Abstract
Background
We report on a retrospective population study aimed at identifying and validating the progression of adapted Diabetes Complications Severity Index (DCSI) for acute coronary syndrome (ACS), ischemic stroke, and mortality in Asian people with type 2 diabetes mellitus (DM).
Methods
Utilizing a Taiwanese national dataset, we included 84 450 type 2 diabetic individuals between 2000 and 2011. The area under the receiver operating characteristic curve (C statistics of logistic model) and the C statistics of the Cox model were used to evaluate whether the progression of diabetic complication status could be a predictor of ACS, ischemic stroke, and death. The optimum threshold for adverse outcomes risk stratification were obtained using Youden's J statistic as the cutoff that gives the highest threshold.
Results
Among the study patients, the C statistics of the logistic model of the progression of the score predictive of ACS, ischemic stroke, and death were 0.72 (95% confidence interval [CI]: 0.71–0.73), 0.84 (95% CI: 0.84–0.85), and 0.66 (95% CI: 0.65–0.67), respectively. The progression of adapted DCSI had moderate discrimination for ACS, ischemic stroke, and death (C statistics = 0.71, 0.72, and 0.75, respectively) based on Cox regression analysis (Harrell C). The optimum threshold of the progression of the score for ACS, ischemic stroke, and death in type 2 DM patients were 0.30, 0.36, and 0.39, respectively.
Conclusions
The acceptable discriminative power of the progression of adapted DCSI for Asian people affected by type 2 DM was demonstrated in a large cohort in Taiwan.
Keywords: Acute Coronary Syndrome, Death, Diabetes Complications Severity Index, Ischemic Stroke, Prediction
1. INTRODUCTION
Diabetes mellitus (DM) is a global medical burden and is clearly linked to cardiovascular and cerebrovascular disease.1, 2, 3 Type 2 DM has been recognized as myocardial infarction equivalent and several micro‐ and macrovasculopathies have already been developed in the early stage of diabetes or even in the prediabetic stage.1, 2, 3, 4, 5, 6 To decrease the adverse cardiovascular events and the associated sequelae, early identification of diabetic subjects at risk of acute cardiovascular events is of paramount importance.
The adapted Diabetes Complications Severity Index (DCSI), which incorporates 7 categories of diabetic complications, is a modified version of a risk scheme without consideration of the laboratory value.7, 8 The use of the adapted DCSI in predicting mortality has been demonstrated and validated with good stratification power.7
Clinically, patients with type 2 DM were approached for the possibility of acute cerebral and cardiovascular complications, such as acute coronary syndrome (ACS) and ischemic stroke based on the symptoms and history taking from the patients. It is possible that adapted DCSI and changes in the score over time were associated with adverse cardiovascular outcomes based on current knowledge.7, 8 To provide clinical feasibility for improving acute adverse cerebro‐cardiovascular events identification in diabetic patients, this study are thus conducted to examine whether progression of diabetes complication score was predictive of all the cardiovascular outcomes.
2. METHODS
2.1. Data source
Data used in this retrospective cohort study were extracted from the Longitudinal Health Insurance Database 2000 (LHID2000). The LHID2000 consists of registration files and original claims data from 1996 to 2011 of 1 000 000 beneficiaries randomly selected from the Taiwan National Health Insurance (TNHI) program.9 The details of the TNHI program and LHID2000 have been thoroughly discussed in previous studies.10, 11 The coding system of the International Classification of Diseases, 9th Revision, Clinical Modification (ICD‐9‐CM) was used to ascertain disease diagnoses. The research ethics committee of China Medical University and Hospital in Taiwan approved the present study (CMUH‐104‐REC2–115).
