Age-period-cohort Model Research Articles

Nationwide trends and incidence of blindness in patients with diabetic retinopathy identified using an age-period-cohort analysis.

To estimate the epidemiologic trends of blindness in patients with diabetic retinopathy (DR) and investigate the age-, period-, and cohort-specific effects on blindness. A total of 2.5 million patients with DR aged 20 years or older were included from the Korean National Health Claims database from 2005 to 2019. Non-proliferative DR/ proliferative DR (NPDR/PDR) cohorts were constructed separately. Participants were identified as having a blindness in at least one eye. The incidence of blindness were calculated using a log-linear Poisson age-period-cohort (APC) analysis model, each effect on blindness were estimated for each study group. The incidence of blindness was 1326.62 per 100,000 in the NPDR group and 3397.57 in the PDR group. The blindness rate sharply decreased after 2011, with annual decreases of 5.6% and 4.4% in the NPDR and the PDR groups, respectively. People born between 1920 and 1930 had the highest overall risk of blindness, with the risk decreasing rapidly after that. For those born after 1980, the risk started to increase in both sexes. Among the APC models, the combination model of age, period, and cohort effects showed the highest explanatory power (0.96). In this nationwide long-term study, blindness in DR was not due to a single epidemiologic cause but rather a combination of biological age, social determinants, and healthcare policies. The increased risk of blindness in individuals in their 20 s and 30 s may even increase in the future and should not be ignored. Therefore, vigilance of younger patients is recommended.

Gender and Education Differences in Mental Health in Belgium: a Bayesian Age-Period-Cohort Analysis

Abstract Background Mental illnesses (MI) constitute a growing public health problem, with reports of a higher prevalence among middle-aged adults (age effect), during specific periods such as economic crisis (period effect) and in recent birth cohorts (cohort effect). However, few studies have assessed whether these age-period-cohort (APC) effects were related to general population trends or if they are driven by changes in the mental health status of more vulnerable groups. Therefore, this study aimed to assess gender and educational differences underlying observed APC effects in the prevalence of MI in the Belgian adult population. Methods This study used data from six cross-sectional waves of the Belgian Health Interview Survey (1997-2018, ∼ 10,000 respondents per wave). Mental health was assessed using the 12-item General Health Questionnaire. Bayesian hierarchical APC models were applied on the prevalence of MI, taking into account gender and educational differences, as well as some potential confounding factors. Results Overall, this study confirmed the higher prevalence of MI in middle-aged adult, in 2013, and in more recent birth cohorts (1980-2000). Accounting for the interaction between gender and education, we found that the MI educational gradient was more pronounced for women than men. Finally, we found that APC trends were different between gender. The odds of having MI tended to be stable across age groups for women, while men showed an increased odds between 50-60 years and a decrease between 60-65 years. Regarding period and cohort effects, the increase odds in 2013 was mainly present in men, while the increase in recent birth cohorts was mainly present in women. Conclusions This study emphasizes that the age-period-cohort effects on mental health are different depending on gender and education, thus enriching our understanding on the underlying mechanisms to better guide future public (mental) health policies. Key messages • This study shows the importance of studying not only the overall trends in mental illness, but also the underlying differences in its distribution. • Insights into gender and educational differences in Age-Period-Cohort effects on mental health is crucial for the development of timely and efficient public health interventions.

Trends and age-period-cohort analysis of leukemia incidence in Qidong from 1972 to 2021

