HR
95% CI
No. of deaths (deaths per 1000 py)
HR
95% CI
197 322 95
1.19 1.64 1.52
(0.96e1.47) (1.35e1.99) (1.16e1.97)
160 222 57
1.00 1.48 1.41
(1.20e1.81) (1.04e1.92)
(10.6) (11.0) (12.5)
(9.5) (9.9) (11.7)
Less and more affluent neighborhoods were distinguished on the basis of the median for neighborhood Carstairs index, with less affluent neighborhoods having more-than-median Carstairs index; a higher Carstairs index represents a more deprived neighborhood.
criteria, is associated with a similar neighborhood effect. Chicholowska et al. showed that individual and neighborhood SES were both independently associated with an increased prevalence of metabolic syndrome using the 2005 ATP III criteria in the Atherosclerosis Risk in Communities (ARIC) study from 1987 to 1999 [25,26]. It is important to understand how residential communities influence mortality, particularly given the demonstrated mortality inequalities between deprived and affluent communities. The first possible reason for this is that the general living environment might influence or shape health exposures and outcomes through its infrastructure. Other reasons include the values placed on health and health-associated factors and the availability of health care. However, the provision of health insurance based on universal coverage of the country as a whole has improved accessibility to medical care, and geographical accessibility is better because of the small size of Korea and the improved transportation among regions. The setting of our study is characterized by universal healthcare coverage and more equal access to other public concerns such as primary education and social services, suggesting that financial barriers have been reduced and access to resources is a less-pronounced determinant of health. Despite reduced financial barriers, socioeconomic disparities for mortality still exist. One possible explanation for this is that more direct psychosocial pathways such as hopelessness, lack of control, or loss of respect arising as a consequence of inequality also affect individual health [27,28]. In addition, a lack of social cohesion or involvement, possibly linked to psychosocial issues, might contribute to the inferior health of the poor in less affluent areas due to the lack of social support from communities. The second possibility, related to the role of SES, suggests that the rich living in more affluent neighborhoods are healthier. The ability of the relatively wealthy to use their knowledge, money, power, prestige, and social connections would be reinforced by residence in more affluent neighborhoods [29]. The wealthy are more likely to embrace prevention and quickly take advantage of the treatment innovations [30]. In more affluent neighborhoods, such knowledge might be readily shared and cultivated among personal networks [31]. By contrast, the greater isolation of individuals with low SES might make it
more difficult for them to seek advice from relatives, friends, or acquaintances. The third hypothesis is that the lack of safe environments reduce the possibility for residents to exercise and increases the possibility of unhealthy and unbalanced nutrition, thereby exacerbating an unhealthy lifestyle [32]. According to a study in the U.S., childhood obesity is associated with neighborhood characteristics. In other words, children in less affluent neighborhoods had easy access to fast food and limited access to healthy food options and parks for engaging in physical activities. In addition, other sociocultural norms regarding unhealthy lifestyles including physical fitness and health promotion vary between neighborhoods and affect the health of the residents as well as their risk for mortality. Before drawing conclusions, the issue of external validity of this study has to be addressed. As we included only dyslipidemia patients with medication, our findings cannot be generalized to all dyslipidemia patients. In other words, some dyslipidemia patients treated with lifestyle modifications and dietary changes are not included in this analysis. These patients might not require medication because of their mild impaired laboratory tests or lack of other requirements for receiving medication. As these patients could constitute a large proportion of those with dyslipidemia in the general population, we cannot guarantee the external validity of the results despite having obtained internal validity using the inclusion criteria of oral medication to overcome the limited information in the claim data. Thus, we suggest that further investigation is warranted to increase the external validity by including all dyslipidemia patients regardless of whether they are treated with medication. This study also has several limitations. First, as we used data from the claim database, we could not consider lifestyle factors and education that also influence mortality. Second, we did not consider changing the neighborhood deprivation status of participants who moved from one residence to another within the study area. Third, the number of deaths considered with regard to diseasespecific mortality was not adequate to evaluate the statistical significance. However, our study also has several advantages. First, to our knowledge, this study was the first to investigate the
214
relationships between individual and neighborhood SES and mortality among dyslipidemia patients with a relatively large sample; this in turn yielded good statistical significance, thereby enabling detection of the effects of neighborhood SES throughout our survival analysis. Second, we analyzed a representative sample of patients with dyslipidemia using the nationwide representative cohort data. Third, we made an effort to increase the homogeneity of our study sample using only data from patients newly diagnosed with dyslipidemia receiving medication. Conclusions Living in a less affluent neighborhood contributes to allcause mortality among dyslipidemia patients. The individual- and neighborhood-level variables cumulatively affect individuals such that the most at-risk individuals are those having both individual- and neighborhood-level risk factors. These findings raise important clinical and public health concerns and indicate that both individual and neighborhood SES should essentially be considered in health-care policies. Source of fundings No funding was received for this study. Furthermore, none of the researchers are associated with privately funded research institutes. Competing interest There are no conflicts of interest to declare. Details of contributors All authors contributed to this study in its entirety and approved this manuscript, agreeing to submit this final version to the Nutrition, Metabolism, and Cardiovascular Diseases. All of them have contributed to writing and analysis of this article. Ethics committee approval We obtained the IRB approval with the data from Institutional Review Board of Graduate School of Public Health in Yonsei University for using the data gathered and the study design (IRB approval number: 2-1040939-AB-N-012014-239). Disclosure None. Acknowledgments Editage has checked and revised English writing in the manuscript.
J. Shin et al.
Appendix A. Supplementary data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.numecd.2015.12.007. Appendix 1. Medication list for dyslipidemia based on the generic names.