2.2. Study population
To evaluate the risk of ACS, ischemic stroke, and death in patients with type 2 DM, we established a type 2 DM cohort and follow them up. We selected the new onset type 2 DM patients ages ≥18 years (ICD‐9‐CM codes 250.x0 and 250.x2) between January 1, 2000 and December 31, 2011. The date of initial type 2 DM diagnosis was defined as the index date. Each study patient received follow‐up until a diagnosis of ACS (ICD‐9‐CM codes 410, 411.1, 411.8), ischemic stroke (ICD‐9‐CM codes 433–437) or death, withdrawal from the TNHI program, or until the end of 2011. To measure the severity of type 2 DM, we used the adapted DCSI.7 The adapted DCSI had 7 categories including retinopathy, nephropathy, neuropathy, cerebrovascular disease, cardiovascular disease, peripheral vascular disease, and metabolic disease.7 The progression of diabetic complications status was defined as a yearly increase in adapted DCSI from the date of diagnosis to the end of follow‐up divided by total of follow‐up years, and the 4 progression groups were defined as a yearly increase in scores less than 0.15 (<50 percentiles), 0.15 to 0.30 (50–65 percentiles), 0.30 to 0.50 (65–80 percentiles), and greater than 0.50 (>80 percentiles) per year. Baseline comorbidities, such as hypertension (ICD‐9‐CM codes 401–405), chronic obstructive pulmonary disease (ICD‐9‐CM codes 491, 492, 496), and hyperlipidemia (ICD‐9‐CM code 272), were identified and included in the analyses.
2.3. Statistical analysis
Demographic data of the study patients were presented as number and percentage for category variables and as mean and standard deviation (SD) for continuous variables. The differences among groups were compared by χ2 test for categorical variables or 1‐way analysis of variance for continuous variables. To address the concern of constant proportionality, we examined the proportional hazard model assumption using a test of scaled Schoenfeld residuals. Results showed that there was no significant relationship between Schoenfeld residuals for the progression of adapted DCSI and follow‐up time (P = 0.91; P = 0.06) in the model evaluating the ACS and ischemic stroke risk. In the model evaluating the death risk throughout the overall follow‐up period, results of the test revealed a significant relationship between Schoenfeld residuals for the progression of adapted DCSI and follow‐up time, suggesting the proportionality assumption was violated. In the subsequent analyses, we stratified the follow‐up duration (≤5 years, >5 years) to deal with the violation of proportional hazard assumption. Univariable and multivariable Cox proportional hazard models were performed to examined whether the progression of type 2 DM complication score could be predictive of ACS, ischemic stroke, and death. Cumulative incidences of ACS, ischemic stroke, and death among the 4 progression groups were assessed using the Kaplan–Meier method, and the differences were determined by conducting a log‐rank test. The predictive performance of the progression of adapted DCSI with regard to ACS, ischemic stroke, and death was examined using the area under the receiver operating characteristic (ROC) curve (AUC) (C‐statistics of logistic model). The C statistics of the Cox model are slightly different from the AUC of the logistics model. We calculate the C statistics of the Cox model with Stata 14.0 (StataCorp, College Station, TX). The optimum threshold for ACS, ischemic stroke, and death risk stratification were obtained using the Youden's J statistic as the cutoff that gives the highest threshold. All of the above analyses were performed using SAS 9.4 software (SAS Institute, Cary, NC). The significance level was set at <0.05 for 2‐sided testing of the P value.
3. RESULTS
This study included 84 450 type 2 DM patents between 2000 and 2011 (Table 1). The mean age was 57.2 years (SD = 13.8 years). Most patients were female (51.8%), and the most common comorbidity was hypertension (57.6%). The major diabetes complication was cardiovascular disease (49.2%). The mean adapted DCSI was 1.19 (SD = 1.49) initially and was 2.64 (SD = 2.20) at the end of the follow‐up. The mean follow‐up period for ACS, ischemic stroke, and death were 5.97 years (SD = 3.53 years), 5.88 years (SD = 3.53 years), and 6.11 years (SD = 3.54 years), respectively (data not shown). Figure 1 illustrates the Kaplan–Meier survival curves of ACS, ischemic stroke, and death in type 2 DM patients stratified by the progression of adapted DCSI. Patients with a higher change in the score (>0.50 per year) were more likely to develop ACS, ischemic stroke, and death as compared to a score of an increase of <0.15 per year (P < 0.001).
