A Study of Morbidity and Mortality related to Diabetes Mellitus Type 2 in Taiwan

1. A Study of Morbidity and Mortality related to Diabetes Mellitus Type 2 in Taiwan Fourteenth International Longevity Risk and Capital Markets Solutions Conference Sep 21, 2018 Hsin-Chung Wang, Department of Finance and Actuarial Science, Aletheia University, Taiwan. Jack C. Yue, Department of Statistics, National Chengchi University, Taiwan. I-Han Wang, Department of Risk Management and Insurance, National Chengchi University, Taiwan.

2. Outline • Motivation • Introduction to Taiwan's National Health Insurance research databases (NHIRD) • one million Longitudinal Health Insurance Database 2005 and one million elderly Longitudinal Health Insurance Database 2005 • Diagnosed with Type II Diabetes (Definition) • Exploratory Data Analysis (EDA): Trend of Incidence, Prevalence and Mortality rates • Modelling • Conclusion

3. Motivation • Crisis of population ageing: • About 89% of older adults (65+) have a least one chronic disease. • The proportion of elderly people with more than three chronic diseases was as high as 50%. • More than 56% of all healthcare costs are due to chronic diseases. • The top three chronic diseases in the elderly over 65 years old were: hypertension (52.3%), cataract (41.3%), and diabetes (24.2%). (Source: 2007 and 2018 Taiwan Longitudinal Study on Aging Survey Report ,Health Promotion Administration, Ministry of Health and Welfare)

4. Motivation • International Diabetes Federation( IDF) Diabetes Atlas 2017: • More than one-third of diabetes cases are estimated to result from population growth and ageing • 8.8% of adults( 20-79 years) are estimated to have diabetes. • In high income countries, approximately 87% to 91% of all people with diabetes are estimated to have type 2 diabetes. • Globally diabetes results in USD 727 billion being spent yearly by people with diabetes only on healthcare, which corresponds to one for every eight dollars spent on healthcare.

5. Motivation • The diabetes prevalence in women is expected to increase to 9.7% in women and to 10.0% in men • The age group 65-79 years shows the highest diabetes prevalence in both women and men.

6. Motivation • National Diabetes Statistics Report, 2017 • The percentage of adults with diabetes increased with age, reaching a high of 25.2% among those aged 65 years • 23.8% of people with diabetes are undiagnosed( about • Diabetes was the seventh leading cause of death in the United States in 2015 • The total direct and indirect estimated cost of diagnosed diabetes in the United States in 2012 was $245 billion. • Average medical expenditures among people with diagnosed diabetes were about 2.3 times higher than expenditures for people without diabetes

7. Motivation Percentage of incident ESRD patients with diabetes as the primary cause of ESRD, by country, 2015 • 2017 United States Renal Data System (USRDS) Annual Data Report • Taiwan, Japan, and the U.S had the highest reported prevalence of treated End-Stage Renal Disease(ESRD)in 2015 • 45% of kidney dialysis patients are caused by diabetes (Taiwan)

8. Motivation • NCD-RisC, 2016 • The age-standardised diabetes prevalence of China, South Korea, Japan, Taiwan, and U.S. male patients from 1980 to 2014. The grow rates are 179%,132%,46%, 67%, and 45% separately. • Taiwan has grown faster than that in Japan and U.S., and become the third highest in East Asia, second only to that in China and South Korea

9. Motivation • NCD-RisC, 2016

10. Motivation • Prevalence of Hyperglycemia by Gender and Age in 2007 • The hyperglycemia may be regard as an early symptom of diabetes Note: Hyperglycemia is defined as blood glucose test value ≧126mg/dl after 8 or more hours on an empty stomach, or use of hypoglycemic agents. Source: 2007 Taiwanese Survey on Hypertension, Hyperglycemia, Hyperlipidemia

