Characteristics of and Variation in Suicide Mortality Related to Retirement during the Great Recession: Perspectives from the National Violent Death Reporting System

BACKGROUND AND OBJECTIVES

Suicide rates typically increase during recessions. However, few studies have explored how recessions impact risk among older adults nearing retirement. This study used a large suicide mortality registry to characterize and quantify suicide related to retirement during the Great Recession (GR).

RESEARCH DESIGN AND METHODS

Data come from the National Violent Death Reporting System (NVDRS, 2004-2017; N=53,298 suicide deaths ages ≥50). We analyzed the text narratives (i.e., descriptions of the most salient circumstances to each suicide) of these decedents using natural language processing (NLP) to identify cases that were "retirement-related" (RR, e.g., anticipating, being unable to, or recently retiring). We used time-series analysis to quantify variation in RR over the GR, and compared these trends to retirees (i.e., decedents whose occupation was "retired") and all decedents aged ≥50. We used content and network analysis to characterize themes represented in the narratives.

RESULTS

There were 878 RR cases (1.6% of suicides aged ≥50) identified by the NLP model; only 52% of these cases were among retirees. RR cases were younger (62 vs. 75 years) and more educated (41.5% vs 24.5% college degree) than retirees. The rate of RR suicide was positively associated with indicators of the GR (e.g., short-term unemployment R2=0.70, p=0.024), but economic indicators were not correlated with the suicide rate among retirees or older adults in general. Economic issues were more central to the narratives of RR cases during the GR compared to other periods.

DISCUSSION AND IMPLICATIONS

Recessions shape suicide risk related to retirement transitions.

Application of wastewater-based surveillance and copula time-series model for COVID-19 forecasts

The objective of this study was to develop a novel copula-based time series (CTS) model to forecast COVID-19 cases and trends based on wastewater SARS-CoV-2 viral load and clinical variables. Wastewater samples were collected from wastewater pumping stations in five sewersheds in the City of Chesapeake VA. Wastewater SARS-CoV-2 viral load was measured using reverse transcription droplet digital PCR (RT-ddPCR). The clinical dataset included daily COVID-19 reported cases, hospitalization cases, and death cases. The CTS model development included two steps: an autoregressive moving average (ARMA) model for time series analysis (step I), and an integration of ARMA and a copula function for marginal regression analysis (step II). Poisson and negative binomial marginal probability densities for copula functions were used to determine the forecasting capacity of the CTS model for COVID-19 forecasts in the same geographical area. The dynamic trends predicted by the CTS model were well suited to the trend of the reported cases as the forecasted cases from the CTS model fell within the 99 % confidence interval of the reported cases. Wastewater SARS CoV-2 viral load served as a reliable predictor for forecasting COVID-19 cases. The CTS model provided robust modeling to predict COVID-19 cases.

RETIREMENT TRANSITIONS AND COMPLETED SUICIDE DURING RECOVERY FROM THE GREAT RECESSION: EVIDENCE FROM THE NVDRS

One million older Americans retire annually. While these transitions are not generally associated with poor mental health, the broader macro-economic context in which retirement transitions take place may shape how they relate to mental health. The objective of this study was to use state-of-the-art natural language processing (NLP) to develop a model to identify retirement transitions from textual data in the National Violent Death Reporting System (NVDRS), and to use that model to examine how the number of suicides related to retirement transitions changed during the recovery from the Great Recession. Data come from the NVDRS (2003 - 2018, n=62,165), a state-based registry of suicide deaths. We used RoBERTa to train a NLP model to identify retirement transitions (e.g., recent retirement, anticipated retirement, unable to retire despite wanting to) from 1,291 annotated sentences from NVDRS text narratives of suicide decedents aged ≥55 (model performance: F1=0.92). Applying this model, 19.35 of every 1,000 suicides among decedents aged ≥55 years mentioned a retirement transition. Decedent characteristics associated with retirement transitions were younger age (< 75 years), having a college education and experiencing financial problems. The probability that a narrative referenced a retirement transition increased 1.495-fold during the Great Recession (2007 - 2009) and declined during recovery (2009-2016) before beginning to increase again. Findings illustrate the utility of NLP methods to identify workforce transitions from NVDRS narratives, and further understanding the impact of macro contextual events like the Great Recession on population mental health.

Work Expectations, Depressive Symptoms, and Passive Suicidal Ideation Among Older Adults: Evidence From the Health and Retirement Study

BACKGROUND AND OBJECTIVES

Employment and work transitions (e.g., retirement) influence mental health. However, how psychosocial contexts such as anticipation and uncertainty about work transitions, irrespective of the transitions themselves, relate to mental health is unclear. This study examined the relationships of work expectations with depressive symptoms, major depression episodes (MDE), and passive suicidal ideation over a 10-year period among the "Baby Boom" cohort of the Health and Retirement Study.