Table 1.
Subject characteristics
| Variable | Change of Adapted DCSI per Year | P Valuea | ||||
|---|---|---|---|---|---|---|
| Subjects | <0.15 | 0.15–0.30 | 0.3–0.5 | >0.5 | ||
| N = 84 450 | N = 43 468 | N = 13 634 | N = 11 272 | N = 16 076 | ||
| Age, y, n (%) | <0.001 | |||||
| ≤49 | 25 736 (30.5) | 15 943 (36.7) | 4244 (31.1) | 2690 (23.9) | 2859 (17.8) | |
| 50–64 | 33 162 (39.3) | 16 815 (38.7) | 5788 (42.5) | 4673 (41.5) | 5886 (36.6) | |
| 65+ | 25 552 (30.3) | 10 710 (24.6) | 3602 (26.4) | 3909 (34.7) | 7331 (45.6) | |
| Mean (SD) | 57.2 (13.8) | 55.0 (14.1) | 56.4 (12.6) | 59.2 (12.7) | 62.6 (13.1) | <0.001 |
| Sex, n (%) | <0.001 | |||||
| Female | 43 739 (51.8) | 22 577 (51.9) | 6590 (48.3) | 5694 (50.5) | 8878 (55.2) | |
| Male | 40 711 (48.2) | 20 891 (48.1) | 7044 (51.7) | 5578 (49.5) | 7198 (44.8) | |
| Comorbidity, n (%) | ||||||
| Hypertension | 48 644 (57.6) | 22 938 (52.8) | 7728 (56.7) | 7009 (62.2) | 10 969 (68.2) | <0.001 |
| COPD | 14 878 (17.6) | 7154 (16.5) | 2184 (16.0) | 2140 (19.0) | 3400 (21.2) | <0.001 |
| Hyperlipidemia | 41 439 (49.1) | 21 280 (49.0) | 7040 (51.6) | 5780 (51.3) | 7339 (45.7) | <0.001 |
| Diabetes complication, n (%) | ||||||
| Retinopathy | 16 489 (19.5) | 2899 (6.67) | 3840 (28.2) | 4015 (35.6) | 5735 (35.7) | <0.001 |
| Nephropathy | 27 670 (32.8) | 7729 (17.8) | 5375 (39.4) | 5650 (50.1) | 8916 (55.5) | <0.001 |
| Neuropathy | 25 081 (29.7) | 7679 (17.7) | 5335 (39.1) | 5064 (44.9) | 7003 (43.6) | <0.001 |
| Cerebrovascular | 20 922 (24.8) | 5623 (12.9) | 2906 (21.3) | 3926 (34.8) | 8467 (52.7) | <0.001 |
| Cardiovascular | 41 527 (49.2) | 15 636 (36.0) | 7759 (56.9) | 7378 (65.5) | 10 754 (66.9) | <0.001 |
| Peripheral vascular disease | 19 651 (23.3) | 4962 (11.4) | 3564 (26.1) | 4242 (37.6) | 6883 (42.8) | <0.001 |
| Metabolic | 3691 (4.37) | 565 (1.30) | 407 (2.99) | 677 (6.01) | 2042 (12.7) | <0.001 |
| Mean adapted DCSI (SD), n (%) | ||||||
| Onset | 1.19 (1.49) | 1.25 (1.58) | 1.03 (1.34) | 1.11 (1.36) | 1.22 (1.45) | <0.001 |
| End of follow‐up | 2.64 (2.20) | 1.42 (1.59) | 2.83 (1.40) | 3.92 (1.62) | 4.89 (2.22) | <0.001 |
Abbreviations: COPD, chronic obstructive pulmonary disease; DCSI, Diabetes Complications Severity Index; SD, standard deviation
Comparison among groups with different change of adapted DCSI per year using χ2 test for categorical variables or 1‐way analysis of variance for continuous variables.
Figure 1.