Simvastatin Clopidogrel Atorvastatin(calcium) Nicotinic acid Pitavastatin calcium Rosuvastatin calcium Rosuvastatin calcium(as rosuvastatin 5 mg) Rosuvastatin calcium(as rosuvastatin 20 mg) Pravastatin sodium Micronized fenofibrate Fluvastatin Ezetimibe Acipimox Probucol Fenofibrate Ethyl linoleate Clopidogrel(besylate) Lovastatin Cerivastatin sodium
References [1] Kim KJ, Choi SI, Lee MS, Kim JA, Chun EJ, Jeon CH. The prevalence and characteristics of coronary atherosclerosis in asymptomatic subjects classified as low risk based on traditional risk stratification algorithm: assessment with coronary CT angiography. Heart 2013;99(15):1113e7. [2] Lee YH, Lee SG, Lee MH, Kim JH, Lee BW, Kang ES, et al. Serum cholesterol concentration and prevalence, awareness, treatment, and control of high low-density lipoprotein cholesterol in the Korea National Health and Nutrition Examination Surveys 2008e2010: beyond the tip of the Iceberg. J Am Heart Assoc 2014; 3(1):446. [3] Office KNS. Korean national health and nutritional examination survey. Available from:. 2015. http://kosis.kr/statisticsList/ statisticsList_01List.jsp?vwcdZMT_ZTITLE&parentIdZD#SubCont [retrieved on 11 June 2015]. [4] Kershaw KN, Osypuk TL, Do DP, De Chavez PJ, Diez Roux AV. Neighborhood-level racial/ethnic residential segregation and incident cardiovascular disease: the multi-ethnic study of atherosclerosis. Circulation 2015;131(2):141e8. [5] Robinson JG, Farnier M, Krempf M, Bergeron J, Luc G, Averna M, et al. Efficacy and safety of alirocumab in reducing lipids and cardiovascular events. N Engl J Med 2015;372(16):1489e99. [6] Sabatine MS, Giugliano RP, Wiviott SD, Raal FJ, Blom DJ, Robinson J, et al. Efficacy and safety of evolocumab in reducing lipids and cardiovascular events. N Engl J Med 2015;372(16):1500e9. [7] FORCE USPST. Lipid disorders in adults (Cholesterol, dyslipidemia). Screening Available from:. 2008. http://www.uspreventiveservice staskforce.org/Page/Document/RecommendationStatementFinal/ lipid-disorders-in-adults-cholesterol-dyslipidemia-screening [retrieved on 11 June 2015]. [8] Strandberg TE, Kolehmainen L, Vuorio A. Evaluation and treatment of older patients with hypercholesterolemia: a clinical review. Jama 2014;312(11):1136e44. [9] Casavalle PL, Lifshitz F, Romano LS, Pandolfo M, Caamano A, Boyer PM, et al. Prevalence of dyslipidemia and metabolic syndrome risk factor in overweight and obese children. Pediatr Endocrinol Rev 2014;12(2):213e23.
Combined effect of individual and neighborhood socioeconomic status on mortality in patients with newly diagnosed dyslipidemia [10] Kinra S, Bowen LJ, Lyngdoh T, Prabhakaran D, Reddy KS, Ramakrishnan L, et al. Sociodemographic patterning of noncommunicable disease risk factors in rural India: a cross sectional study. BMJ 2010;341. c4974. [11] Rodriguez CJ, Daviglus ML, Swett K, Gonzalez HM, Gallo LC, Wassertheil-Smoller S, et al. Dyslipidemia patterns among Hispanics/Latinos of diverse background in the United States. Am J Med 2014;127(12):1186e94. e1. [12] Pursnani A, Massaro J, D’Agostino R, O’Donnell C, Hoffmann U. Guideline-based statin allocation to distinguish both cardiovascular events and coronary artery calcification in the community: a comparison of the 2004 NCEP ATP III and 2013 ACC/AHA cholesterol treatment guideline. J Am Coll Cardiol 2015;65 (10_S). [13] Kavousi M, Leening MJ, Nanchen D, Greenland P, Graham IM, Steyerberg EW, et al. Comparison of application of the ACC/AHA guidelines, adult treatment panel III guidelines, and European society of cardiology guidelines for cardiovascular disease prevention in a European cohort. Jama 2014;311(14):1416e23. [14] Kim J. Strategies to enhance the use of National Health Insurance claims database in generating health statistics. Technical report by Health Insurance Review and Assessment Services 2005. in Korean. [15] Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40(5):373e83. [16] Morgan O, Baker A. Measuring deprivation in England and Wales using 2001 Carstairs scores. Health Stat Q 2006;31:28e33. [17] Gyu Lee Sang. The effect of neighborhood socioeconomic factors on spatial mortality and individual health status. PhD diss. Yonsei University; 2002. [18] Kapral MK, Wang H, Mamdani M, Tu JV. Effect of socioeconomic status on treatment and mortality after stroke. Stroke 2002;33(1): 268e73. [19] Kumral E, Evyapan D, Gokcay F, Karaman B, Orman M. Association of baseline dyslipidemia with stroke recurrence within five-years after ischemic stroke. Int J Stroke 2014;9(Suppl A100):119e26. [20] Cannon CP. Mixed dyslipidemia, metabolic syndrome, diabetes mellitus, and cardiovascular disease: clinical implications. Am J Cardiol 2008;102(12a):5le9l. [21] Madssen E, Laugsand LE, Wiseth R, Morkedal B, Platou C, Vatten L, et al. Risk of acute myocardial infarction: dyslipidemia more
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
215
detrimental for men than women. Epidemiology 2013;24(5): 637e42. Honjo K, Iso H, Nakaya T, Hanibuchi T, Ikeda A, Inoue M, et al. Impact of neighborhood socioeconomic conditions on the risk of stroke in Japan. J Epidemiol 2015;25(3):254e60. Brown AF, Liang LJ, Vassar SD, Stein-Merkin S, Longstreth Jr WT, Ovbiagele B, et al. Neighborhood disadvantage and ischemic stroke: the Cardiovascular Health Study (CHS). Stroke 2011; 42(12):3363e8. Brown AF, Liang LJ, Vassar SD, Merkin SS, Longstreth Jr WT, Ovbiagele B, et al. Neighborhood socioeconomic disadvantage and mortality after stroke. Neurology 2013;80(6):520e7. Chichlowska KL, Rose KM, Diez-Roux AV, Golden SH, McNeill AM, Heiss G. Individual and neighborhood socioeconomic status characteristics and prevalence of metabolic syndrome: the Atherosclerosis Risk in Communities (ARIC) Study. Psychosom Med 2008;70(9):986e92. Chichlowska KL, Rose KM, Diez-Roux AV, Golden SH, McNeill AM, Heiss G. Life course socioeconomic conditions and metabolic syndrome in adults: the Atherosclerosis Risk in Communities (ARIC) Study. Ann Epidemiol 2009;19(12):875e83. Boyce WT, Chesterman EA, Winkleby MA. Psychosocial predictors of maternal and infant health among adolescent mothers. Am J Dis Child 1991;145(3):267e73. Williams DR, Yu Yan, Jackson JS, Anderson NB. Racial differences in physical and mental health: socio-economic status, stress and discrimination. J Health Psychol 1997;2(3):335e51. Link BG, Northridge ME, Phelan JC, Ganz ML. Social epidemiology and the fundamental cause concept: on the structuring of effective cancer screens by socioeconomic status. Milbank Q 1998;76(3): 375e402. Roos LL, Traverse D, Turner D. Delivering prevention: the role of public programs in delivering care to high-risk populations. Med Care 1999;37(6). JS264-78. Roos LL, Magoon J, Gupta S, Chateau D, Veugelers PJ. Socioeconomic determinants of mortality in two Canadian provinces: multilevel modelling and neighborhood context. Soc Sci Med 2004;59(7):1435e47. Li X, Sundquist J, Calling S, Zöller B, Sundquist K. Neighborhood deprivation and risk of cervical cancer morbidity and mortality: a multilevel analysis from Sweden. Gynecol Oncol 2012;127(2):283e9.