RESEARCH DESIGN AND METHODS

Analysis was limited to 13,247 respondents aged 53-70 observed from 2008 to 2018. Past-year depressive symptoms, MDE, and passive suicidal ideation were indexed using the Composite International Diagnostic Interview-Short Form. Expectations regarding working full-time after age 62 were assessed using a probability scale (0%-100%). Mixed-effect logistic regressions with time-varying covariates were used to assess the relationship of work expectations with mental health, accounting for demographics, health status, and functioning, and stratified by baseline employment status.

RESULTS

At baseline, higher work expectations were inversely associated with depressive symptoms. Longitudinally, higher expectations were associated with lower odds of depressive symptoms (odds ratio [OR] = 0.93, 95% CI: 0.91, 0.94). This association was more pronounced among respondents not working at baseline (ORNot working = 0.93 vs ORWorking = 0.96). Greater uncertainty (i.e., expectations near 50%) was also inversely associated with depressive symptoms. Results were similar for past-year MDE and passive suicidal ideation.

DISCUSSION AND IMPLICATIONS

Expectations (overall likelihood and uncertainty), as indicators of psychosocial context, provide insight into the processes that link work transitions with depression risk.

Language in popular American culture constructs the meaning of healthy and unhealthy eating: Narratives of craveability, excitement, and social connection in movies, television, social media, recipes, and food reviews

Many people want to eat healthier but struggle to do so, in part due to a dominant perception that healthy foods are at odds with hedonic goals. Is the perception that healthy foods are less appealing than unhealthy foods represented in language across popular entertainment media and social media? Six studies analyzed dialogue about food in six cultural products - creations of a culture that reflect its perspectives - including movies, television, social media posts, food recipes, and food reviews. In Study 1 (N = 617 movies) and Study 2 (N = 27 television shows), healthy foods were described with fewer appealing descriptions (e.g., "couldn't stop eating"; d = 0.59 and d = 0.37, respectively) and more unappealing descriptions (e.g., "I hate peas"; d = -.57 and d = -.63, respectively) than unhealthy foods in characters' speech from the film and television industries. Using sources with richer descriptive language, Studies 3-6 analyzed popular American restaurants' Facebook posts (Study 3, N = 2275), recipe descriptions from Allrecipes.com (Study 4, N = 1000), Yelp reviews from six U.S. cities (Study 5, N = 4403), and Twitter tweets (Study 6, N = 10,000) for seven specific themes. Meta-analytic results across Studies 3-6 showed that healthy foods were specifically described as less craveworthy (d = 0.51, 95% CI: 0.44-0.59), less exciting (d = 0.40, 95% CI: 0.31-0.49), and less social (d = 0.36, 95% CI: 0.04-0.68) than unhealthy foods. Machine learning methods further generalized patterns across 1.6 million tweets spanning 42 different foods representing a range of nutritional quality. These data suggest that strategies to encourage healthy choices must counteract pervasive narratives that dissociate healthy foods from craveability, excitement, and social connection in individuals' everyday lives.

A Data Science Approach to Estimating the Frequency of Driving Cessation Associated Suicide in the US: Evidence From the National Violent Death Reporting System

Driving cessation is a common transition experienced by aging adults that confers both a symbolic and literal loss of independence due to the central role of automobiles for mobility in the US. Prior research has shown that driving cessation has negative implications for mental health, social participation, and access to healthcare. Given these sequelae of driving cessation and prior work showing that late-life transitions related to independence (e.g., transitioning into residential care) are associated with suicide, we sought to estimate the frequency of driving cessation associated suicide. Data include suicide (n = 59,080) and undetermined (n = 6,862) deaths aged ≥55 from the National Violent Death Reporting System (NVDRS, 2003-2017). Each case in the NVDRS has both quantitative data (e.g., demographic characteristics) and qualitative text narratives, derived from coroner/medical examiner reports, which describe the most salient circumstances and features of each death. To identify cases associated with driving cessation, we employed a supervised random forest algorithm to develop a Natural Language Processing (NLP) classifier. Identified driving cessation associated cases were then categorized and characterized using descriptive statistics and qualitative content analysis. From 2003 to 2017, there were an estimated 305 cases of suicide/undetermined deaths associated with driving cessation in the NVDRS, representing 0.04% of all cases. Cases associated with driving cessation were older, more likely to be male, more likely to have a physical health problem, more likely to have experienced a recent crisis, and more likely to have lived in a rural county than other decedents. Qualitative analysis identified functional impairment, alcohol-related driving limitations, loss of employment, and recent car accidents as common themes among cases associated with driving cessation. This analysis illustrates the utility of NLP in identifying novel correlates of suicide in later life. Although driving cessation associated suicide is a rare outcome, further research is warranted on understanding the conditions under which driving cessation is associated with suicidal behavior, and how to support the well-being of aging adults during these types of major life transitions.