Kaplan–Meier curves of the cumulative incidence of acute coronary syndrome (A), ischemic stroke (B), and death (C) in the 4 groups of the adapted DCSI change (0–0.15, 0.15–0.3, 0.3–0.5, and > 0.5 per year) among type 2 diabetic patients. Abbreviations: DCSI, Diabetes Complications Severity Index
Risk of ACS, ischemic stroke, and death increased with increasing changes in the score over time is shown in Table 2. Compared with type 2 DM patients diagnosed with mildly progressive diabetes (an increase in the adapted DCSI of <0.15 per year), the adjusted hazard ratio (aHR) of ACS increased with the progression of diabetes complication score (aHR = 1.49 and 4.53 for an increase in the score of 0.30–0.50 and > 0.50 per year, respectively). A significant trend was also found with regard to the ACS risk associated with the progression of type 2 DM complication score (P for trend <0.001). The incidence of ischemic stroke increased from 2.12 per 1000 person‐years in type 2 DM patients with a change in the score of 0 to 0.15 per year to 49.7 per 1000 person‐years for those with a change in the score of >0.5 per year. The corresponding aHR of ischemic stroke was 17.9 (95% confidence interval [CI]: 16.2–19.7) for patients with a change in the score of >0.5 per year compared to those with a change in the score of 0 to 0.15 per year. Among type 2 DM patients, the aHR of death was 2.13 (95% CI: 2.01–2.25) for those with a change in the score of >0.5 per year compared to those with a change in the score of 0 to 0.15 per year. In the first 5‐year follow‐up, the type 2 DM patients with the progression of adapted DCSI of >0.5 per year had a higher risk of death compared with the type 2 DM patients with the progression of adapted DCSI of <0.15 per year (aHR = 1.77, 95% CI: 1.66–1.90). Moreover, the risk of death among patients with a change in the score of 0.30 to 0.50 and > 0.50 per year was significantly higher than those with a change in the score of 0 to 0.15 per year after 5 years of follow‐up.
Table 2.
Incidence and hazard ratios of adapted DCSI change for ACS, ischemic stroke, and death in diabetes cohort
| Change in Adapted DCSI per Year | N | No. of Events | Ratea | Crude HR | 95% CI | Adjusted HRb | 95% CI |
|---|---|---|---|---|---|---|---|
| ACS | |||||||
| <0.15 | 43 468 | 914 | 4.06 | 1 | (Reference) | 1 | (Reference) |
| 0.15–0.30 | 13 634 | 398 | 3.57 | 0.83 | (0.74‐0.94)c | 0.75 | (0.67–0.85)d |
| 0.30–0.50 | 11 272 | 658 | 7.86 | 1.87 | (1.69–2.06)d | 1.49 | (1.35–1.65)d |
| >0.50 | 16 076 | 2145 | 25.7 | 6.31 | (5.84–6.82)d | 4.53 | (4.18–4.90)d |
| P for trend | <0.001 | <0.001 | |||||
| Ischemic stroke | |||||||
| <0.15 | 43 468 | 481 | 2.12 | 1 | (Reference) | 1 | (Reference) |
| 0.15–0.30 | 13 634 | 365 | 3.28 | 0.83 | (0.74–0.94)c | 1.35 | (1.18–1.55)d |
| 0.30–0.50 | 11 272 | 832 | 10.0 | 1.87 | (1.69–2.06)d | 3.77 | (3.37–4.22)d |
| >0.50 | 16 076 | 3768 | 49.7 | 6.31 | (5.84–6.82)d | 17.9 | (16.2–19.7)d |
| P for trend | <0.001 | <0.001 | |||||
| Death | |||||||
| <0.15 | 43 468 | 2333 | 10.2 | 1 | (Reference) | 1 | (Reference) |
