Abstract
Background: There is considerable interest in the trending discrepancy between ischemic heart disease (IHD) and heart failure (HF) in vital statistics. Clinically, acute myocardial infarction (AMI) and stroke are closely associated with HF, but their contribution to HF as the underlying cause of death (UCD) is unclear.
Methods and Results: In 1990 and 1992–1993, we enrolled a total of 140,420 residents of Japanese nationality (aged 40–69 years) from 11 public health center areas. We prospectively examined the occurrence of cardiovascular disease (CVD), including AMI, sudden cardiac death within 1 h (SCD), and stroke, and analyzed the 14,375 participants without a history of CVD at baseline who died during the 20-year follow-up. A time-dependent Cox proportional hazards model was used to estimate hazard ratios and the population attributable fraction (PAF) of AMI, AMI+SCD, stroke, and CVD for deaths due to HF, IHD, and cerebrovascular disease as the UCD, adjusted for individuals’ lifestyles and comorbid conditions. The PAF of AMI for HF deaths was 2.4% (95% confidence interval [CI] 1.7–2.9%), which increased to 12.0% (95% CI 11.6–12.2%) for AMI+SCD. The PAF of CVD-attributed HF deaths was estimated to be 17.6% (95% CI 15.9–18.9%).
Conclusions: HF as the UCD was partly explained by CVD. The data imply that most HF deaths reported in vital statistics may be associated with underlying causes other than CVD.
Vital statistics in the US show that age-adjusted mortality rates for ischemic heart disease (IHD) and cerebrovascular disease are decreasing; in contrast, the mortality rate for heart failure (HF) is increasing.1,2 The effects of aging cannot fully explain this discrepancy, and the underlying reasons are not well understood. Although acute myocardial infarction (AMI) is known to be a major cause of HF, and stroke is also associated with increased risk of major adverse cardiovascular events including HF,3 the magnitude of their contributions to HF deaths has not been analyzed.
On the vital statistics, certain underlying causes of death (UCD), such as HF, have been incorrect in identifying the actual underlying cause.4 HF assigned on death certificates is often used to describe a mode of death.5 Therefore, the UCD is unlikely to be ascertained from a report of HF on vital statistics. When the death certificate diagnosis is validated, HF as the UCD is primarily reclassified as coronary artery disease (CAD) in the US and other countries.4,6 The Atherosclerosis Risk in Communities (ARIC) study found a trend towards increased CAD mortality as a result of HF reclassification, indicating that HF deaths include many causes of death related to CAD.6
Aside from the accuracy of the diagnosis of HF deaths, it is important to understand the extent to which the development of cardiovascular disease (CVD) is involved in the HF deaths reported in the vital statistics. Consequently, we focused on subjects who died during our prospective study and tried to analyze the effect of CVD on the death certificate diagnosis of HF and other CVD using a time-dependent Cox proportional hazards model. We aimed to clarify how much incident CVD contributed to HF coded by the International Classification of Diseases (ICD) on death certificates, which reflects vital statistics.
Methods
Study Population
The Japan Public Health Center-based Prospective (JPHC) Study began in 1990 (Cohort I) and continued in 1993–1994 (Cohort II) in 11 public health center areas.7 The total study population consisted of 140,420 residents of Japanese nationality, aged 40–59 years (Cohort I) or 40–69 years (Cohort II). Participants were asked about their lifestyle habits at the baseline examination. After excluding 2 communities without CVD registration (n=23,524), we selected individuals who responded to the baseline questionnaire (n=95,405). For our analysis, we omitted those who had a history of myocardial infarction or stroke (n=2,126) and excluded those who were alive (n=68,223). Furthermore, we excluded those who moved out of the community because we could not conduct a follow-up to identify incident outcomes (n=10,699). Ultimately, 14,375 individuals who died with a UCD assigned on their death certificate were analyzed (Supplementary Figure).
This study was conducted in accordance with the Declaration of Helsinki, and the study protocol, including the informed consent procedure of the JPHC Study, was approved by the human ethics review committees of the National Cancer Center, Faculty of Medicine, Oita University (Approval #1523) and each registered hospital.
