Testing Persistence of Cohort Effects in the Epidemiology of Suicide: an Age-Period-Cohort Hysteresis Model

The Insight Study: A Case-Control Psychological Autopsy Study Examining Factors Associated With Suicide in Middle-Aged People

INTRODUCTION

Despite current suicide prevention strategies, suicide rates are increasing in many Western countries in the 45-60 age group. To comprehend the high risk of suicide in this group, it is essential to gain insight into its risk factors.

METHODS

A case-control psychological autopsy study examined the differences between individuals who died by suicide (i.e., suicide cases) and those with psychological problems (i.e., living controls) between the ages of 45 and 60 years.

RESULTS

Eighty-two informants were interviewed regarding 48 suicide cases and 23 controls. Controls were significantly more likely than suicide cases to have discussed the subject of euthanasia (45.5% vs.15.2%), to experience physical problems (95.7% vs. 70.8%), and to undergo treatment for these problems (73.9% vs. 39.6%). Suicide cases were more likely to be diagnosed with a depressive episode, depressive disorder (87.5% vs. 60.9%), but less likely to have current treatment for psychiatric disorders (70.8% vs. 95.7%). Suicide cases were more likely than controls to experience problems at work (85.4% vs. 60.9%), future financial problems (59.0% vs. 22.2%), and concerns about these problems (56.8% vs. 22.2).

CONCLUSION

The results suggest possibilities for suicide prevention among middle-aged individuals through targeted interventions addressing mental healthcare, occupational settings, and financial distress.

An age-period-cohort approach to studying long-term trends in obesity and overweight in England (1992-2019)

OBJECTIVE

This study aims to understand long-term trends in obesity and overweight in England by estimating life-course transitions as well as historical and birth cohort trends for both children and adults.

METHODS

Data on individuals aged 5 to 85 years old from the Health Survey for England were used, covering the period 1992 to 2019 and birth cohorts born between 1909 and 2013. Individual BMI values were classified as healthy weight, overweight, or obesity. Trends were compared, and an age-period-cohort model was estimated using logistic regression and categorical age, period, and cohort groups.

RESULTS

There was significant variation in age trajectories by birth cohorts for healthy weight and obesity prevalence. The odds of having obesity compared with a healthy weight increased consistently with age, increased throughout the study period (but faster between 1992 and 2001), and were higher for birth cohorts born between 1989 and 2008. The odds of having overweight showed an inverted U-shape among children, increased through adulthood, have been stable since 2012, and were considerably higher for the youngest birth cohort (2009-2013).

CONCLUSIONS

Younger generations with higher overweight prevalence coupled with increasing obesity prevalence with age suggest that obesity should remain a high priority for public health policy makers in England.

Farm fatalities in Northern Ireland agriculture: What fifty years of data tell us

Agriculture is one of the most hazardous sectors in terms of fatal and non-fatal accidents. This paper utilises an administrative dataset that recorded farm fatalities in Northern Ireland over a 50 year timeframe (1968-2017) to undertake an age-period analysis of accident related mortality rates by sex, cause of death, season, and day of the week. Public policies aimed to improve farm safety should consider that fatalities due to animals have increased while the incidence of deaths due to vehicles and equipment has substantially decreased over the years although it is still the primary cause of death. With respect to age, elderly still actively involved in farming and children in the spring and at week-ends are most exposed to the risk of a fatal accident. Overall, men die on farms five times more than women.

APCI: An R and Stata Package for Visualizing and Analyzing Age-Period-Cohort Data

Social scientists have frequently attempted to assess the relative contribution of age, period, and cohort variables to the overall trend in an outcome. We develop an R package APCI (and Stata command apci) to implement the age-period-cohort-interaction (APC-I) model for estimating and testing age, period, and cohort patterns in various types of outcomes for pooled cross-sectional data and multi-cohort panel data. Package APCI also provides a set of functions for visualizing the data and modeling results. We demonstrate the usage of package APCI with empirical data from the Current Population Survey. We show that package APCI provides useful visualization and analytical tools for understanding age, period, and cohort trends in various types of outcomes.

Understanding long-term trends in smoking in England, 1972-2019: an age-period-cohort approach

BACKGROUND AND AIMS

Smoking prevalence has been falling in England for more than 50 years, but remains a prevalent and major public health problem. This study used an age-period-cohort (APC) approach to measure lifecycle, historical and generational patterns of individual smoking behaviour.

DESIGN

APC analysis of repeated cross-sectional smoking prevalence data obtained from three nationally representative surveys.