| 0.15–0.30 | 13 634 | 655 | 5.83 | 0.54 | (0.49–0.59)d | 0.47 | (0.44–0.52)d |
| 0.30–0.50 | 11 272 | 1002 | 11.8 | 1.10 | (1.02–1.18)e | 0.81 | (0.75–0.87)d |
| >0.50 | 16 076 | 3190 | 35.4 | 3.41 | (3.23–3.59 d | 2.13 | (2.01–2.25)d |
| P for trend | <0.001 | <0.001 | |||||
| Follow‐up period ≦ 5 years | |||||||
| Death | |||||||
| <0.15 | 22 060 | 1755 | 35.5 | 1 | (Reference) | 1 | (Reference) |
| 0.15–0.30 | 2272 | 169 | 17.9 | 0.39 | (0.33–0.45) d | 0.37 | (0.31–0.43)d |
| 0.30–0.50 | 2569 | 307 | 35.2 | 0.87 | (0.77–0.98)e | 0.71 | (0.63–0.80)d |
| >0.50 | 7775 | 1784 | 93.4 | 2.62 | (2.45–2.80)d | 1.77 | (1.66–1.90)d |
| P for trend | <0.001 | <0.001 | |||||
| Follow‐up period >5 years | |||||||
| Death | |||||||
| <0.15 | 21 408 | 578 | 3.23 | 1 | (Reference) | 1 | (Reference) |
| 0.15–0.30 | 11 362 | 486 | 4.73 | 1.27 | (1.13–1.44d | 1.08 | (0.96–1.22) |
| 0.30–0.50 | 8703 | 695 | 9.08 | 2.56 | (2.29–2.86)d | 1.82 | (1.63–2.04)d |
| >0.50 | 8301 | 1406 | 19.8 | 5.79 | (5.26–6.38)d | 3.68 | (3.33–4.06)d |
| P for trend | <0.001 | <0.001 |
Abbreviations: ACS, acute coronary syndrome; CI = confidence interval; Crude HR, crude hazard ratio; DCSI, Diabetes Complications Severity Index
Rate per 1000 person‐years.
Adjusted for age, sex, comorbidities of hypertension, chronic obstructive pulmonary disease, and hyperlipidemia.
p<0.01
p<0.001
p<0.05
To compare the discriminative ability of the progression of adapted DCSI for ACS, ischemic stroke, and death, the sensitivity, specificity, the C statistic of the logistic model, the Harrell's C statistic of Cox model, and the optimal cutoff points for ACS, ischemic stroke, and all‐cause mortality using change in adapted DCSI are shown in Table 3. Among type 2 DM patients, the C statistic of the logistic model of the progression of adapted DCSI predictive of ACS, ischemic stroke, and death were 0.72 (95% CI: 0.71–0.73) (Figure 2A), 0.84 (95% CI: 0.84–0.85) (Figure 2B), and 0.66 (95% CI: 0.65–0.67) (Figure 2C), respectively. The progression of adapted DCSI had moderate discrimination for ACS, ischemic stroke, and death (C statistics = 0.71, 0.72, 0.75, respectively) based on Cox regression analysis (Harrell C). The optimum threshold of the progression of adapted DCSI for ACS, ischemic stroke, and death in Type 2 DM patients were 0.30, 0.36, and 0.39, respectively.
Table 3.
Sensitivity, specificity, optimum threshold, and predictive ability (C statistics and the 95% confidence interval) for adapted DCSI change correlated to ACS, ischemic stroke, and mortality in type 2 diabetes mellitus patients
| Sensitivity | Specificity | Optimum Threshold | C Statistic (95% CI) of Logistic Model | Harrell's C Statistic of Cox Model | P Value | |
|---|---|---|---|---|---|---|
| ACS, adapted DCSI change | 0.68 | 0.69 | 0.30 | 0.72 (0.71–0.73) | 0.71 | <0.001 |
| Ischemic stroke, adapted DCSI change | 0.81 | 0.76 | 0.36 | 0.84 (0.84–0.85) | 0.72 | <0.001 |
| Death, adapted DCSI change | 0.52 | 0.77 | 0.39 | 0.66 (0.65–0.67) | 0.75 | <0.001 |
Abbreviations: ACS, acute coronary syndrome; CI, confidence interval; DCSI, Diabetes Complications Severity Index
Figure 2.