Available online at www.sciencedirect.com
Nutrition, Metabolism & Cardiovascular Diseases journal homepage: www.elsevier.com/locate/nmcd
Combined effect of individual and neighborhood socioeconomic status on mortality in patients with newly diagnosed dyslipidemia: A nationwide Korean cohort study from 2002 to 2013 J. Shin b,c,1, K.H. Cho a,b,1, Y. Choi a,b, S.G. Lee b,d, E.-C. Park b,c, S.-I. Jang b,c,* a
Department of Public Health, Graduate School, Yonsei University, South Korea Institute of Health Services Research, College of Medicine, Yonsei University, South Korea c Department of Preventive Medicine, College of Medicine, Yonsei University, South Korea d Department of Hospital Management, Graduate School of Public Health, Yonsei University, South Korea b
Received 3 July 2015; received in revised form 9 November 2015; accepted 14 December 2015 Available online 23 December 2015
KEYWORDS Socioeconomic status; Dyslipidemia; Hypercholesterolemia; Hypertriglyceridemia; Neighborhood; Mortality
Abstract Background and Aim: The study aims to determine whether dyslipidemia patients living in less affluent neighborhood are at a higher risk of mortality compared to those living in more affluent neighborhoods. Methods and results: A population-based cohort study was conducted using a stratified representative sampling from the National Health Insurance claim data from 2002 to 2013. The target subjects comprise patients newly diagnosed with dyslipidemia receiving medication. We performed a survival analysis using the Cox proportional hazard model. Of 11,946 patients with dyslipidemia, 1053 (8.8%) subjects died during the follow-up period. Of the dyslipidemia patients earning a middle-class income, the adjusted HR in less affluent neighborhoods was higher than that in the more affluent neighborhoods compared to the reference category of high individual SES in more affluent neighborhoods (less affluent; hazard ratio (HR) Z 1.64, 95% confidence interval (CI): 1.35e1.99 vs. more affluent; HR Z 1.48, 95% CI: 1.20e1.81, respectively). We obtained consistent results in patients with lower income, wherein the adjusted HR in less affluent neighborhoods was higher than that in more affluent neighborhoods (less affluent; HR Z 1.52, 95% CI: 1.16e1.97 vs. more affluent; HR Z 1.41, 95% CI: 1.04e1.92, respectively). Conclusion: Living in a less affluent neighborhood contributes to higher mortality among dyslipidemia patients. The individual- and neighborhood-level variables cumulatively affect individuals such that the most at-risk individuals include those having both individual- and neighborhood-level risk factors. These findings raise important clinical and public health concerns and indicate that neighborhood SES approaches should be essentially considered in health-care policies similar to individual SES. ª 2015 The Italian Society of Diabetology, the Italian Society for the Study of Atherosclerosis, the Italian Society of Human Nutrition, and the Department of Clinical Medicine and Surgery, Federico II University. Published by Elsevier B.V. All rights reserved.
* Corresponding author. Department of Preventive Medicine and Institute of Health Services Research, Yonsei University College of Medicine, 50, Yonsei-ro, Seodaemun-gu, Seoul, 120-752, South Korea. Tel.: þ82 2 2228 1863; fax: þ82 2 392 8133. E-mail address: [email protected] (S.-I. Jang). 1 J. Shin and K.H. Cho contributed equally to the manuscript. http://dx.doi.org/10.1016/j.numecd.2015.12.007 0939-4753/ª 2015 The Italian Society of Diabetology, the Italian Society for the Study of Atherosclerosis, the Italian Society of Human Nutrition, and the Department of Clinical Medicine and Surgery, Federico II University. Published by Elsevier B.V. All rights reserved.