Measuring the predictability of life outcomes with a scientific mass collaboration

How predictable are life trajectories? We investigated this question with a scientific mass collaboration using the common task method; 160 teams built predictive models for six life outcomes using data from the Fragile Families and Child Wellbeing Study, a high-quality birth cohort study. Despite using a rich dataset and applying machine-learning methods optimized for prediction, the best predictions were not very accurate and were only slightly better than those from a simple benchmark model. Within each outcome, prediction error was strongly associated with the family being predicted and weakly associated with the technique used to generate the prediction. Overall, these results suggest practical limits to the predictability of life outcomes in some settings and illustrate the value of mass collaborations in the social sciences.

Suicide Among Older Adults Living in or Transitioning to Residential Long-term Care, 2003 to 2015

Importance

Almost 25% of Medicare beneficiaries live in residential long-term care (LTC) (eg, independent or assisted living facility or nursing home). There are few reliable statistics on completed suicide in LTC, in part because of data limitations.

Objectives

To estimate the number of suicides associated with residential LTC (ie, among persons in a facility, transitioning into or out of a facility, or otherwise associated with LTC) among adults 55 and older and, secondarily, to identify whether machine learning tools could improve the quality of suicide surveillance data.

Design, Setting, and Participants

Cross-sectional epidemiologic study (conducted in 2018) of restricted-access data from the National Violent Death Reporting System (NVDRS) (2003-2015) using restricted-access case narratives from suicides and undetermined deaths among adults 55 years and older in 27 states. Participants were all suicides and undetermined deaths (N = 47 759) among persons 55 years and older.

Exposure

Long-term care cited in the coroner/medical examiner case narrative, whether as a reason for self-harm or the injury location, identified using machine learning natural language processing (NLP) algorithms plus manual review of texts.

Main Outcomes and Measures

Number and characteristics (eg, demographics, health history, and means of injury) of suicides associated with LTC. The κ statistic was used to estimate the reliability of the existing NVDRS injury location codes relative to cases identified by the algorithm.

Results

Among 47 759 persons 55 years and older (median age, 64 years; 77.6% male; 90.0% non-Hispanic white), this study identified 1037 suicide deaths associated with LTC, including 428 among older adults living in LTC, 449 among older adults transitioning to LTC, and 160 otherwise associated with LTC. In contrast, there were only 263 cases coded with the existing NVDRS location code "supervised residential facility," which had poor agreement with cases that the algorithm identified as occurring in LTC (κ statistic, 0.30; 95% CI, 0.26-0.35).

Conclusions and Relevance

Over a 13-year period, approximately 2.2% of suicides among adults 55 years and older were associated with LTC in some manner. Clinicians, administrators, and policy makers should consider ways to promote the mental health and well-being of older adults experiencing functioning limitations and their families. Natural language processing may be a useful way to improve abstraction of variables in the NVDRS.

Semi-supervised Learning with Induced Word Senses for State of the Art Word Sense Disambiguation

Word Sense Disambiguation (WSD) aims to determine the meaning of a word in context, and successful approaches are known to benefit many applications in Natural Language Processing. Although supervised learning has been shown to provide superior WSD performance, current sense-annotated corpora do not contain a sufficient number of instances per word type to train supervised systems for all words. While unsupervised techniques have been proposed to overcome this data sparsity problem, such techniques have not outperformed supervised methods. In this paper, we propose a new approach to building semi-supervised WSD systems that combines a small amount of sense-annotated data with information from Word Sense Induction, a fully-unsupervised technique that automatically learns the different senses of a word based on how it is used. In three experiments, we show how sense induction models may be effectively combined to ultimately produce high-performance semi-supervised WSD systems that exceed the performance of state-of-the-art supervised WSD techniques trained on the same sense-annotated data. We anticipate that our results and released software will also benefit evaluation practices for sense induction systems and those working in low-resource languages by demonstrating how to quickly produce accurate WSD systems with minimal annotation effort.

Gas Chromatographic Determination of Butachlor in Formulations: Collaborative Study

An isothermal gas chromatographic (GC) method for measuring butachlor in formulations was tested by 15 laboratories. Samples were prepared in acetone and determined using a GC column of 10% SP-2250 on 100-120 mesh Supelcoport, and triphenyl phosphate as internal standard. Collaborators made single determinations on each of 2 similar materials (matched pairs) on 2 days for 2 pairs of samples. The study generated 56 matched pairs which were used in Youden’s matched pair calculations. Coefficients of variation (CV,) for liquid formulations were 1.06 and 0.98%. The method was simple to use and did not reveal any interferences in the samples tested. The method has been adopted official first action as an AOAC-CIPAC method.