SETTING

England (1972-2019).

PARTICIPANTS

Individuals aged 18-90 years.

MEASUREMENTS

We studied relative odds of current smoking in relation to age in single years from 18 to 90, 24 groups of 2-year survey periods (1972-73 to 2018-19) and 20 groups of 5-year birth cohorts (1907-11 to 1997-2001). Age and period rates were studied for two groups of birth cohorts: those aged 18-25 years and those aged over 25 years.

FINDINGS

Relative to age 18, the odds of current smoking increased with age until approximately age 25 [odds ratio (OR) = 1.48, 95% confidence interval (CI) = 1.41-1.56] and then decreased progressively to age 90 (OR = 0.06, 95% CI = 0.04-0.08). They also decreased almost linearly with period relative to 1972-73 (for 2018-19: OR = 0.30, 95% CI = 0.26-0.34) and with birth cohort relative to 1902-06, with the largest decreased observed for birth cohort 1992-96 (OR = 0.44, 95% CI = 0.35-0.46) and 1997-2001 (OR = 0.35, 95% CI = 0.74-0.88). Smoking declined in the 18-25 age group by an average of 7% over successive 2-year periods and by an average of 5% in those aged over 25.

CONCLUSIONS

Smoking in England appears to have declined over recent decades mainly as a result of reduced smoking uptake before age 25, and to a lesser extent to smoking cessation after age 25.

An Age-Period-Cohort Approach to Analyse Late-Life Depression Prevalence in Six European Countries, 2004-2016

Late-life depression is a condition that affects an ever-growing share of the population in ageing societies. While depression prevalence varies across countries for a myriad of reasons, generational factors, expressed in the shared experience of birth cohorts, may also play a part in such differentials. This paper describes the presence of age, period, and cohort (APC) effects in late-life depression prevalence trends (for adults aged 50 and above) for selected countries in Europe, using the Survey of Health and Ageing and Retirement of Europe (SHARE). We analysed six countries during the 2004-2016 period: Denmark, Sweden, and Germany, with a lower baseline prevalence, and Italy, Spain, and France, with a higher baseline prevalence. By applying a set of APC statistical models to visualise linear and nonlinear effects, we found that all countries followed a J-shaped curve when describing the transversal and longitudinal age trajectories of late-life depression. We also found a combination of nonlinear effects present in Germany, France and Sweden in males, indicating that younger male cohorts had a higher relative risk of depression. In females, we found nonlinear cohort effects, indicating that younger and older cohorts presented a higher risk of depression in Sweden and Germany and a lower risk in Spain. The presence of an increased risk for younger male cohorts may be indicative of a new trend in some countries, which may reduce the sex gap in prevalence. Future analysis should focus on the causes and mechanisms that lead to differential risks across cohorts.

Cohort Variation in U.S. Violent Crime Patterns from 1960 to 2014: An Age-Period-Cohort-Interaction Approach

Objectives

Previous research in criminology has overlooked that cohort effects on crime should be age-time-specific (Ryder in Am Sociol Rev 30(6):843-861, 1965) and consequently assumed cohort effects to be the same across the life course. The current study addresses these limitations by modeling cohort effects as the differential impacts of social change depending on age groups. With this new operationalization that is closely tied to Ryder's conceptualization, we examine both inter-cohort differences and intra-cohort dynamics in violent crime.

Methods

We use the age-period-cohort-interaction (APC-I) model developed by Luo and Hodges (Sociol Methods Res 2020) to analyze the UCR age-specific arrest statistics for robbery, aggravated assaults, and homicide from 1960 to 2014. Specifically, we estimate and test two types of cohort variation: average cohort deviations and life-course dynamics.

Results

Our findings reveal varying degrees of cohort deviations at different ages. The early boomers (born between 1945 and 1954) and the late boomers (born between 1955 and 1964) demonstrate different intra-cohort dynamics of robbery arrest, and the violence epidemic cohorts' (born between 1975 and 1984) high risks of homicide arrest appear to be driven by cohort deviations at young ages.

Conclusions

The APC-I framework introduced in this study provides new insights into the dynamic aspect of cohort effects on violent crime that have not been examined in the criminological literature. Criminological studies on cohort effects would benefit by shifting away from the problematic assumption of constant and additive cohort effects to the dynamic and interactive approach represented by the APC-I framework.

Suicide mortality in Spain (1984-2018): age-period-cohort analysis

OBJECTIVE

To assess the effects of age, period and cohort suicide mortality trend in Spain (1984-2018).