208
Introduction Dyslipidemia is a common noncommunicable disease in Korea; in the Korean society, the incidence and prevalence of the disease increase with increasing age [1,2]. According to the Korean National Health and Nutritional Examination Survey (KHANES) in 2013, the prevalence of hypercholesterolemia and hypertriglyceridemia was 14.9% and 17.1%, respectively, among the respondents of age >30 years [3]. Upon comparing the current statistics of prevalence to those reported earlier (KHANES, 1998: 10.0% of subjects had hypercholesterolemia and 10.2% hypertriglyceridemia), we found the need for prevention of this continuous steep increase in the number of patients with dyslipidemia. Dyslipidemia can result in other noncommunicable diseases such as metabolic syndrome, cardiovascular diseases, and stroke [4e6]. Dyslipidemia might impose a large burden on human health in a developed society; however, it can be prevented and controlled by early screening and treatment with appropriate dietary strategies at both individual and population levels. According to the U.S. Preventive Services Task Force in 2008, several guidelines have been suggested for lipid disorder in adults [7]. However, whether the screening for lipid disorder is cost-effective is still debatable. In addition, the risk factors for dyslipidemia are well known [8,9]. Among them, individual socioeconomic status (SES) and neighborhood SES are associated with prevalence [10,11]. Many other diseases associated with dyslipidemia such as cardiovascular disease, stroke, and metabolic syndrome have been investigated with regard to the cross-level effects between individual and neighborhood SES. However, few studies have related these factors to dyslipidemia especially in Asians. Therefore, in this study, we first examined the effects of individual and neighborhood SES on the overall mortality rate from 2003. Then, we further investigated the crosslevel effects between individual and neighborhood SES.
J. Shin et al.
medication to investigate the association between combined individual and neighborhood SES and mortality. Ethical approval for this study was granted by the Institutional Review Board of the Graduate School of Public Health at Yonsei University. The requirement for informed consent was waived because the study was based on routinely collected administrative and claim data. Study sample selection
Methods
Among the enrolled subjects, 18,219 with newly diagnosed dyslipidemia (E78; International Classification of Disease (ICD), 10th edition) between 2003 and 2006 were selected. We confirmed that diagnoses were new by verifying a lack of dyslipidemia claims between 2002 and 2005, followed by an initial dyslipidemia claim between 2003 and 2006 and an absence of dyslipidemia in the health history prior to the year of diagnosis. Through this verification process, we excluded 3567 patients who did not have a first claim during 2003e2006. Besides, 2706 patients were excluded due to lack of medication compliance, while 370 patients of age <20 years were excluded. We intended to select true dyslipidemia patients among all the enrollees. Generally, diagnosis of dyslipidemia is made on the basis of clinical guidelines involving various laboratory tests, physical examinations, and history taking including cigarette smoking and family history of premature coronary heart disease [12,13]. However, we could not obtain this concrete information in claim data. Although the concordance rate of claim data to real medical record is high in Korea [14], whether selection of the target population using the main diagnosis code of International classification of Diseases (ICD)-10 codes in claim data is appropriate is still debatable. Thus, we only selected dyslipidemia patients with oral medications (Appendix 1) in order to increase the accuracy of ICD-10 code in the National Health Insurance claim data and to increase the homogeneity of the target population with risk adjustment. The final study sample was 11,946 patients with dyslipidemia (Fig. 1). We observed subjects for a minimum of 8 years from 2006 and a maximum of 11 years from 2003.
Data source
Covariates
This study used data from the Korean National Health Insurance (KNHI) claims database from 2002 to 2013 and the 2005 Korean Census. The National Health Insurance Corporation collects cohort data that are representative of the country’s population. These data include information from 1,025,340 subjects who represent a stratified random sample selected based on age, sex, region, health insurance type, income quintile, and individual total medical costs in 2002. The database includes information on reimbursement for each medical service including basic patient information, identifier for the clinic or hospital, disease code, costs incurred, results of health screening, individual/ family health history, health behaviors, and information related to cause of death. We conducted a cohort study of newly diagnosed dyslipidemia participants with
Age (20e49, 50e59, 60e69, or 70 years), sex, residential area (metropolitan, urban, or rural), Charlson comorbidity index (CCI; 0, 1, 2, or 3) [15], medical history of hypertension and diabetes, disability (normal, mild, and severe), and number of health screenings during the follow-up period (1, 2, 3, or 4) were determined in all patients. Only the comorbidity component of the CCI was calculated, and all diagnostic information was collected from inpatient and outpatient billing data within the diagnosis year. Individual SES We used average monthly insurance premium for household income as a proxy variable. In Korea, the type of
Combined effect of individual and neighborhood socioeconomic status on mortality in patients with newly diagnosed dyslipidemia
Figure 1
209
Flow chart for sampling the target population.
health insurance is classified as national health insurance or medical aid. People can qualify for medical aid if their single-family household income is <$600 per month; otherwise, they have national health insurance. People who have national health insurance based on employment pay a monthly insurance premium according to annual salary, and people who are self-employed pay for their premium based on the value of their property. People qualified for national health insurance were distributed from 1 to 100 percentile, and people who had medical aid were classified into the 0 percentile. We categorized individual household income into three groups (low, 0e20 percentile; middle, 21e80 percentile; and high, 81e100 percentile).
calculated at Si (city), Gun (county), and Gu (borough) level by merging these four basic indicators, similar to the calculation of the Carstairs index. Si, Gun, and Gu are Korean geographical units covering all small areas in Korea. We calculated z-scores at Si, Gun, and Gu levels using the mean and standard deviation of the four indicators. The z-score was calculated by subtracting the mean from the observed value for each indicator and divided by standard deviation; then, the four standardized z-scores were summed. “Less affluent” and “more affluent” neighborhoods were distinguished based on median neighborhood deprivation index.
Neighborhood SES
The outcome variable was survival time, starting from the date of diagnosis until either the date of death or the end of the study, and we defined mortality as death by any cause, as identified by death certificate data in the national death registry.