Objective: To describe the epidemiological characteristics and trends of leukemia incidence in Qidong between 1972 and 2021, and provide guidelines for prevention and control measures and strategies. Methods: The cancer registry data was collected and analyzed on leukemia incidence during 1972-2021 in Qidong by sex, age and time. Crude incidence rate (CR), China age-standardized rate (ASRC), world age-standardized rate (ASRW), and average annual change percentage (AAPC) was calculated by Joinpoint software. Age-period-cohort (APC) model was used to analyze the influence of age, period and birth cohort on the changes in the incidence trend of leukemia patients. Results: From 1972 to 2021, there were 2 948 patients with leukemia in Qidong, accounting for 2.00% of all cancer new cases, CR of leukemia was 5.26/105, ASRC was 4.34/105, ASRW was 4.35/105. The truncated incidence of 35-64 years old was 5.29/105, the cumulative incidence rate between the ages of 0 and 74 years old was 0.40%, the cumulative risk was 0.40%. There were 1 608 male patients, the CR, ASRC, and the ASRW were 5.81/105, 4.88/105 and 4.85/105. The number of female patients were 1 340, and the CR, ASRC, and the ASRW were 4.71/105, 3.86/105 and 3.91/105, respectively. Temporal trends indicated significant upward trends in ASRC among both gender, males and females with AAPC values of 1.41% (P＜0.001), 1.15% (P＜0.001), and 1.73% (P＜0.001), respectively. The results of the APC model showed that the average net drift value of leukemia incidence in all age groups was 1.57% (95% CI, 1.24%-1.89%), and the highest value of local drift was 3.20% (95% CI, 1.63%-4.78%) in the 80~ years old group. The incidence of leukemia increased with age. With the passage of time, the risk of leukemia incidence increased gradually compared with the rate ratio of leukemia incidence (risk ratio [RR], 1.00) in 1992-1996, the RR of leukemia incidence increased from 0.70 during 1972-1976 to 1.57 during 2017-2021. The later the cohort was born, the greater the risk of leukemia incidence compared with the relative risk of leukemia incidence (RR, 1.00) in 1952-1956 cohort, the RR of leukemia incidence increased from 0.24 in the 1892-1896 cohort to 2.73 in the 2017-2021 cohort. Conclusions: The incidence of the leukemia has presented a rising trend in the past fifty years. Leukemia incidence increased with age, and the period and cohort effects on the risk of incidence increase. Further research is needed to investigate the risk factors related to leukemia.

Dynamic trends of ischemic heart disease mortality attributable to high low-density lipoprotein cholesterol: a joinpoint analysis and age-period-cohort analysis with predictions

AimsThe purpose of this study was to analyze the dynamic trends of ischemic heart disease (IHD) mortality attributable to high low-density lipoprotein cholesterol (LDL-C).MethodsData on IHD mortality attributable to high LDL-C from 1990 to 2021 were extracted from the global disease burden database. Joinpoint software was used to estimate the average annual percentage change (AAPC) in the age-standardized mortality rate (ASMR). An age‒period‒cohort model was used to analyze the impacts of age, period, and cohort on these changes. The Bayesian framework was used to predict IHD mortality attributable to high LDL-C from 2022 to 2040.ResultsThe overall ASMR of IHD attributable to high LDL-C decreased from 50. 479 per 100,000 people in 1990 to 32.286 per 100,000 people in 2021, and ASMR of IHD attributable to high LDL-C was higher in males than in females. The longitudinal age curves of the overall IHD mortality attributable to high LDL-C showed a monotonic upward trend, especially after 65 years of age. The period and cohort effect relative risk (RR) values of overall IHD mortality attributable to high LDL-C showed a downward trend. The overall ASMR of IHD attributable to high LDL-C is predicted to show a downward trend, and male IHD mortality attributable to high LDL-C is expected to be higher than that of females.ConclusionThis study revealed a sustained decrease in IHD mortality attributable to high LDL-C over three decades, with a continued decline expected. Despite this, gender disparities persist, with males experiencing higher mortality rates and elderly individuals remaining a vulnerable group.