Measurements
We used a self-administered questionnaire to ascertain smoking status, alcohol drinking status, sports and regular exercise habits, and medical history. Smoking status was classified as non-smoker, ex-smoker, smoking <20 cigarettes/day, and smoking ≥20 cigarettes/day. Alcohol drinking status was defined as an alcohol drinking habit of at least once a week. Individuals who participated in sports and regular exercise at least 1–2 times a week were considered to have the habit. Missing data on smoking status, alcohol drinking status, and sports and regular exercise were treated as unknown. The use of medications for hypertension, diabetes, cancer (stomach, lung, colon, liver, breast, uterine, and other cancers), hepatitis, and kidney disease (yes/no) was identified by the question, “Have the following conditions been treated by physicians?”, with hypertension, diabetes, cancer, hepatitis, and kidney disease offered as potential responses.
Follow-up
The median follow-up period was 18.5 years from 1990 (Cohort I) and 1993 (Cohort II) to the end of 2009 (Cohort I) or 2012 (Cohort II). Person-years were calculated as the period from the date of baseline to that of death.
The UCD was determined according to the International Classification of Diseases, ninth revision (ICD-9) until 1995, and translated into the corresponding International Classifications of Diseases (ICD), 10th revision (ICD-10) codes or coded by the ICD-10 after that. All UCD codes were provided with permission from the Japanese Ministry of Health, Labour and Welfare. Deaths from HF, IHD, cerebrovascular disease, and other CVD were defined as ICD-9/ICD-10 codes 428/I50; 410–414, 429.2/I20–I25; 430–438/I60–I69; and 390–459/I00–I99 (except for the above codes), respectively. The remaining UCDs were considered non-CVD deaths.
Determinants of the Occurrence of AMI, Sudden Cardiac Death Within 1 h, and Stroke
The records of 78 major hospitals located in communities capable of treating patients with AMI and stroke were systematically surveyed for the occurrence of AMI, sudden cardiac death within 1 h of the onset of the event (SCD), and stroke. Medical records of possible stroke hospitalizations were reviewed by physicians blinded to the exposures. In addition, 5- and 10-year follow-up questionnaires were sent to participating hospitals to complete non-fatal AMI, SCD, and stroke surveillance.
AMI was confirmed from patient medical records according to the criteria of the MONICA Project (monitoring trends and determinants in cardiovascular disease),8 and strokes were confirmed according to the National Survey of Stroke criteria9 (Supplementary Table 1). In the present study, AMI+SCD was considered CAD.
Finally, CVD was defined as the presence of AMI, SCD, or stroke. If participants had AMI, SCD, and stroke, the event that occurred first was identified as the outcome.
Statistical Analysis
Proportions were calculated for each UCD for age, smoking status, alcohol drinking status, exercise habit, history of illness, and occurrence of AMI, AMI+SCD, stroke, and CVD including those events.
We used a time-dependent Cox proportional hazards model to examine the association between the occurrence of AMI, AMI+SCD, or stroke and the UCD (i.e., HF, IHD, cerebrovascular disease, other CVD, and non-CVD). Because the date of event occurrence varied among individuals, we adopted the time-dependent model to fit the association between event occurrence and the UCD. Hazard ratios (HR) and 95% confidence intervals (CIs) were calculated. The population attributable fraction (PAF) was calculated using the formula p × (HR − 1) / HR, using adjusted HRs,10 where p represents the percentage of the event (AMI, AMI+SCD, stroke) in each UCD. HRs were calculated by creating a model that adjusted for sex and age while stratifying by study community and a model that further adjusted for smoking status, alcohol drinking status, sports and regular exercise habit, and illnesses, including hypertension, diabetes, cancer, hepatitis, and kidney disease (“multivariable-adjusted HRs”). The PAFs reported were calculated using the multivariable-adjusted HRs. In addition, multivariable-adjusted HRs and PAFs were calculated for each of 3 age groups (40–49, 50–59, and 60–69 years). For those who experienced an AMI+SCD or stroke, the time from onset to death was divided into the following categories: <1, 1–3, and >3 years.
Statistical significance was assumed at P<0.05 (two-tailed). SAS version 9.4 (SAS Institute Inc., Cary, NC, USA) was used for all analyses.