We used a modified Carstairs index [16] as a measurement for neighborhood SES using census data from 2005. When originally calculating the Carstairs index, four variables from census data were used based on previous studies: 1) residents in households headed by an unskilled worker, 2) unemployed males, 3) residents overcrowded, and 4) residents without a car. However, as we could not obtain data from the census with regard to whether or not residents owned cars, similar to the case in Lee’s study [17], we replaced “residents without car” with “residence not owner-occupied.” Neighborhood deprivation index was
Outcome variable
Statistical analysis Descriptive statistics were computed for all variables as frequencies and percentages for categorical variables using a Chi-squared test. Survival probability for all-cause mortality was estimated by the KaplaneMeier product limit method with log-rank tests to stratify SES. In order to
210
J. Shin et al.
investigate the association between combined individual and neighborhood SES and mortality, we performed survival analysis using a Cox proportional hazard model. However, with the structure of SES being hierarchical, we also considered using a frailty model that included random effects to deal with covariates of a hierarchy and denoted and tested variance and p-value for mortality among neighborhoods. The variance was 0.036 and p-value 0.19. Therefore, the frailty model was not considered ultimately. The proportional hazard assumptions were tested using the scaled Schoenfeld residuals, and no violation was
found. All statistical analyses were performed using the SAS 9.3 software. Results Demographic characteristics Of the 11,946 eligible subjects, 1053 (8.8%) died during the study period (Table 1). Statistical differences were observed between the two groups for all of the individual patient characteristics including age, sex, health insurance
Table 1 Baseline characteristics of newly diagnosed dyslipidemia patients. (N, (%))
Characteristics
Total
Alive
N Z 11,946
N Z 10,893
Age, n (%) 20e49 3614 3548 50e59 3476 3331 60e69 3206 2899 70 1650 1115 Sex, n (%) Male 5453 4921 Female 6493 5972 Health insurance type, n (%) National health insurance 11,874 10,836 Medical aid 72 57 Income, n (%) 20 percentile 1499 1347 21e80 percentile 6207 5663 81 percentile 4240 3883 Carstairs index, n (%) Less affluent 6670 6056 More affluent 5276 4837 Individual household income level: neighborhood deprivation, n (%) HigheMore affluent 2009 1849 Highe Less affluent 2231 2034 MiddleeMore affluent 2679 2457 MiddleeLess affluent 3528 3206 LoweMore affluent 588 531 LoweLess affluent 911 816 Residential area, n (%) Metropolitan 5833 5326 Urban 4925 4509 Rural 1188 1058 Charlson comorbidity index, n (%) 0e1 7248 6858 2 2593 2312 3 1189 1028 4 916 695 Diabetes, n (%) Yes 4675 4165 No 3873 3647 Hypertension, n (%) Yes 8073 7246 No 3873 3647 Disability, n (%) Normal 11,117 10,236 Mild disability 663 553 Severe disability 166 104 Number of health screenings during the follow-up period, n (%) 1 4809 4074 2 2041 1912 3 5096 4907
Dead (91.2)
N Z 1053
P-value (8.8) <0.001
(98.2) (95.8) (90.4) (67.6)
66 145 307 535
(1.8) (4.2) (9.6) (32.4)
(90.2) (92.0)
532 521
(9.8) (8.0)
0.0009
(91.3) (79.2)
1038 15
(8.7) (20.8)
0.0003
(89.9) (91.2) (91.6)
152 544 357
(10.1) (8.8) (8.4)
0.123
(90.8) (91.7)
614 439
(9.2) (8.3)
0.090
(92.0) (91.2) (91.7) (90.9) (90.3) (89.6)
160 197 222 322 57 95
(8.0) (8.8) (8.3) (9.1) (9.7) (10.4)
0.245
(91.3) (91.6) (89.1)
507 416 130
(8.7) (8.5) (10.9)
0.022
(94.6) (89.2) (86.5) (75.9)
390 281 161 221
(5.4) (10.8) (13.5) (24.1)
<0.001
(89.1) (94.2)
510 226
(10.9) (5.8)
<0.001
(89.8) (94.2)
827 226
(10.2) (5.8)
<0.001
(92.1) (83.4) (62.7)
881 110 62
(7.9) (16.6) (37.4)
<0.001
(84.7) (93.7) (96.3)
735 129 189
(15.3) (6.3) (3.7)
<0.001
Combined effect of individual and neighborhood socioeconomic status on mortality in patients with newly diagnosed dyslipidemia
type, residential area, CCI, major comorbidities, disability, and number of health screenings during the follow-up period. In terms of the combined variables of interest, between individual income and neighborhood SES, the reference group of high-income residents from more affluent neighborhoods had 2009 patients (16.8%). Among them, 160 patients (8.8%) died during the follow-up period. By contrast, 95 (10.4%) of 911 low-income patients in less affluent areas died during the study. More than three-quarters of the study subjects had diabetes or hypertension, and 29.6% had both. During the study, 11.9% of the dyslipidemia patients with both noncommunicable diseases died, while only 4.8% of patients with only dyslipidemia died.
Survival analysis KaplaneMeier analysis showed no distinctive feature regarding the cross-level effect between individual and neighborhood SES (Fig. 2; p-value < 0.014 by log-rank test). However, according to the Cox repression analysis for the separate effects of individual and neighborhood SES (Table 2), the adjusted hazard ratios (HRs) for middle and low SES were 1.34 (95% confidence interval (CI): 1.10e1.64 and 1.43 (1.25e1.64), respectively, compared to the high-
211
income reference group. Moreover, the adjusted HR for living in a less affluent neighborhood was 1.13 (95% CI: 1.00e1.29). Other demographic characteristics such as older age category, male sex, more severely disabled status, and higher CCI also had statistically significant higher adjusted HRs. In addition, dyslipidemia patients living in rural areas (HR Z 1.30, 95% CI: 1.07e1.59) had higher adjusted HRs compared to those living in metropolitan areas. By contrast, healthy behaviors such as participation in a national health examination either twice (HR Z 0.43, 95% CI: 0.36e0.52) or thrice (HR Z 0.27, 95% CI: 0.23e0.32) during the follow-up were associated with lower adjusted HR. In terms of the combined variables of interest between individual income and neighborhood deprivations, we set the high-income group from more affluent neighborhoods as the reference category (Table 3). Compared to the reference category, patients with high income living in less affluent neighborhoods had a higher mortality risk (HR Z 1.19, 95% CI: 0.96e1.47) regardless of statistical indifference. Among dyslipidemia patients with middle income, the adjusted HR in less affluent neighborhoods was higher than that of those living in more affluent neighborhoods (less affluent; HR Z 1.64, 95% CI: 1.35e1.99 vs. more affluent; HR Z 1.48, 95% CI: 1.20e1.81, respectively). We obtained consistent results in patients with low income, as the adjusted HR in less affluent neighborhoods was higher than that in more affluent neighborhoods (less
Figure 2 Survival probability for all-cause mortality stratified to individual household income according to less or more affluent neighborhoods.