Receiver operating characteristic (ROC) curve for adapted Diabetes Complications Severity Index change in predicting acute coronary syndrome (A), ischemic stroke (B), and death (C) among type 2 diabetic patients
4. DISCUSSION
This is a study devised to assess the possible advantages of using the progression of adapted DCSI for ACS, ischemic stroke, and mortality prediction in Asian patients affected by type 2 DM. The 84 450 study patients were opportunely chosen from a Taiwanese national registry and followed until occurrence of the end points. After adjustments for some variables, the higher stroke rate was observed with a change in the score of >0.5 per year. An optimum threshold of 0.30, 0.36, and 0.39 was chosen as discriminant of ACS, ischemic stroke, and mortality incidence among type 2 DM patients, respectively. Moreover, the acceptable predictive capability of the progression of the score makes it a possible predictor of adverse events in these patients.
The major novelty of this article is the predominantly Asian population investigated, in which the dynamic change of a preexisting score has been validated. The sample size is very large, the follow‐up duration is reasonable, and the study covers an important topic—the gap in adequate cerebro‐cardiovascular adverse outcomes risk prediction models in Asian cohorts, making our data potentially clinically relevant and important.
Several risk prediction models for cardiovascular events and death in patients with type 2 DM have been previously described.12, 13, 14, 15, 16 The adapted DCSI, a modified version of the original DCSI score, has already been validated in predicting mortality in diabetic subjects and has been applied in some Asian populations.17 Nonetheless, the role of the above score for stratifying cerebrovascular and cardiovascular events have never been explored. Our study, based on a large longitudinal cohort, presents results from a cultural background and validated the changes in the adapted DCSI over time for clinical outcomes prediction, which goes beyond earlier findings specifically developed to predict mortality in diabetic subjects.
Because this score captures several major determinants for ACS, ischemic stroke, and mortality, some might argue that any score that captures some of these determinants would show similar results. Furthermore, other risk tools for cardiovascular events and death were also recommended for this particular type of patient.12, 13, 14, 15, 16 It is quite possible that another scoring scheme may outperform the above model in this unique patient population. The purpose of the study was to validate and expand the current risk prediction model of the adapted DCSI and changes in the score over time for ACS, ischemic stroke, and death in this particular population. With this risk tool, the recommended actions for physicians to follow are clear and helpful in clinical practice, especially while taking care of diabetic patients with dynamic changes of the adapted DCSI of 0.3 or greater per year, so that the optimal screening and preventive therapies can be initiated early.
4.1. Limitations
Despite the work strengths, several limitations of the study should be noted. First, this study and the prediction model deals specifically with an Asian population, and the applicability of the model in other ethnic populations is uncertain. Second, by using the entire cohort to derive the prediction model, the model's generalizability might be restricted due to the lack of a validation cohort despite the ROC analysis giving an impressive picture of the predictive value of the scoring scheme. Third, due to the limitations of this nationwide dataset, comparing various specific cardiovascular risk calculators, such as Framingham or American College of Cardiology/American Heart Association, with the progression of adapted DCSI for projecting cardiovascular outcomes were not attempted. Finally, no data are provided on glucose control, glycohemoglobin, smoking, alcohol consumption, and other lifestyle behaviors, which might limit the interpretation of the risks of adverse cardiovascular and cerebrovascular disease in this population.
5. CONCLUSION
We described the acceptable predictive performance of the progression of adapted DCSI for Asian type 2 DM patients in a large cohort in Taiwan.
Conflicts of interest
The authors declare no potential conflicts of interest.
Hu W‐S, Lin C‐L. Use of the progression of adapted Diabetes Complications Severity Index to predict acute coronary syndrome, ischemic stroke, and mortality in Asian patients with type 2 diabetes mellitus: A nationwide cohort investigation. Clin Cardiol. 2018;41:1038–1043. 10.1002/clc.22991
Funding information This study was supported in part by the Taiwan Ministry of Health and Welfare Clinical Trial and Research Center of Excellence (MOHW106‐TDU‐B‐212‐113 004).