Results
Table 1 presents population characteristics by UCD in men and women. There were 14,375 deaths: 495 with HF, 818 with IHD, 1,437 with cerebrovascular disease, 790 with other CVD, and 10,835 with non-CVD as the UCD. Deaths from HF were more common in individuals in their 60 s than from other causes of death. Smoking and diabetes were more common among deaths due to IHD. The proportion of ICD-9 diagnoses among the UCDs was highest for HF deaths.
Table 1. Population Characteristics According to Underlying Cause of Death
| |
Underlying cause of death (n=14,375) |
| HF |
IHD |
Cerebrovascular
disease |
Other CVD |
Non-CVD |
| n |
495 |
818 |
1,437 |
790 |
10,835 |
| Male sex (%) |
56.0 |
66.9 |
59.6 |
60.3 |
65.0 |
| Age group (%) |
| 40–49 years |
12.7 |
16.5 |
19.0 |
15.8 |
18.0 |
| 50–59 years |
32.9 |
37.0 |
39.0 |
44.7 |
42.2 |
| 60–69 years |
54.3 |
46.5 |
42.0 |
39.5 |
39.9 |
| Smoking status (%) |
| Never |
60.4 |
51.3 |
54.8 |
53.0 |
53.0 |
| Ex-smoker |
2.6 |
3.1 |
4.7 |
5.8 |
4.9 |
| <20 cigarettes/day |
13.7 |
14.3 |
14.0 |
13.4 |
13.4 |
| ≥20 cigarettes/day |
22.2 |
30.1 |
25.6 |
26.8 |
27.7 |
| Unknown |
1.0 |
1.2 |
0.9 |
0.9 |
1.1 |
| Alcohol drinking status (%) |
| Regular drinker |
34.1 |
39.1 |
40.8 |
42.8 |
42.7 |
| Sports and regular exercise habit (%) |
| ≥1–2 times/week |
15.6 |
17.4 |
16.0 |
17.2 |
17.2 |
| Illnesses (%) |
| Hypertension |
25.9 |
30.7 |
31.5 |
33.3 |
22.2 |
| Diabetes |
11.9 |
13.8 |
9.3 |
9.0 |
8.9 |
| Cancer |
3.4 |
2.3 |
2.6 |
1.5 |
3.2 |
| Hepatitis |
2.6 |
1.7 |
2.5 |
1.4 |
4.8 |
| Kidney disease |
2.0 |
2.7 |
4.0 |
3.5 |
3.1 |
| Proportion of ICD-9 diagnosis (%) |
13.7 |
4.4 |
7.7 |
3.7 |
6.8 |
CVD, cardiovascular disease; HF, heart failure; ICD-9, International Classification of Diseases, 9th revision; IHD, ischemic heart disease.
Table 2 summarizes incident AMI, AMI+SCD, stroke, and CVD by the UCD. The frequency of AMI, AMI+SCD, stroke, and CVD prior to death is shown according to the UCD. AMI+SCD and CVD accounted for 12.9% and 24.4%, respectively, of deaths due to HF as the UCD. AMI+SCD accounted for 39.9% of deaths due to IHD, and stroke accounted for 70.4% of deaths due to cerebrovascular disease as the UCD.
Table 2. Occurrence of AMI, AMI+SCD, Stroke and CVD According to the Underlying Cause of Death
| Event |
Underlying cause of death (n=14,375) |
HF
(n=495) |
IHD
(n=818) |
Cerebrovascular
disease (n=1,437) |
Other CVD
(n=790) |
Non-CVD
(n=10,835) |
| AMI |
18 (3.6) |
220 (26.9) |
14 (1.0) |
12 (1.5) |
138 (1.3) |
| AMI+SCD |
64 (12.9) |
326 (39.9) |
24 (1.7) |
50 (6.3) |
152 (1.4) |
| Stroke |
63 (12.7) |
80 (9.8) |
1,011 (70.4) |
131 (16.6) |
782 (7.2) |
| CVD |
121 (24.4) |
374 (45.7) |
1,023 (71.2) |
177 (22.4) |
909 (8.4) |
Data show n (%). The percentages in parentheses were calculated using the number of each underlying cause of death as the denominator. Acute myocardial infarction (AMI) included definite or possible AMI. Sudden cardiac death (SCD) meant a death of unknown origin that occurred within 1 h of the onset of the event. Cardiovascular disease (CVD) was defined as AMI, SCD, or stroke. HF, heart failure; IHD, ischemic heart disease.