212
J. Shin et al.
Table 2 Hazard ratios for all-cause mortality among newly diagnosed dyslipidemia patients. Unadjusted Characteristics
HR
Age (years) 20e49 1.00 50e59 2.34 60e69 5.46 70 21.40 Sex Male 1.24 Female 1.00 Health insurance type National health insurance 1.00 Medical aid 2.40 Individual household income 20 percentile 1.04 21e80 percentile 0.95 81 percentile 1.00 Carstairs index Less affluent 1.11 More affluent 1.00 Residential area Metropolitan 1.00 Urban 0.97 Rural 1.28 Diabetes Yes 1.45 No 1.00 Hypertension Yes 1.75 No 1.00 Charlson comorbidity indexa 0e1 1.00 2 2.08 3 2.65 4 5.06 Disability Normal 1.00 Mild disability 2.29 Severe disability 5.68 Health screenings during the follow-up period 1 1.00 2 0.40 3 0.23 a
Adjusted 95% CI
HR
95% CI
(1.75e3.12) (4.19e7.13) (16.57e27.63)
1.00 2.98 6.40 22.28
(2.22e4.00) (4.86e8.42) (17.41e29.90)
(1.10e1.40)
(1.44e4.00)
2.16 1.00 1.00 1.40
(1.91e2.45)
(0.82e2.39)
(0.82e1.32) (0.81e1.12)
1.43 1.34 1.00
(1.25e1.64) (1.10e1.64)
(0.99e1.26)
1.13 1.00
(1.00e1.29)
(0.85e1.11) (1.05e1.55)
1.00 1.03 1.30
(0.90e1.17) (1.07e1.59)
(1.28e1.63)
1.11 1.00
(0.98e1.25)
(1.51e2.02)
0.81 1.00
(0.69e0.94)
(1.78e2.42) (2.20e3.18) (4.29e5.97)
1.00 1.69 1.72 2.68
(1.45e1.97) (1.43e2.07) (2.26e3.18)
(1.87e2.79) (4.39e7.36)
1.00 1.50 2.39
(1.23e1.84) (1.84e3.11)
(0.33e0.48) (0.19e0.27)
1.00 0.43 0.27
(0.36e0.52) (0.23e0.31)
Calculated by extracting diabetes, hypertension, and hyperlipidemia among comorbidity components.
affluent; HR Z 1.52, 95% CI: 1.16e1.97 vs. more affluent; HR Z 1.41, 95% CI: 1.04e1.92, respectively). We also calculated the HRs for disease-specific mortality including cerebrovascular and ischemic heart disease; however, the observed number of deaths for these specific disease categories was too small to evaluate the HRs with an appropriate statistical significance (Supplementary Table 1).
Discussion This study examined the cross-level effects of individual and neighborhood SES on the risk of all-cause mortality in dyslipidemia patients in a population-based study using data from the national health insurance system in Korea. According to the results, dyslipidemia patients with low
individual income living in less affluent neighborhoods had significantly increased all-cause mortality rates after adjustments were made for a variety of comorbidities. Our findings are consistent with the results of another study that investigated the relationship between individual SES and prevalence of dyslipidemia [18]. Dyslipidemia is considered to be associated with cardio- and cerebrovascular diseases, [19e21] both of which are also influenced by individual and neighborhood SES. For example, according to a recent study in Japan, stroke mortality rate in less affluent neighborhoods is higher than that in more affluent neighborhoods [22]. According to another study in the U.S., a higher risk of incidence and mortality from ischemic stroke was observed in the most less affluent neighborhoods [23,24]. Moreover, metabolic syndrome, with dyslipidemia being one of the diagnostic
Combined effect of individual and neighborhood socioeconomic status on mortality in patients with newly diagnosed dyslipidemia
213
Table 3 HRs of mortality according to individual household income in less and more affluent neighborhoods. Less affluent neighborhoods All-cause mortality
Individual household income High (81 percentile) Middle (21e80 percentile) Low (20 percentile)
More affluent neighborhoods
No. of deaths (deaths per 1000 py)
HR
95% CI
No. of deaths (deaths per 1000 py)
HR
95% CI
197 322 95
1.19 1.64 1.52
(0.96e1.47) (1.35e1.99) (1.16e1.97)
160 222 57
1.00 1.48 1.41
(1.20e1.81) (1.04e1.92)
(10.6) (11.0) (12.5)
(9.5) (9.9) (11.7)
Less and more affluent neighborhoods were distinguished on the basis of the median for neighborhood Carstairs index, with less affluent neighborhoods having more-than-median Carstairs index; a higher Carstairs index represents a more deprived neighborhood.