Liquid Chromatographic Method for Determination of Oxythioquinox in Technical and Formulated Products: Collaborative Study

A liquid chromatographic method for determination of oxythioquinox (Morestan®) in oxythioquinox technical and formulated products has been developed and collaboratively studied in 14 laboratories. Samples are dissolved in chloroform containing n-valerophenone as an internal standard, diluted with acetonitrile, and analyzed by reverse phase chromatography. Collaborators analyzed blind duplicate samples of oxythioquinox technical and 25 WP. Coefficients of variation were 1.06 and 1.72% for the technical and 25 WP samples, respectively. The method has been adopted official first action.

Liquid Chromatographic Determination of Methiocarb in Technical and Formulated Products: Collaborative Study

An LC method for the determination of methiocarb in methiocarb technical and formulated products has been subjected to a collaborative study with 9 participating collaborators. Formulations are extracted with acetonitrile and analyzed by reverse phase chromatography, with acetophenone as an internal standard. Collaborators were furnished samples of technical, 75% wettable powder, 75% seed treater, 75% concentrate, and 50% hopper box treater. Coefficient of variation values obtained on the 5 samples were 0.71, 0.83, 0.62, 1.57, and 0.82%, respectively. The method has been adopted official first action.

Gas Chromatographic Determination of Fluazifop-butyl in Formulations: Collaborative Study

A method is described for the determination of butyl 2-(4-(5-trifluoromethyl- 2-pyridinyloxy)phenoxy) propanoate in technical and formulated material by gas chromatography (GC). Samples of technical or formulated material are dissolved in methylene chloride containing dibenzyl succinate as internal standard. The solution is injected into a gas chromatograph fitted with a flame ionization detector and a 6 ft × 2 mm glass column. The column is packed with 3% OV-17 on 100–120 mesh Chromosorb WHP and is maintained at 230°C. Peak areas obtained at the retention times of the internal standard and the active ingredient are measured with an integrator. The quantity of fluazifop-butyl is determined by comparing the internal standard and active ingredient peak areas with the peak areas obtained from a calibration solution containing known amounts of internal standard and pure active ingredient. Four samples were chosen for collaboration: technical fluazifop-butyl, a 4 lb/US gal. emulsifiable concentrate, a 560 g/L oil concentrate, and a 5% granular. Fourteen collaborators carried out replicate determinations on each sample. Coefficients of variation between laboratories were 0.80 for the technical, 0.88 for the emulsifiable concentrate, 1.03 for the oil concentrate, and 2.04 for the granular. Parallel peak height determinations showed substantially higher coefficients of variation for all except the granular sample. The method has been adopted official first action.

Liquid Chromatographic Determination of Benomyl in Wettable Powders: Collaborative Study

A simple and rapid determination of benomyl in wettable powders was tested by 10 collaborators. Formulations of benomyl are extracted into acetonitrile containing 3% n-butyl isocyanate, filtered, and chromatographed on a reverse phase (C-18) column with a minimum of 1000 plates. Benomyl is detected at 290 nm and quantitated by peak height. Three closely matched pairs of benomyl formulations were assayed along with a duplicate pair of 9 8% technical benomyl. The coefficients of variation were 1.20% for the 50% wettable powder, 5.89% for the 10% benomyl formulation with 50% captan, and 2.49% for a formulation of 6% benomyl with 10% captan and 50% lindane. Analysis of the technical benomyl gave a coefficient of variation of 1.52%. Collaborators stated the method was simple, rapid, and reproducible, and the statistical analysis showed it to be precise and free from expected interferences. The method has been adopted official first action.

Atomic Absorption Spectrophotometric Determination of Chelated Iron in Iron Chelate Concentrates: Collaborative Study

An atomic absorption Spectrophotometric (AAS) method for determining chelated iron in iron chelate concentrates was collaboratively studied by 13 laboratories. Nonchelated iron was selectively precipitated as ferric hydroxide from an aqueous solution of the sample by adjusting to pH 8.5 with sodium hydroxide. A filtered portion of the sample solution was diluted with 0.5N HC1, and the iron was determined by AAS using standards containing disodium EDTA. Six pairs of iron chelate materials (12 samples) were analyzed and the average coefficients of variation were iron EDTA 3.52%, iron HEDTA 4.16%, iron DTPA 3.55%, iron EDDHA 3.52%, iron citrate 2.86%, and iron DPS 8.47%. The method has been adopted official first action.