Table 3 presents the time-dependent sex- and age-adjusted, as well as multivariable-adjusted HRs of AMI, AMI+SCD, and stroke for each UCD, stratified by study community. Although HRs adjusted for sex and age decreased slightly after adjustment, the multivariable-adjusted HR for AMI occurrence was 2.99 (95% CI 1.85–4.86) for deaths due to HF; it increased to 13.3 (95% CI 10.0–17.6) when SCD was also considered. Similarly, the PAF of AMI for HF deaths was 2.4% (95% CI 1.7–2.9), but that for AMI+SCD increased to 12.0% (95% CI 11.6–12.2). CVD attributed to HF deaths was estimated to be 17.6% (95% CI 15.9–18.9) of the PAF. AMI+SCD and stroke were strongly associated with deaths due to IHD and cerebrovascular disease, respectively. The occurrence of stroke was associated with other CVD. We analyzed the results separately for men and women but found few differences (Supplementary Tables 2,3). In addition, we reanalyzed data restricted to deaths assigned by ICD-10 diagnosis to exclude the influence of the ICD revision (Supplementary Table 4). The HR for AMI+SCD decreased slightly for HF as the UCD, but the difference was not significant.
Table 3. Association of the Occurrence of AMI, AMI+SCD, Stroke and CVD With the Recorded Underlying Causes of Death
| Event |
Underlying cause of death (n=14,375) |
HF
(n=495) |
IHD
(n=818) |
Cerebrovascular
disease (n=1,437) |
Other CVD
(n=790) |
Non-CVD
(n=10,835) |
| AMI |
| No. cases |
18 |
220 |
14 |
12 |
138 |
| Sex- and age-adjusted HR (95% CI) |
3.14 (1.94, 5.07) |
27.0 (22.9, 31.9) |
0.76 (0.45, 1.29) |
1.12 (0.63, 1.98) |
0.94 (0.79, 1.11) |
| Multivariable-adjusted HR (95% CI) |
2.99 (1.85, 4.86) |
25.7 (21.7, 30.5) |
0.72 (0.42, 1.22) |
1.05 (0.59, 1.86) |
0.96 (0.81, 1.14) |
| PAF (95% CI) (%) |
2.4 (1.7, 2.9) |
25.8 (25.6, 26.0) |
−0.4 (−1.3, 0.2) |
0.1 (−1.1, 0.7) |
−0.1 (−0.3, 0.2) |
| AMI+SCD |
| No. cases |
64 |
326 |
24 |
50 |
152 |
| Sex- and age-adjusted HR (95% CI) |
13.4 (10.2, 17.7) |
53.0 (45.4, 61.7) |
1.32 (0.88, 1.99) |
4.92 (3.67, 6.60) |
1.04 (0.88, 1.22) |
| Multivariable-adjusted HR (95% CI) |
13.3 (10.0, 17.6) |
52.8 (45.0, 61.9) |
1.25 (0.83, 1.88) |
4.70 (3.49, 6.31) |
1.06 (0.90, 1.25) |
| PAF (95% CI) (%) |
12.0 (11.6, 12.2) |
39.1 (39.0, 39.2) |
0.3 (−0.3, 0.8) |
5.0 (4.5, 5.3) |
0.1 (−0.2, 0.3) |
| Stroke |
| No. cases |
63 |
80 |
1,011 |
131 |
782 |
| Sex- and age-adjusted HR (95% CI) |
1.74 (1.32, 2.28) |
1.22 (0.96, 1.55) |
44.3 (39.0, 50.4) |
2.19 (1.80, 2.66) |
0.93 (0.87, 1.00) |
| Multivariable-adjusted HR (95% CI) |
1.72 (1.31, 2.26) |
1.15 (0.91, 1.46) |
45.3 (39.9, 51.6) |
2.10 (1.72, 2.55) |
0.94 (0.87, 1.01) |
| PAF (95% CI) (%) |
5.3 (3.0, 7.1) |
1.3 (−1.0, 3.1) |
68.8 (68.6, 69.0) |
8.7 (7.0, 10.1) |
−0.5 (−1.1, 0.1) |
| CVD |
| No. cases |
121 |
374 |
1,023 |
177 |
909 |
| Sex- and age-adjusted HR (95% CI) |
3.59 (2.89, 4.47) |
9.74 (8.39, 11.3) |
40.2 (35.3, 45.7) |
2.80 (2.35, 3.34) |
0.94 (0.88, 1.01) |
| Multivariable-adjusted HR (95% CI) |
3.57 (2.87, 4.45) |
9.38 (8.06, 10.9) |
41.2 (36.2, 46.9) |
2.69 (2.25, 3.21) |
0.95 (0.88, 1.02) |