Burden of non-communicable diseases attributable to high temperature in a changing climate from 1990 to 2019: a global analysis

BackgroundWith global climate change, the health threats of ambient high temperature have received widespread attention. However, latest spatio-temporal patterns of the non-communicable diseases (NCDs) burden attributable to high temperature have not been systematically reported. We aimed to analyze vulnerable areas and populations based on a detailed profile for the NCDs burden attributable to high temperature globally.MethodsWe obtained data from the Global Burden of Diseases (GBD) Study (2019) to describe the temporal and spatial patterns of NCDs burden attributable to high temperature globally from 1990–2019. Then we analyzed the differences by region, sex, and socio-demographic index (SDI). Finally, the age‑period‑cohort (APC) model was utilized to explore the age, period, and cohort effects of NCDs mortality caused by high temperature.ResultsIn 2019, the number of deaths and Disability-adjusted life years (DALYs) from high-temperature-related NCDs was about 150,000 and 3.4 million globally, of which about 70% were in South Asia and North Africa and Middle East, and the burden was higher in men. Among 204 countries and territories, the highest age-standardized mortality rate (ASMR) and age-standardized DALY rate (ASDR) were observed in Oman and United Arab Emirates, respectively. The global burden showed an upward trend from 1990 to 2019, with an EAPC of 3.66 (95%CI: 3.14–4.18) for ASMR and 3.68 (95%CI: 3.16-4.21) for ASDR. Cardiovascular diseases were the main contributors to the global burden of high-temperature-related NCDs in 2019. The age and period effect in APC model showed an increasing trend globally. There was a significant negative correlation between SDI and both ASMR (r = -0.17) and ASDR (r = -0.20) from 1990 to 2019.ConclusionThere was an increasing trend of the global burden of high-temperature-related NCDs. The burden was likely to be higher in males and the elderly, as well as in countries and regions with less economically and socially developed and in tropical climates. Surveillance and prevention measures should be implemented with a focus on these vulnerable areas and susceptible populations.

Tracking multidrug resistant tuberculosis: a 30-year analysis of global, regional, and national trends.

To provide valuable insights for targeted interventions and resource allocation, our analysis delved into the multifaceted burden, trends, risks, and projections of multi drug resistant tuberculosis (MDR-TB). This research employed data from the Global Burden of Disease (GBD) 2019 dataset, which used a comparative risk assessment to quantify the disease burden resulting from risk factors. Initially, this database was utilized to extract details concerning the disability-adjusted life years (DALYs), mortality, incidence, and the number of individuals afflicted by MDR-TB. Subsequently, regression analyses were conducted using the Joinpoint program to figure average annual percent change (AAPC) to ascertain the trend. Thirdly, the age-period-cohort model (APCM) was adopted to analyze evolutions in incidence and mortality. Finally, utilizing the Nordpred model within R software, we projected the incidence and mortality of MDR-TB from 2020 to 2030. MDR-TB remained a pressing global health concern in regions with lower socio-demographic indexes (SDI), where the AAPC in DALYs topped 7% from 1990 to 2019. In 2019, the cumulative DALYs attributed to MDR-TB tallied up to 4.2 million, with India, the Russian Federation, and China bearing the brunt. Notably, the incidence rates have shown a steadfast presence over the past decade, and a troubling forecast predicts an uptick in these areas from 2020 to 2030. Additionally, the risk of contracting MDR-TB grew with advancing age, manifesting most acutely among men aged 40+ in lower SDI regions. Strikingly, alcohol consumption had been identified as a significant contributor, surpassing the impacts of smoking and high fasting plasma glucose, leading to 0.7 million DALYs in 2019. A robust strategy is needed to end tuberculosis (TB) by 2030, especially in lower SDI areas.

Global temporal trends and projections of acute hepatitis E incidence among women of childbearing age: Age-period-cohort analysis 2021