11. Motivation • According to the data released by Ministry of Health and Welfare • The position of diabetes in the male and female causes of death rankings in our country has increased from 12th and 9th in 1981 to 6th and 3rd places in 2016. • The crude death rate of diabetes per 100,000 population dropped from 44.7 in 2005 to 34.9 in 2010 and then rose to 42.8 in 2016 for men. • It fell from 47.8 in 2005 to 36.1 in 2010 and then rose to 41.9 in 2016 for women. Source: Ministry of Health and Welfare

12. Motivation • Mortality rate of diabetes mellitus in Taiwan from 2005 to 2016 (per 100,000 people)

13. Motivation • In recent years, life expectancy has continued to grow in Taiwan. Population ageing has become more apparent, and people have paid more attention to life planning after retirement. • The annual medical expenses for the elderly (65+)are five times that of Taiwanese people aged 0-64. • We expect that the loss of life and medical expenses caused by diabetes and its complications will increase the financial burden on the community.

14. NHIRD • Taiwan launched a single-payer National Health Insurance program on March 1, 1995. As of 2014, 99.9% of Taiwan’s population were enrolled. • There are approximately 25.68 million individuals in this registry • Longitudinal Health Insurance Database 2005 (LHID2005) and The elderly (65+) Longitudinal Health Insurance Database 2005 • In this study, we used inpatient and outpatient claims dataset collected during 1996–2013. • Both data sets are one-million random samples of Taiwan people, one for the group of ages 0~99 and the other for ages 65~99, accounting for about 4.6% and 45.7% of Taiwan’s populations in each age groups.

15. Definition • Claims datasets from 2003 to 2013 were searched to identify any outpatient visit or inpatient admission with diabetes as one of the diagnoses (International Classification of Diseases, 9th Revision, Clinical Modification, ICD-9-CM code 250). • Classification 1: Patients with type 2 diabetic(ICD code 250) but who were type 1 diabetic (ICD code 250.×1, 250.×3) were excluded from this study. • Classification 2: Patients in the classification 1 and who takes refillable ( continuous ) prescription for chronic illnesses.

16. Definition • The age-specific incidence rate of Type II Diabetes: • = , where = the age groups; = number of new diabetes diagnosed in age group ( i.e. who have a diabetes medical record in the previous years must be taken away from the total number of cancer in age group ); = person of the Taiwan’s population deduct who have the diabetes medical record in the previous year in age group. • The age-specific prevalence rate of Type II Diabetes: • = , where = the age groups; = number were found to have diabetes diagnosed in age group ; = person in group of the Taiwan’s population. • The diabetes mortality=death number of diabetes divided by the Taiwan’s population in age group.

17. Definition • There are different problems of diabetes prevalence and incidence researches in previous studies: • using interview survey • observation time is too short • insufficient samples • the different criteria for diabetes identification • the high proportion of diabetes patients with treatment dropout • We use twoone million large samples to reduce small sample size problems and by means of conduct exploratory data analysis (EDA) to evaluate if the observation periods, the number of outpatient visits and the diabetes classification definition are appropriate reduce the problems.

18. Exploratory Data Analysis (EDA) • The observation periods: one-year, two-year, or three-year diabetes-free observation period (For example: 2012 year Taiwan Female) • The longer the observation period The lower incidence rate

19. Exploratory Data Analysis (EDA) • The observation periods: one-year, two-year, or three-year diabetes-free observation period (For example: 2012 year Taiwan Female 65+) • There are about 80% new diabetes people in two-year observation periods still in one-year observation periods

20. Exploratory Data Analysis (EDA) • Underwriting needs: The insured only required to submit a two-year health report (Taiwan) • We choose two year as the observation period in this study. • Lin et al. (2005) shows that : • The accuracy of diabetes diagnosis in NHI claims data in Taiwan • was 74.6%. • The accuracy of cases with 4 outpatient visits was 96.1% • The number of outpatient visits was the factor most associated with diagnostic accuracy.