| PAF (95% CI) (%) |
17.6 (15.9, 18.9) |
40.8 (40.0, 41.5) |
69.5 (69.2, 69.7) |
14.1 (12.5, 15.4) |
−0.5 (−1.1, 0.1) |
Hazard ratios (HRs) were calculated using a time-dependent Cox proportional hazard model stratified by the study community. The model was further adjusted for smoking, alcohol drinking, sports and exercise habit, and illnesses (hypertension, diabetes, cancer, hepatitis, kidney disease). The population attributable fraction (PAF) was estimated using multivariable-adjusted HRs. CI, confidence interval; SCD, sudden cardiac death within 1 h. Other abbreviations as in Table 2.
Age-specific multivariable-adjusted HRs for and PAFs of AMI+SCD and stroke for the UCD are shown in Figure 1. The HRs for AMI+SCD were higher for HF deaths among younger age groups. In contrast, stroke occurred more frequently in older age groups. The PAFs of HF deaths were higher in the 40–49- and 50–59-year age groups for AMI and in the 60–69-year age group for stroke. Large contributions of the occurrence of AMI+SCD and stroke to IHD and cerebrovascular disease deaths, respectively, were found in each age group.
Concerning the occurrence of AMI+SCD and stroke, the time from onset to death was divided into 3 periods: <1, 1–3, and >3 years (Figure 2). Less than 1 year after event onset, the percentage of HF deaths among those with AMI+SCD was 6.3%, which decreased to 2.8% after 3 years. Similarly, only 1.1% of those who had a stroke had HF assigned as the UCD; however, the percentage increased to 4.7% after 3 years. AMI+SCD or stroke was assigned as the UCD mainly if the patient died within 1 year of developing AMI+SCD or stroke.
Discussion
The results of this prospective study demonstrate how the presence of CVD prior to death influences HF, IHD, and cerebrovascular disease as the UCD. Briefly, the PAF of AMI for HF as the UCD was only 2.4% (95% CI 1.7–2.9), but for AMI+SCD it increased to 12.0% (95% CI 11.6–12.2). CVD contributed to 17.6% (95% CI 15.9–18.9) of HF as the UCD. To the best of our knowledge, this is the first prospective study that has tried to clarify the association of CVD with HF as the UCD.
The association between AMI+SCD and HF deaths was stronger in the younger age group. However, the PAF was not necessarily greater because of the lower frequency. In contrast, the association between stroke and HF deaths was stronger in the older age group. Stroke occurrence leading to HF deaths may be characteristic of elderly patients with a variety of comorbid conditions.
As stated in clinical guidelines for HF, HF is caused by various illnesses, such as hypertension, CAD, valvular heart disease, and diabetes.11 Nonetheless, HF assigned by death certificates is often used to represent a mode of death,5 and it is hard to determine the underlying cause. AMI is considered a major cause of HF, and stroke was associated with post-stroke major adverse cardiovascular events, including incident HF.3 Therefore, we tried to find a relationship between HF deaths and AMI; AMI+SCD; stroke; and CVD in a population-based prospective study.