criteria, is associated with a similar neighborhood effect. Chicholowska et al. showed that individual and neighborhood SES were both independently associated with an increased prevalence of metabolic syndrome using the 2005 ATP III criteria in the Atherosclerosis Risk in Communities (ARIC) study from 1987 to 1999 [25,26]. It is important to understand how residential communities influence mortality, particularly given the demonstrated mortality inequalities between deprived and affluent communities. The first possible reason for this is that the general living environment might influence or shape health exposures and outcomes through its infrastructure. Other reasons include the values placed on health and health-associated factors and the availability of health care. However, the provision of health insurance based on universal coverage of the country as a whole has improved accessibility to medical care, and geographical accessibility is better because of the small size of Korea and the improved transportation among regions. The setting of our study is characterized by universal healthcare coverage and more equal access to other public concerns such as primary education and social services, suggesting that financial barriers have been reduced and access to resources is a less-pronounced determinant of health. Despite reduced financial barriers, socioeconomic disparities for mortality still exist. One possible explanation for this is that more direct psychosocial pathways such as hopelessness, lack of control, or loss of respect arising as a consequence of inequality also affect individual health [27,28]. In addition, a lack of social cohesion or involvement, possibly linked to psychosocial issues, might contribute to the inferior health of the poor in less affluent areas due to the lack of social support from communities. The second possibility, related to the role of SES, suggests that the rich living in more affluent neighborhoods are healthier. The ability of the relatively wealthy to use their knowledge, money, power, prestige, and social connections would be reinforced by residence in more affluent neighborhoods [29]. The wealthy are more likely to embrace prevention and quickly take advantage of the treatment innovations [30]. In more affluent neighborhoods, such knowledge might be readily shared and cultivated among personal networks [31]. By contrast, the greater isolation of individuals with low SES might make it
more difficult for them to seek advice from relatives, friends, or acquaintances. The third hypothesis is that the lack of safe environments reduce the possibility for residents to exercise and increases the possibility of unhealthy and unbalanced nutrition, thereby exacerbating an unhealthy lifestyle [32]. According to a study in the U.S., childhood obesity is associated with neighborhood characteristics. In other words, children in less affluent neighborhoods had easy access to fast food and limited access to healthy food options and parks for engaging in physical activities. In addition, other sociocultural norms regarding unhealthy lifestyles including physical fitness and health promotion vary between neighborhoods and affect the health of the residents as well as their risk for mortality. Before drawing conclusions, the issue of external validity of this study has to be addressed. As we included only dyslipidemia patients with medication, our findings cannot be generalized to all dyslipidemia patients. In other words, some dyslipidemia patients treated with lifestyle modifications and dietary changes are not included in this analysis. These patients might not require medication because of their mild impaired laboratory tests or lack of other requirements for receiving medication. As these patients could constitute a large proportion of those with dyslipidemia in the general population, we cannot guarantee the external validity of the results despite having obtained internal validity using the inclusion criteria of oral medication to overcome the limited information in the claim data. Thus, we suggest that further investigation is warranted to increase the external validity by including all dyslipidemia patients regardless of whether they are treated with medication. This study also has several limitations. First, as we used data from the claim database, we could not consider lifestyle factors and education that also influence mortality. Second, we did not consider changing the neighborhood deprivation status of participants who moved from one residence to another within the study area. Third, the number of deaths considered with regard to diseasespecific mortality was not adequate to evaluate the statistical significance. However, our study also has several advantages. First, to our knowledge, this study was the first to investigate the
214
relationships between individual and neighborhood SES and mortality among dyslipidemia patients with a relatively large sample; this in turn yielded good statistical significance, thereby enabling detection of the effects of neighborhood SES throughout our survival analysis. Second, we analyzed a representative sample of patients with dyslipidemia using the nationwide representative cohort data. Third, we made an effort to increase the homogeneity of our study sample using only data from patients newly diagnosed with dyslipidemia receiving medication. Conclusions Living in a less affluent neighborhood contributes to allcause mortality among dyslipidemia patients. The individual- and neighborhood-level variables cumulatively affect individuals such that the most at-risk individuals are those having both individual- and neighborhood-level risk factors. These findings raise important clinical and public health concerns and indicate that both individual and neighborhood SES should essentially be considered in health-care policies. Source of fundings No funding was received for this study. Furthermore, none of the researchers are associated with privately funded research institutes. Competing interest There are no conflicts of interest to declare. Details of contributors All authors contributed to this study in its entirety and approved this manuscript, agreeing to submit this final version to the Nutrition, Metabolism, and Cardiovascular Diseases. All of them have contributed to writing and analysis of this article. Ethics committee approval We obtained the IRB approval with the data from Institutional Review Board of Graduate School of Public Health in Yonsei University for using the data gathered and the study design (IRB approval number: 2-1040939-AB-N-012014-239). Disclosure None. Acknowledgments Editage has checked and revised English writing in the manuscript.
J. Shin et al.
Appendix A. Supplementary data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.numecd.2015.12.007. Appendix 1. Medication list for dyslipidemia based on the generic names.
Simvastatin Clopidogrel Atorvastatin(calcium) Nicotinic acid Pitavastatin calcium Rosuvastatin calcium Rosuvastatin calcium(as rosuvastatin 5 mg) Rosuvastatin calcium(as rosuvastatin 20 mg) Pravastatin sodium Micronized fenofibrate Fluvastatin Ezetimibe Acipimox Probucol Fenofibrate Ethyl linoleate Clopidogrel(besylate) Lovastatin Cerivastatin sodium
References [1] Kim KJ, Choi SI, Lee MS, Kim JA, Chun EJ, Jeon CH. The prevalence and characteristics of coronary atherosclerosis in asymptomatic subjects classified as low risk based on traditional risk stratification algorithm: assessment with coronary CT angiography. Heart 2013;99(15):1113e7. [2] Lee YH, Lee SG, Lee MH, Kim JH, Lee BW, Kang ES, et al. Serum cholesterol concentration and prevalence, awareness, treatment, and control of high low-density lipoprotein cholesterol in the Korea National Health and Nutrition Examination Surveys 2008e2010: beyond the tip of the Iceberg. J Am Heart Assoc 2014; 3(1):446. [3] Office KNS. Korean national health and nutritional examination survey. Available from:. 2015. http://kosis.kr/statisticsList/ statisticsList_01List.jsp?vwcdZMT_ZTITLE&parentIdZD#SubCont [retrieved on 11 June 2015]. [4] Kershaw KN, Osypuk TL, Do DP, De Chavez PJ, Diez Roux AV. Neighborhood-level racial/ethnic residential segregation and incident cardiovascular disease: the multi-ethnic study of atherosclerosis. Circulation 2015;131(2):141e8. [5] Robinson JG, Farnier M, Krempf M, Bergeron J, Luc G, Averna M, et al. Efficacy and safety of alirocumab in reducing lipids and cardiovascular events. N Engl J Med 2015;372(16):1489e99. [6] Sabatine MS, Giugliano RP, Wiviott SD, Raal FJ, Blom DJ, Robinson J, et al. Efficacy and safety of evolocumab in reducing lipids and cardiovascular events. N Engl J Med 2015;372(16):1500e9. [7] FORCE USPST. Lipid disorders in adults (Cholesterol, dyslipidemia). Screening Available from:. 2008. http://www.uspreventiveservice staskforce.org/Page/Document/RecommendationStatementFinal/ lipid-disorders-in-adults-cholesterol-dyslipidemia-screening [retrieved on 11 June 2015]. [8] Strandberg TE, Kolehmainen L, Vuorio A. Evaluation and treatment of older patients with hypercholesterolemia: a clinical review. Jama 2014;312(11):1136e44. [9] Casavalle PL, Lifshitz F, Romano LS, Pandolfo M, Caamano A, Boyer PM, et al. Prevalence of dyslipidemia and metabolic syndrome risk factor in overweight and obese children. Pediatr Endocrinol Rev 2014;12(2):213e23.