Acute Hepatitis E (AHE) poses a significant threat to global public health, particularly among women of childbearing age (WCBA), who are at heightened risk for severe pregnancy-related complications. This study aimed to delineate the temporal trends and project future incidence of AHE in WCBA, providing insights crucial for targeted prevention and control strategies. Data on AHE incidence from the Global Health data 2021. The Age-Period-Cohort (APC) model was applied to analyze trends across different age groups, periods, and birth cohorts, and the Bayesian APC model was utilized for forecasting future epidemiological trajectories. Globally, AHE incidence numbers among WCBA rose from 2,831,075 in 1992 to 3,420,786 in 2021, while the age-standardized incidence rate (ASIR) declined from 194.66 to 179.54 per 100,000 with a global net drift of -0.28%. However, high SDI regions showed a contrasting trend with a positive net drift of 0.02%. The age effect was consistent across SDI regions and global, showing a decrease with advancing age, while unfavorable period and cohort effects were exhibited in high-SDI region. At the national level, locations exhibited varying trends of change. The BAPC model predicted a total of 3,759,384 AHE global cases in WCBA by 2030, with an expected mild increase in the ASIR. The outlook for the management and containment of AHE is grim in certain countries, including India. The study revealed a complex epidemiological landscape of AHE in WCBA, with increasing global incidence numbers juxtaposed against a declining ASIR. The AHE burden by 2030 remain severe among WCBA. Young WCBA and high SDI region merit particular attention. The findings underscore the need for region-specific strategies to curb the projected rise in AHE incidence and align with the 2030 WHO goals.

Evolution of Esophageal Cancer Incidence Patterns in Hong Kong, 1992-2021: An Age-Period-Cohort and Decomposition Analysis.

To elucidate the historical trends, underlying causes and future projections of esophageal cancer incidence in Hong Kong. Utilizing the Age-Period-Cohort (APC) model, we analyzed data from the Hong Kong Cancer Registry (1992-2021) and United Nations World Population Prospects 2022 Revision. Age-standardized incidence rates were computed, and APC models evaluated age, period, and cohort effects. Bayesian APC modeling, coupled with decomposition analysis, projected future trends and identified factors influencing incidence. Between 1992 and 2021, both crude and age-standardized incidence rates of esophageal cancer witnessed significant declines. Net drifts exhibited pronounced downward trends for both sexes, with local drift diminishing across all age groups. Period and cohort rate ratios displayed a consistent monotonic decline for both sexes. Projections indicate a continued decline in esophageal cancer incidence. Population decomposition analysis revealed that epidemiological changes offset the increase in esophageal cancer cases due to population growth and aging. The declining trend of esophageal cancer in Hong Kong is influenced by a combination of age, period, and cohort. Sustaining and enhancing these positive trends requires continuous efforts in public health interventions.

Trends and future projections of incidence rate and mortality of rheumatoid arthritis in China: a systematic analysis based on GBD 2021 data.

This study aims to evaluate the trends in rheumatoid arthritis (RA) in China from 1990 to 2021 by analyzing data from the Global Burden of Disease (GBD) 2021 study and to predict the trends for the next 25years. Age-standardized incidence rates (ASIR) and age-standardized mortality rates (ASMR) were calculated, and the estimated annual percentage change was used to illustrate differences in age distribution among various populations. Age-period-cohort (APC) analysis and Bayesian APC (BAPC) models were employed to forecast the burden of RA in China from 2022 to 2046. From 1990 to 2021, the ASIR of RA in China increased from 11.6 to 13.7, with a significantly higher ASIR in females than in males. Despite the increase in incidence, the ASMR related to RA decreased from 0.7 to 0.5. Predictions using the BAPC model indicate that the incidence of RA will continue to rise, with an expected ASIR of approximately 16.4 by 2046, and the total number of RA cases is projected to reach around 342,000. In terms of mortality, the ASMR is expected to decline to 0.3 by 2046, although the total number of deaths might reach about 40,000. The incidence of RA in China has significantly increased over the past 30years. Although the incidence rate and the total number of RA cases may continue to rise in the future, the mortality rate of RA has been consistently declining. Key Points • Over the past 30years, the incidence of RA in China has significantly increased; although the incidence rate and total number of cases may continue to rise, the mortality rate has been consistently declining.