In general, a validation study of the cause of death reported on death certificates was performed based on a review of medical records by a panel of physicians.6,12,13 However, we previously found that medical information, such as a past history of myocardial infarction or stroke, was insufficient in a retrospective investigation;14 therefore, it was difficult to validate the UCD in some cases, especially for HF deaths, nearly half of which lacked such information.13 Although the study population was limited to a cohort, we believe this study was able to compensate for the shortcomings of such retrospective studies, particularly in determining whether and to what extent the onset of CVD was involved in HF deaths, which were assigned on the death certificates without sufficient evidence.
Few studies have examined the validity of HF as the UCD as opposed to CAD.14 The ARIC Study demonstrated that HF as the UCD was redistributed to CAD in 37.1% of deaths and to stroke in 4.4% of deaths using coarsened exact matching.4,6 Validation studies conducted in Japan estimated that deaths from HF as the UCD may also include deaths from CAD, up to approximately 20% of HF deaths,14 which is close to our estimates. Alternatively, according to a statistical model-based estimation, most HF deaths as the UCD were redistributed to CAD in developed countries.5
The Kyoto Congestive Heart Failure Registry reported causes of death for 3,717 discharged patients with acute decompensated HF, followed by a median follow-up period of 470 days.15 In that study, 848 patients died during follow-up, with 324 (38.2%) classified as HF deaths. That study suggested that some deaths with HF as the UCD were clinical HF, but we do not know how many such deaths on the death certificates involved patients with diagnosed HF in the present study.
When we examined the UCD in those who experienced AMI+SCD or stroke, we found that in cases of death within 1 year of onset, the same UCD tended to be assigned, but the proportion of non-CVD deaths as the UCD increased by more than half after 1 year. In addition, we demonstrated that the proportion of HF deaths increased for AMI+SCD with a shorter time to death, but increased for stroke with a longer time to death. The addition of such acute deaths resulted in a larger PAF for HF, suggesting that acute deaths tended to be recorded as HF on death certificates in Japan. Indeed, the validation study demonstrated that “acute heart failure” as the UCD on death certificates involved much more validated sudden deaths than “heart failure”.16
Vital statistics trends provide important information for future health policy. Therefore, it is necessary to understand what is responsible for the recent increase in HF deaths, but the evidence is very limited. Furthermore, it is unclear whether the increase in patients with clinical HF reflects the vital statistics trends. At the very least, this study was able to show that the development of CVD was associated with HF deaths, but only in part, and further research is needed to determine what else is responsible for the increasing HF mortality on the vital statistics.
Strengths and Limitations
The strength of our study was that the JPHC study registered CAD and stroke based on data from the main hospital for 20-year periods. Despite the strengths of this study, several limitations should be noted. First, the results of this study would apply only to the same age group, with the population of the present study comprising residents in the study area aged 40–69 years. Second, the JPHC study recruited patients from 9 communities as the study population, but those areas were located in rural areas of Japan. Consequently, the results may be different between urban and rural areas in Japan. Third, CVD registration was predominantly performed based on hospitalization. Therefore, some cases of CVD were possibly missed if patients were in a hospital we did not investigate them, or when patients died from CVD outside of hospital. Fourth, we excluded those who had a history of CVD and could not estimate the effect of past CVD occurrences on HF deaths. Accordingly, when applied to the general population, our findings may result in an underestimation. Fifth, although we analyzed all UCD data, including deaths coded by the ICD-9 and ICD-10, and a sensitivity analysis was performed to exclude the ICD-9, we could not identify an effect of ICD revision. Nonetheless, our estimates could also have been affected by changes in the ICD rules. Finally, we did not have information on atrial fibrillation, which is also considered to be a major cause of HF.
Conclusions
HF as the UCD was, to some extent, explained by CVD, including SCDs, but not enough to fully understand the cause related to HF as the UCD in Japan. Our data suggest that strategies to reduce mortality rates from HF are limited to CVD risk control alone. Future research is needed to understand the major cause related to HF as the UCD in Japan.
Acknowledgments
The authors acknowledge the members of the study groups. JPHC members (as of 2021) are listed at https://epi.ncc.go.jp/en/jphc/781/8896.html.