21. Exploratory Data Analysis (EDA) • Prevalence rates : 4 Outpatient Visits Per Year • The 5- year average prevalence of Type 2 diabetes increased from 2.3% ( age 45-49) to 15.9% ( age 75-79), and then decreased to 5.6% for women.

22. Exploratory Data Analysis (EDA) • Incidence rates Comparison : 4 Outpatient Visits Per Year Vs. First Time Takes Refillable Prescription •  The incidence rates of these two judging criteria are almost equal. Ps: Lin C.C., Lai M.S., Syu C.Y., Chang S.C., Tseng F.Y., (2005), “Accuracy of diabetes diagnosis in health insurance claims data in Taiwan”, J Formos Med Assoc 104: 157–163.

23. Methodology • New judging criteria of incidence rates : • Classification 3: Patients no diabetes claim records (ICD code 250 ) in the previous 2 years and who first time takes refillable ( continuous ) prescription as the new diabetics. • LC model: the logarithm of central mortality rate satisfies with and , are age related parameters ( ), and represents the time related parameter. Note that is the general mortality level, is the decline in mortality at age x, and is usually a linear function in time.

24. Methodology • The partial SMR: is one way to deal with estimating mortality rates of small populations, by adding information from other (large) population to correct possible bias - was the estimated value of the heterogeneity parameter via:

25. Trend of Incidence Rate • LC model has good fitness ability. • The 11-year average incidence rate of Type 2 diabetes increased from 0.42%( age 45-49) to 1.79% ( age 70-79), and then decreased to 0.54% for women. • The 11-year average incidence rate of Type 2 diabetes increased from 0.73%( age 45-49) to 1.65% ( age 65-69), and then decreased to 0.64% for men.

26. Trend of Incidence Rate • For example: 2012 year • Smoothing the incidence rates (or graduation) is often necessary to reduce the fluctuations between ages. • Graduation methods: LC model, PSMR, or PSMR+LC

27. Trend of Incidence Rate • single age : for example 2012 year • Both male and female in the incidence rates have an increasing phenomenon yearly.

28. Trend of Incidence Rate • Forecast: •  Greater the increase in higher age

29. To judge elder diabetes patients whether dead : • Condition 1:Withdrawal in Registry for beneficiaries dataset (ID) and no outpatient records for two consecutive years • Condition 2: When the last outpatient record of the year is emergency, and no outpatient records for two consecutive years in the future. • Condition 3: last inpatient records is more than 30 days and no outpatient records for two consecutive years in the future. • Condition 4: the Catastrophic Illnesses (CI) patients and no outpatient records for two consecutive years in the future. • Condition 5: The code is death, suicide, and discharged under critical condition and no outpatient records for two consecutive years in the future in Inpatient expenditures by admissions dataset (DD) • Condition 6: The code is death in Catastrophic illness dataset (HV) and no outpatient records for two consecutive years in the future Mortality Rate

30. Elderly Mortality Rate • Taiwan age-specific diabetes mortality rates (Observed 4 Outpatient Visits Per Year ) • The mortality rates are increasing with age increases. • The mortality rates have the fluctuation phenomenon with a stable path from 2006 to 2011

31. Conclusion • The incidence rates of Type II diabetes is gradually increasing by year. • The prevalence of type 2 diabetes patients declines with ageafter age 75, but it increased noticeably by year. • The mortality rates of diabetes change with a stable path. • The adult diabetic population will have a significant increase in Taiwan • The loss of life and medical expenses caused by diabetes and its complications will increase the financial burden on individuals and countries.

32. Conclusion • In this study, we found out that if we consider those patients who have an at least 2 years diabetes-free observation period and at least 4 outpatient claims with the diagnosis of diabetes (or one time continuous prescription for chronic diseases) as the definition of incidence, the diabetes incidence would be the most stable. • In the future, we will explore how these research results can be used to design insurance products associated with diabetes, which can help the insured and their families to face the consequence.

33. Thank you for your attention.