The Transformation of Traditional Peasants: An Age–Period–Cohort Analysis of the Modernity of Peasants in China

Using the data of Chinese General Social Survey from 2010 to 2018, this study measured the modernity of Chinese peasants from four dimensions: equality and autonomy, positive attitude, general confidence, and open-mindedness. Hierarchical age–period–cohort models were introduced to assess the effect of age, period, and cohort on modernity and to further explore the heterogeneities in the effect of education and off-farm work experience on modernity among various birth cohorts. Age effects consistently show significance, with the level of individual modernity decreasing as age increases. However, the rate of decline gradually slows down. Period effects are generally significant and contribute to the advancement of modernity. Cohort effects are significant in most cohorts. The modernity of peasants born between 1941 and 1960 show a consistent decline, reaching its lowest point in the cohort 1961–1970 and then increasing steadily among those born after 1970. Possible explanations were proposed from the perspective of changes in the land property system and household registration system in China. We also found that the positive effect of education is stronger for younger cohorts, whereas the positive effect of off-farm experience is stronger for older cohorts. This study enriches Inkeles’s research on individual modernization and reveals the transformation process of traditional peasants in developing countries.

Expanding our understanding of long-term trends in alcohol abstention and consumption in England (2001-19) using two age-period-cohort approaches.

Alcohol consumption has decreased in England in recent decades, while alcohol-specific death rates have remained relatively stable. Age-period-cohort (APC) models offer the potential for understanding these paradoxical trends. This study aimed to use an APC model approach to measure long-term trends in alcohol abstention and consumption in England from 2001 to 2019. The study used grouped and proxy-variable APC models of repeat cross-sectional survey data, set in England (2001-19). Participants were residents in England aged 13 years or over who took part in the Health Survey for England. Outcome variables were alcohol abstention and consumption in units. We created nine age groups (13-15, 16-17, 18-24, 25-34, until 65-74 and 75+, reference 45-54 years), four periods (2001-04, 2005-09, 2010-14 to 2015-19, reference 2005-09) and 18 5-year birth cohorts (1915-19 to 2000-04, reference 1960-64). We proxied age effects (systolic and diastolic blood pressure), period effects (alcohol affordability, internet usage and household alcohol expenditure) and birth cohort effects (prevalence of smoking and prevalence of overweight). The odds of abstaining were considerably larger at young ages, 13-15 years [odds ratio (OR) = 5.38; 95% confidence interval (CI) = 4.50-6.43], were lowest during the first period, 2001-04 (OR = 0.83; 95% CI = 0.79-0.86) and had a U-shaped pattern by birth cohort. For units of alcohol, the incidence rate ratio (IRR) increased until age 18-24 years (IRR = 1.41, 95% CI = 1.34-1.48) and decreased afterwards, were highest during the first period, 2001-04 (IRR = 1.07; 95% CI = 1.05-1.08) and showed an inverted J-shape by birth cohort. Our proxy variable approach revealed that using blood pressure measures, alcohol affordability and prevalence of overweight as proxies resulted in APC effects that differed from our base-case model. However, internet usage, household expenditure on alcohol and smoking prevalence resulted in APC effects similar to our base-case model. The discrepancy between decreasing alcohol consumption and increasing alcohol-related deaths observed in England from 2001 to 2019 may, in part, be explained by the halt in abstention trends since 2010 and a slight consumption decline since 2001.

Trend analysis and prediction of injury incidence in China from 1990 to 2019 based on Bayesian age–period–cohort model

BackgroundInjury is a major challenge to global public health. Analysing the trend of injury incidence in China from 1990 to 2019 and predicting future trends in incidence can provide a...

Age-period-cohort analysis of epidemiological trends in pelvic fracture in China from 1992 to 2021 and forecasts for 2046.