Sources of Funding
This study was supported by the National Cancer Center Research and Development Fund (since 2011) and a Grant-in-Aid for Cancer Research from the Ministry of Health, Labour and Welfare of Japan (from 1989 to 2010).
Disclosures
K.Y. is a member of Circulation Journal’s Editorial Team.
IRB Information
The present study was approved by the Ethics Committee of the Faculty of Medicine, Oita University (Reference no. 1523).
Supplementary Files
Please find supplementary file(s);
https://doi.org/10.1253/circj.CJ-22-0805
References
- 1.
Shah NS, Molsberry R, Rana JS, Sidney S, Capewell S, O’Flaherty M, et al. Heterogeneous trends in burden of heart disease mortality by subtypes in the United States, 1999–2018: Observational analysis of vital statistics. BMJ 2020; 370: m2688.
- 2.
Pagidipati NJ, Gaziano TA. Estimating deaths from cardiovascular disease: A review of global methodologies of mortality measurement. Circulation 2013; 127: 749–756.
- 3.
Sposato LA, Lam M, Allen B, Shariff SZ, Saposnik G; Paradise Study Group. First-ever ischemic stroke and incident major adverse cardiovascular events in 93 627 older women and men. Stroke 2020; 51: 387–394.
- 4.
Stevens GA, King G, Shibuya K. Deaths from heart failure: Using coarsened exact matching to correct cause-of-death statistics. Popul Health Metr 2010; 8: 6.
- 5.
Ahern RM, Lozano R, Naghavi M, Foreman K, Gakidou E, Murray CJ. Improving the public health utility of global cardiovascular mortality data: The rise of ischemic heart disease. Popul Health Metr 2011; 9: 8.
- 6.
Snyder ML, Love SA, Sorlie PD, Rosamond WD, Antini C, Metcalf PA, et al. Redistribution of heart failure as the cause of death: The Atherosclerosis Risk in Communities Study. Popul Health Metr 2014; 12: 10.
- 7.
Tsugane S, Sawada N. The JPHC study: Design and some findings on the typical Japanese diet. Jpn J Clin Oncol 2014; 44: 777–782.
- 8.
Tunstall-Pedoe H, Kuulasmaa K, Amouyel P, Arveiler D, Rajakangas AM, Pajak A. Myocardial infarction and coronary deaths in the World Health Organization MONICA Project: Registration procedures, event rates, and case-fatality rates in 38 populations from 21 countries in four continents. Circulation 1994; 90: 583–612.
- 9.
Walker AE, Robins M, Weinfeld FD. The National Survey of Stroke: Clinical findings. Stroke 1981; 12(Suppl 1): I13–I44.
- 10.
Rockhill B, Newman B, Weinberg C. Use and misuse of population attributable fractions. Am J Public Health 1998; 88: 15–19.
- 11.
Tsutsui H, Isobe M, Ito H, Ito H, Okumura K, Ono M, et al. JCS 2017/JHFS 2017 guideline on diagnosis and treatment of acute and chronic heart failure: Digest version. Circ J 2019; 83: 2084–2184.
- 12.
Lloyd-Jones DM, Martin DO, Larson MG, Levy D. Accuracy of death certificates for coding coronary heart disease as the cause of death. Ann Intern Med 1998; 129: 1020–1026.
- 13.
Rao C, Yang G, Hu J, Ma J, Xia W, Lopez AD. Validation of cause-of-death statistics in urban China. Int J Epidemiol 2007; 36: 642–651.
- 14.
Saito I. Review of death certificate diagnosis of coronary heart disease and heart failure in Japan. Nihon Koshu Eisei Zasshi 2004; 51: 909–916.
- 15.
Kitai T, Miyakoshi C, Morimoto T, Yaku H, Murai R, Kaji S, et al. Mode of death among Japanese adults with heart failure with preserved, midrange, and reduced ejection fraction. JAMA Netw Open 2020; 3: e204296.
- 16.
Yamashita T, Ozawa H, Aono H, Hosokawa H, Saito I, Ikebe T. Heart disease deaths on death certificates re-evaluated by clinical records in a Japanese city. Jpn Circ J 1997; 61: 331–338.