METHODS

Mortality and population data were obtained from the National Institute of Statistics. The analysis of the effect of age, period of death and birth cohort on the evolution of suicide mortality in the period 1984-2018 was performed using a web tool for age-period-cohort analysis provided by the Division of Cancer Epidemiology and Genetics of the National Cancer Institute of the USA.

RESULTS

Rates increase with age (age effect) in both sexes. The period effect shows, in males, an increase over the period 1984-1998 followed by a significant decrease until 2018. In females, rates remain stable over the period 1987-2002, decrease during 2007-2012 (p < 0.05) and eventually stabilise. In both males and females, the risk decreases in each successive birth cohort between 1904 and 1939. Subsequently, the risks increase until the birth cohort of the period 1964-1974 after which the risk decreases for males and remains stable for females.

CONCLUSION

A better understanding of the effects of the birth cohort could open new doors in suicide prevention.

Age period cohort analysis: a review of what we should and shouldn't do

Context: Age, period and birth cohort (APC) effects have been known for decades in biological, health and social sciences. However, exact collinearity between these three (Age = Year - Birth Year) leads to difficulty estimating these effects. It is thus impossible to estimate linear components of these effects without strong assumptions about at least one of these. This is problematic for anyone interested in APC patterns. Attempts to 'solve' this identification problem without strong assumptions are, in fact, making hidden unintended assumptions.Objective: Provide an overview of what APC effects are and the nature of the identification problem, before reviewing and critiquing methodological literature across the health and social sciences. I also present an argument for what researchers should do.Method: Non-systematic review of methodological literature across health and social sciences.Results: Recommendations include considering non-linearities around linear APC effects and stating strong and explicit theory-based assumptions. Mechanical solutions to the identification problem do not work.Conclusion: These recommendations acknowledge there is a 'line of solutions' of possible combinations of APC effects, and not a single answer that can be estimated empirically. None of these methods solve the identification problem - rather they acknowledge that methods are limited by assumptions.

Age-period-cohort analysis with a constant-relative-variation constraint for an apportionment of period and cohort slopes

Age-period-cohort analysis of incidence and/or mortality data has received much attention in the literature. To circumvent the non-identifiability problem inherent in the age-period-cohort model, additional constraints are necessary on the parameters estimates. We propose setting the constraint to reflect the different nature of the three temporal variables: age, period, and birth cohort. There are two assumptions in our method. Recognizing age effects to be deterministic (first assumption), we do not explicitly incorporate the age parameters into constraint. For the stochastic period and cohort effects, we set a constant-relative-variation constraint on their trends (second assumption). The constant-relative-variation constraint dictates that between two stochastic effects, one with a larger curvature gets a larger (absolute) slope, and one with zero curvature gets no slope. We conducted Monte-Carlo simulations to examine the statistical properties of the proposed method and analyzed the data of prostate cancer incidence for whites from 1973-2012 to illustrate the methodology. A driver for the period and/or cohort effect may be lacking in some populations. In that case, the CRV method automatically produces an unbiased age effect and no period and/or cohort effect, thereby addressing the situation properly. However, the method proposed in this paper is not a general purpose model and will produce biased results in many other real-life data scenarios. It is only useful in situations when the age effects are deterministic and dominant, and the period and cohort effects are stochastic and minor.

Differences between patterns of suicide in East Asia and the West. The importance of sociocultural factors

OBJECTIVE

To compare suicide rates and characteristics in East Asian and Western countries.

METHOD

Data from five East Asian jurisdictions and five English-speaking Western countries were obtained from national offices responsible for collection of statistics. Data were graphed to demonstrate differences between East Asian and Western age patterns of suicide. Relevant literature was reviewed, including observations concerning suicide rates in various non-English-speaking European countries.

RESULTS

Age patterns and rates of suicide differ between East and West and across time, but with Japan's and the United States patterns progressively becoming more similar.

CONCLUSIONS

Sociocultural factors affect suicide patterns and rates; choice of method and cohort effects affect rates. Strategies for prevention of suicide should give attention to culture-related precipitants of suicide.