This study explored the epidemiological trends in pelvic fracture (PF) in China from 1992 to 2021, analyze their relationships with age-period-cohort (APC) factors, and predict the trends of PF from 2022 to 2046. Incidence and years lived with disabilities (YLDs) of PF among sexes in China from 1992 to 2021 were obtained through the 2021 Global Burden of Disease (GBD) database. Trends in the incidence and YLDs were described, and a joinpoint regression model was used. The APC model was used to explore the effects of age, period, and cohort on the incidence and YLDs. Nordpred forecasted the incidence and YLDs in China from 2022 to 2046. In 2021, there were an estimated 0.63 million incidence cases and 0.33 million of YLDs, respectively. The number and age-standardized rate (ASR) of incidence and YLDs were both gradually increased. The average annual percent change (AAPC) in incidence and YLDs for men were 0.26% and -0.17%, respectively. For women, the AAPC values for incidence and YLDs were -0.03% and -0.57% (p < 0. 001), respectively. The relative risk (RR) of PF increases with age, with the lowest risk in those aged 10-14 years for incidence and aged 1-4 for YLDs and the highest risk in those aged >95 years for incidence and aged 90-94 years for YLDs. The period effect showed a totally increase in the risk across the general, male, and female populations. Cohort effects indicated a totally significant decline for both incidence and YLDs. The predicted incidence and YLDs of PF in China from 2022 to 2046 showed an initial rise, followed by a decline, with 2029 and 2034 being the turning point for incidence and YLDs, respectively. The characteristics of pelvic fracture incidence and YLDs in China are complex. Thus, primary prevention measures must be strengthened. Raising awareness about osteoporosis prevention, enhancing public health education, and promoting good dietary and hygiene habits are appropriate preventive measures for PF in China.

A matter of time, not generations: rising emotional attachment to the European Union 1991–2023. An age period cohort analysis

ABSTRACT The literature argues that successive generations have become progressively more attached to the EU, due to having experienced increasing levels of integration during their impressionable adolescent years. This generational view of EU attachment assumes that after birth cohorts have outgrown their impressionable years, they do not become more attached to the EU, regardless of period-based changes. However, newer theories of European integration emphasise period-based influences that shape EU attachment across all age cohorts, rather than solely impacting adolescents during their formative years. We use an age period cohort model with repeated cross-sectional Eurobarometer data from 1991 to 2023, showing that all EU-citizens have on average become more attached to the EU over time, rather than later-born birth cohorts being more attached to the EU based on their generational membership. Contrary to what the literature assumes, this indicates that no ‘generations of Europeans’ exist, but that EU-attachment increases for all cohorts over time. Contrary to long-standing assumptions in the literature, this indicates that efforts to increase emotional attachment to the EU can proceed faster than generational change occurs.

Trends in cancer mortality among the elderly in China, 2005-2035.

With the appearance and aggravation of the aging society, cancer has become one of the major problems that threaten the life and health of Chinese residents seriously. To explore the cancer epidemiological characteristics among the elderly in China from 2005 to 2016, and to provide strategies for cancer prevention and treatment. Stratified analysis was conducted on the cancer mortality data of the elderly aged ≥60 years in China, which were selected from the Chinese Cancer Registry Annual Report. Joinpoint regression model was used to calculted average annual percentage change (AAPC) to estimate the time trends. Age-period-cohort (APC) model was used to explore the age, period and birth cohort effect on the risk of cancer death. Bayesian age-period-cohort (BAPC) model was used to predict trends in cancer mortality among elderly by gender and region to 2035. 2005-2016, cancer mortality in the elderly in China showed a decreasing trend (AAPC = -1.2%, P<0.001). Cancer mortality in rural areas was higher than that in urban areas, but the urban-rural difference gradually narrowed (t = 6.1, P<0.01). The APC model showed that cancer mortality increased with age. The relative risk (RR) for the period effect decreased. RR was lower for the later- born cohort than that for the earlier-born cohort in rural areas. Lung cancer mortality ranked first in both male and female, and showed an increasing trend among female in the 60-64, 80-84 and ≥85 age groups (AAPC60-64 = 1.0%, AAPC80-84 = 0.8%, AAPC≥85 = 2.0%, all P<0.05). By 2035, cancer mortality for the elderly was predicted to decline nationally, by sex and in rural areas, while rising in urban areas. Cancer mortality in the elderly in China showed a decreasing trend from 2005 to 2016, but it was still higher than the world average. Early cancer screening is important, especially in the elderly male and in rural areas.