Combined effect of individual and neighborhood socioeconomic status on mortality in patients with newly diagnosed dyslipidemia [10] Kinra S, Bowen LJ, Lyngdoh T, Prabhakaran D, Reddy KS, Ramakrishnan L, et al. Sociodemographic patterning of noncommunicable disease risk factors in rural India: a cross sectional study. BMJ 2010;341. c4974. [11] Rodriguez CJ, Daviglus ML, Swett K, Gonzalez HM, Gallo LC, Wassertheil-Smoller S, et al. Dyslipidemia patterns among Hispanics/Latinos of diverse background in the United States. Am J Med 2014;127(12):1186e94. e1. [12] Pursnani A, Massaro J, D’Agostino R, O’Donnell C, Hoffmann U. Guideline-based statin allocation to distinguish both cardiovascular events and coronary artery calcification in the community: a comparison of the 2004 NCEP ATP III and 2013 ACC/AHA cholesterol treatment guideline. J Am Coll Cardiol 2015;65 (10_S). [13] Kavousi M, Leening MJ, Nanchen D, Greenland P, Graham IM, Steyerberg EW, et al. Comparison of application of the ACC/AHA guidelines, adult treatment panel III guidelines, and European society of cardiology guidelines for cardiovascular disease prevention in a European cohort. Jama 2014;311(14):1416e23. [14] Kim J. Strategies to enhance the use of National Health Insurance claims database in generating health statistics. Technical report by Health Insurance Review and Assessment Services 2005. in Korean. [15] Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40(5):373e83. [16] Morgan O, Baker A. Measuring deprivation in England and Wales using 2001 Carstairs scores. Health Stat Q 2006;31:28e33. [17] Gyu Lee Sang. The effect of neighborhood socioeconomic factors on spatial mortality and individual health status. PhD diss. Yonsei University; 2002. [18] Kapral MK, Wang H, Mamdani M, Tu JV. Effect of socioeconomic status on treatment and mortality after stroke. Stroke 2002;33(1): 268e73. [19] Kumral E, Evyapan D, Gokcay F, Karaman B, Orman M. Association of baseline dyslipidemia with stroke recurrence within five-years after ischemic stroke. Int J Stroke 2014;9(Suppl A100):119e26. [20] Cannon CP. Mixed dyslipidemia, metabolic syndrome, diabetes mellitus, and cardiovascular disease: clinical implications. Am J Cardiol 2008;102(12a):5le9l. [21] Madssen E, Laugsand LE, Wiseth R, Morkedal B, Platou C, Vatten L, et al. Risk of acute myocardial infarction: dyslipidemia more
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
215
detrimental for men than women. Epidemiology 2013;24(5): 637e42. Honjo K, Iso H, Nakaya T, Hanibuchi T, Ikeda A, Inoue M, et al. Impact of neighborhood socioeconomic conditions on the risk of stroke in Japan. J Epidemiol 2015;25(3):254e60. Brown AF, Liang LJ, Vassar SD, Stein-Merkin S, Longstreth Jr WT, Ovbiagele B, et al. Neighborhood disadvantage and ischemic stroke: the Cardiovascular Health Study (CHS). Stroke 2011; 42(12):3363e8. Brown AF, Liang LJ, Vassar SD, Merkin SS, Longstreth Jr WT, Ovbiagele B, et al. Neighborhood socioeconomic disadvantage and mortality after stroke. Neurology 2013;80(6):520e7. Chichlowska KL, Rose KM, Diez-Roux AV, Golden SH, McNeill AM, Heiss G. Individual and neighborhood socioeconomic status characteristics and prevalence of metabolic syndrome: the Atherosclerosis Risk in Communities (ARIC) Study. Psychosom Med 2008;70(9):986e92. Chichlowska KL, Rose KM, Diez-Roux AV, Golden SH, McNeill AM, Heiss G. Life course socioeconomic conditions and metabolic syndrome in adults: the Atherosclerosis Risk in Communities (ARIC) Study. Ann Epidemiol 2009;19(12):875e83. Boyce WT, Chesterman EA, Winkleby MA. Psychosocial predictors of maternal and infant health among adolescent mothers. Am J Dis Child 1991;145(3):267e73. Williams DR, Yu Yan, Jackson JS, Anderson NB. Racial differences in physical and mental health: socio-economic status, stress and discrimination. J Health Psychol 1997;2(3):335e51. Link BG, Northridge ME, Phelan JC, Ganz ML. Social epidemiology and the fundamental cause concept: on the structuring of effective cancer screens by socioeconomic status. Milbank Q 1998;76(3): 375e402. Roos LL, Traverse D, Turner D. Delivering prevention: the role of public programs in delivering care to high-risk populations. Med Care 1999;37(6). JS264-78. Roos LL, Magoon J, Gupta S, Chateau D, Veugelers PJ. Socioeconomic determinants of mortality in two Canadian provinces: multilevel modelling and neighborhood context. Soc Sci Med 2004;59(7):1435e47. Li X, Sundquist J, Calling S, Zöller B, Sundquist K. Neighborhood deprivation and risk of cervical cancer morbidity and mortality: a multilevel analysis from Sweden. Gynecol Oncol 2012;127(2):283e9.