A Cause-of-Death Decomposition of Young Adult Excess Mortality

We propose a method to decompose the young adult mortality hump by cause of death. This method is based on a flexible shape decomposition of mortality rates that separates cause-of-death contributions to the hump from senescent mortality. We apply the method to U.S. males and females from 1959 to 2015. Results show divergence between time trends of hump and observed deaths, both for all-cause and cause-specific mortality. The study of the hump shape reveals age, period, and cohort effects, suggesting that it is formed by a complex combination of different forces of biological and socioeconomic nature. Male and female humps share some traits in all-cause shape and trend, but they also differ by their overall magnitude and cause-specific contributions. Notably, among males, the contributions of traffic and other accidents were progressively replaced by those of suicides, homicides, and poisonings; among females, traffic accidents remained the major contributor to the hump.

A longitudinal comparison of age patterns and rates of suicide in Hong Kong, Taiwan and Japan and two Western countries

Suicide data relating to 1979-2014 were obtained from three East Asian jurisdictions (Taiwan, Hong Kong, Japan) and two 'Western' countries (Australia, New Zealand). Rates and age patterns of suicide have changed markedly since 1979. Graphs of these patterns largely remained either upward-sloping, bimodal or flat (uniform) over the 36 years, male commonly differing from female, and East Asian patterns more like each other than those of the Western countries. Japan's male middle-aged suicide rate reached a peak in 1999-2003, which, like increased rates among working age males in Hong Kong and Taiwan, has been attributed largely to consequences of Asian financial crises. Male to female ratios of suicide rates have remained higher in the Western countries, but late life suicide rates have decreased to varying extents in all five jurisdictions. Identifying reasons for differences between jurisdictions in their suicide rates and patterns at particular times, and over time, is likely to point to factors (period, cohort, psychosocial or cultural) that protect against or foster suicidal ideation. This avenue of research may assist in identifying ways of preventing suicide.

Changes in age patterns of suicide in Australia, the United States, Japan and Hong Kong

The patterns of association between age and suicide rate vary by country, subpopulation and gender, and over time. To shed light on factors associated with these differences, we analysed suicide data from four populations, two 'Western' (Australia, the United States [US]) and two Asian (Japan and Hong Kong). We computed suicide rates in five-year age-groups (between 10 and 14 years and 85+ years) for men and women separately, and present graphical representations of the age patterns during selected five-year periods. Rates and age patterns differed markedly, as did gender patterns except in Hong Kong. In 1964-8, male suicide rates in Australia and US were represented by upward-sloping graphs, whereas in Japan the pattern was bimodal. By 1979-83, male patterns in Australia and US were bimodal, but Japan's was trimodal, including a middle-age peak reached in 1994-98. In contrast, female age patterns in the Western countries were shallowly convex or uniform, while in Hong Kong and Japan the upward-sloping graphs became, over time, less steep; by 2009-13, the pattern in Japan was uniform (flat). In recent decades, suicide rates of older men in Australia, US and Japan, and older women in Japan and Hong Kong, have fallen considerably. Suicide rates of men aged 45-64 in Australia and US also fell, though by 2009-13 the US rate had risen again. The suicide rate of Australian men in their twenties halved between 1994-98 and 2009-13, while rates for younger men and women in Japan have risen since 1994-98. In Hong Kong, suicide rates of young men have increased. Age patterns of suicide likely reflect period and cohort effects shaped by socioeconomic stressors, availability of health and welfare services, access to lethal methods of suicide, and other factors. Greater understanding of their impact on age patterns of suicide can result in potential preventive solutions.

The hierarchical age-period-cohort model: Why does it find the results that it finds?

It is claimed the hierarchical-age–period–cohort (HAPC) model solves the age–period–cohort (APC) identification problem. However, this is debateable; simulations show situations where the model produces incorrect results, countered by proponents of the model arguing those simulations are not relevant to real-life scenarios. This paper moves beyond questioning whether the HAPC model works, to why it produces the results it does. We argue HAPC estimates are the result not of the distinctive substantive APC processes occurring in the dataset, but are primarily an artefact of the data structure—that is, the way the data has been collected. Were the data collected differently, the results produced would be different. This is illustrated both with simulations and real data, the latter by taking a variety of samples from the National Health Interview Survey (NHIS) data used by Reither et al. (Soc Sci Med 69(10):1439–1448, 2009) in their HAPC study of obesity. When a sample based on a small range of cohorts is taken, such that the period range is much greater than the cohort range, the results produced are very different to those produced when cohort groups span a much wider range than periods, as is structurally the case with repeated cross-sectional data. The paper also addresses the latest defence of the HAPC model by its proponents (Reither et al. in Soc Sci Med 145:125–128, 2015a). The results lend further support to the view that the HAPC model is not able to accurately discern APC effects, and should be used with caution when there appear to be period or cohort near-linear trends.