Global, regional and national time trends in incidence of adverse effects of medical treatment, 1990–2019: an age–period–cohort analysis from the Global Burden of Disease 2019 study

BackgroundCurrent adverse effects of medical treatment (AEMT) incidence estimates rely on limited record reviews and underreporting surveillance systems. This study evaluated global and national longitudinal patterns in AEMT incidence from...

An Age–Period–Cohort Framework for Profit and Profit Volatility Modeling

The greatest source of failure in portfolio analytics is not individual models that perform poorly, but rather an inability to integrate models quantitatively across management functions. The separable components of age–period–cohort models provide a framework for integrated credit risk modeling across an organization. Using a panel data structure, credit risk scores can be integrated with an APC framework using either logistic regression or machine learning. Such APC scores for default, payoff, and other key rates fit naturally into forward-looking cash flow estimates. Given an economic scenario, every applicant at the time of origination can be assigned profit and profit volatility estimates so that underwriting can truly be account-level. This process optimizes the most fallible part of underwriting, which is setting cutoff scores and assigning loan pricing and terms. This article provides a summary of applications of APC models across portfolio management roles, with a description of how to create the models to be directly integrated. As a consequence, cash flow calculations are available for each account, and cutoff scores can be set directly from portfolio financial targets.

Birth cohort effects on diagnosed atrial fibrillation incidence: nationwide cohort study from 1980 to 2018

BackgroundThe incidence of atrial fibrillation (AF) shows substantial temporal trends, but the contribution of birth cohort effects is unknown. These effects refer to the relationship between birth year and the...

Long-term trends in incidence, mortality and burden of liver cancer due to specific etiologies in Hubei Province

Liver cancer, a chronic non-communicable disease, represents a serious public health problem. Long-term trends in the burden of liver cancer disease are heterogeneous across regions. Incidence and mortality of liver cancer, based on the Global Burden of Disease, were collected from the Chinese Centre for Disease Control and Prevention. Age–period–cohort model was utilized to reveal the secular trends and estimate the age, period and cohort effects on primary liver cancer due to specific etiologies. Both the age-standardized incidence and mortality rate of liver cancer in Hubei province were on the rise, although there were discrepancies between gender groups. From age–period–cohort analysis, both incidence and mortality of liver cancer due to Hepatitis B virus were the highest in all age groups. The incidence of all liver cancer groups increased with time period in males, while this upward trend was observed in females only in liver cancer due to alcohol use group. Cohort effects indicated the disease burden of liver cancer decreased with birth cohorts. Local drifts showed that the incidence of liver cancer due to specific etiologies was increasing in the age group of males between 40 and 75 years old. The impact of an aging population will continue in Hubei Province. the disease burden of liver cancer will continue to increase, and personalized prevention policies must be adopted to address these changes.

Discrete-Time Survival Models with Neural Networks for Age–Period–Cohort Analysis of Credit Risk

Survival models have become popular for credit risk estimation. Most current credit risk survival models use an underlying linear model. This is beneficial in terms of interpretability but is restrictive for real-life applications since it cannot discover hidden nonlinearities and interactions within the data. This study uses discrete-time survival models with embedded neural networks as estimators of time to default. This provides flexibility to express nonlinearities and interactions between variables and hence allows for models with better overall model fit. Additionally, the neural networks are used to estimate age–period–cohort (APC) models so that default risk can be decomposed into time components for loan age (maturity), origination (vintage), and environment (e.g., economic, operational, and social effects). These can be built as general models or as local APC models for specific customer segments. The local APC models reveal special conditions for different customer groups. The corresponding APC identification problem is solved by a combination of regularization and fitting the decomposed environment time risk component to macroeconomic data since the environmental risk is expected to have a strong relationship with macroeconomic conditions. Our approach is shown to be effective when tested on a large publicly available US mortgage dataset. This novel framework can be adapted by practitioners in the financial industry to improve modeling, estimation, and assessment of credit risk.