REFERENCES
- 1. International Diabetes Federation . Key findings. http://www.idf.org. Accessed February 2, 2014.
- 2. Martín‐Timón I, Sevillano‐Collantes C, Segura‐Galindo A, Del Cañizo‐Gómez FJ. Type 2 diabetes and cardiovascular disease: have all risk factors the same strength? World J Diabetes. 2014;5:444–470. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Geiss LS, Herman WH, Smith PJ. Mortality in non‐insulin‐dependent diabetes In: Harris MI, Cowie CC, Stern MP, et al, eds. Diabetes in America. 2nd ed Washington, DC: National Diabetes Data Group, National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases; 1995:233–257. [Google Scholar]
- 4. World Health Organization . Diabetes. http://www.who.int/en/news-room/fact-sheets/detail/diabetes.
- 5. Roglic G, Unwin N, Bennett PH, et al. The burden of mortality attributable to diabetes: realistic estimates for the year 2000. Diabetes Care. 2005;28:2130–2135. [DOI] [PubMed] [Google Scholar]
- 6. Miettinen H, Lehto S, Salomaa V, et al. Impact of diabetes on mortality after the first myocardial infarction. The FINMONICA Myocardial Infarction Register Study Group . Diabetes Care. 1998;21:69–75. [DOI] [PubMed] [Google Scholar]
- 7. Young BA, Lin E, Von Korff M, et al. Diabetes complications severity index and risk of mortality, hospitalization, and healthcare utilization. Am J Manag Care. 2008;14:15–23. [PMC free article] [PubMed] [Google Scholar]
- 8. Chang HY, Weiner JP, Richards TM, Bleich SN, Segal JB. Validating the adapted Diabetes Complications Severity Index in claims data. Am J Manag Care. 2012;18:721–726. [PubMed] [Google Scholar]
- 9. National Health Insurance Research Database , Taiwan. http://nhird.nhri.org.tw/en/index.html. Accessed May 10, 2017.
- 10. Hu WS, Lin CL. Association between cataract and risk of incident atrial fibrillation: a nationwide population‐based retrospective cohort study. Mayo Clin Proc. 2017;92:370–375. [DOI] [PubMed] [Google Scholar]
- 11. Lin SY, Chen DC, Lin CL, et al. Risk of acute coronary syndrome in patients with cervical spondylosis. Atherosclerosis. 2018;271:136–141. [DOI] [PubMed] [Google Scholar]
- 12. Stevens RJ, Kothari V, Adler AI, Stratton IM. The UKPDS risk engine: a model for the risk of coronary heart disease in type II diabetes (UKPDS 56). Clin Sci (Lond) 2001;101:671–679. [PubMed] [Google Scholar]
- 13. Selby JV, Karter AJ, Ackerson LM, Ferrara A, Liu J. Developing a prediction rule from automated clinical databases to identify high‐risk patients in a large population with diabetes. Diabetes Care. 2001;24:1547–1555 [DOI] [PubMed] [Google Scholar]
- 14. Clarke PM, Gray AM, Briggs A, et al. A model to estimate the lifetime health outcomes of patients with type 2 diabetes: the United Kingdom Prospective Diabetes Study (UKPDS) Outcomes Model (UKPDS no. 68). Diabetologia. 2004;47:1747–1759 [DOI] [PubMed] [Google Scholar]
- 15. Wells BJ, Jain A, Arrigain S, Yu C, Rosenkrans WA Jr, Kattan MW. Predicting 6‐year mortality risk in patients with type 2 diabetes. Diabetes Care. 2008;31:2301–2306. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Yang X, So WY, Tong PC, et al. Development and validation of an all‐cause mortality risk score in type 2 diabetes. Arch Intern Med. 2008;168:451–457. [DOI] [PubMed] [Google Scholar]
- 17. Wu CX, Tan WS, Toh MPHS, Heng BH. Derivation and validation of a risk index to predict all‐cause mortality in type 2 diabetes mellitus. J Endocrinol Metab. 2012;2:88–95. [Google Scholar]