Heat Exposure and Cause-Specific Hospital Admissions in Spain: A Nationwide Cross-Sectional Study.

BACKGROUND: More frequent and intense exposure to extreme heat conditions poses a serious threat to public health. However, evidence on the association between heat and specific diagnoses of morbidity is still limited. We aimed to comprehensively assess the short-term association between cause-specific hospital admissions and high temperature, including the added effect of temperature variability and heat waves and the effect modification by humidity and air pollution. METHODS: We used data on cause-specific hospital admissions, weather (i.e., temperature and relative humidity), and air pollution [i.e., fine particulate matter with aerodynamic diameter ≤2.5µm (PM2.5), fine particulate matter with aerodynamic diameter ≤10µm (PM10), NO2, and ozone (O3)] for 48 provinces in mainland Spain and the Balearic Islands between 1 January 2006 and 31 December 2019. The statistical analysis was performed for the summer season (June-September) and consisted of two steps. We first applied quasi-Poisson generalized linear regression models in combination with distributed lag nonlinear models (DLNM) to estimate province-specific temperature-morbidity associations, which were then pooled through multilevel univariate/multivariate random-effect meta-analysis. RESULTS: High temperature had a generalized impact on cause-specific hospitalizations, while the added effect of temperature variability [i.e., diurnal temperature range (DTR)] and heat waves was limited to a reduced number of diagnoses. The strongest impact of heat was observed for metabolic disorders and obesity [relative risk (RR) = 1.978; 95% empirical confidence interval (eCI): 1.772, 2.208], followed by renal failure (1.777; 95% eCI: 1.629, 1.939), urinary tract infection (1.746; 95% eCI: 1.578, 1.933), sepsis (1.543; 95% eCI: 1.387, 1.718), urolithiasis (1.490; 95% eCI: 1.338, 1.658), and poisoning by drugs and nonmedicinal substances (1.470; 95% eCI: 1.298, 1.665). We also found differences by sex (depending on the diagnosis of hospitalization) and age (very young children and the elderly were more at risk). Humidity played a role in the association of heat with hospitalizations from acute bronchitis and bronchiolitis and diseases of the muscular system and connective tissue, which were higher in dry days. Moreover, heat-related effects were exacerbated on high pollution days for metabolic disorders and obesity (PM2.5) and diabetes (PM10, O3). DISCUSSION: Short-term exposure to heat was found to be associated with new diagnoses (e.g., metabolic diseases and obesity, blood diseases, acute bronchitis and bronchiolitis, muscular and connective tissue diseases, poisoning by drugs and nonmedicinal substances, complications of surgical and medical care, and symptoms, signs, and ill-defined conditions) and previously identified diagnoses of hospital admissions. The characterization of the vulnerability to heat can help improve clinical and public health practices to reduce the health risks posed by a warming planet. https://doi.org/10.1289/EHP13254.

Ambient temperature and risk of cardiovascular and respiratory adverse health outcomes: a nationwide cross-sectional study from Spain.

AIMS: We assessed the association of temperature and temperature variability with cause-specific emergency hospitalizations and mortality from cardiovascular and respiratory diseases in Spain, as well as the effect modification of this association by individual and contextual factors. METHODS AND RESULTS: We collected data on health (hospital admissions and mortality), weather (temperature and relative humidity), and relevant contextual indicators for 48 Spanish provinces during 2004-2019. The statistical analysis was separately performed for the summer (June-September) and winter (December-March) seasons. We first applied a generalized linear regression model with quasi-Poisson distribution to estimate daily province-specific temperature-health associations, and then we fitted multilevel multivariate meta-regression models to the evaluate effect modification of the contextual characteristics on heat- and cold-related risks. High temperature increased the risk of mortality across all cardiovascular and respiratory diseases, with the strongest effect for hypertension (relative risk (RR) at 99th temperature percentile vs. optimum temperature: 1.510 [95% empirical confidence interval {eCI} 1.251 to 1.821]), heart failure (1.528 [1.353 to 1.725]), and pneumonia (2.224 [1.685 to 2.936]). Heat also had an impact on all respiratory hospitalization causes (except asthma), with similar risks between pneumonia (1.288 [1.240 to 1.339]), acute bronchitis and bronchiolitis (1.307 [1.219 to 1.402]), and chronic obstructive pulmonary disease (1.260 [1.158 to 1.372]). We generally found significant risks related to low temperature for all cardiovascular and respiratory causes, with heart failure (RR at 1st temperature percentile vs. optimum temperature: 1.537 [1.329 to 1.779]) and chronic obstructive pulmonary disease (1.885 [1.646 to 2.159]) exhibiting the greatest risk for hospitalization, and acute myocardial infarction (1.860 [1.546 to 2.238]) and pneumonia (1.734 [1.219 to 2.468]) for mortality. Women and the elderly were more vulnerable to heat, while people with secondary education were less susceptible to cold compared to those not achieving this educational stage. Results from meta-regression showed that increasing heating access to the highest current provincial value (i.e. 95.6%) could reduce deaths due to cold by 59.5% (57.2 to 63.5). CONCLUSION: Exposure to low and high temperatures was associated with a greater risk of morbidity and mortality from multiple cardiovascular and respiratory conditions, and heating was the most effective societal adaptive measure to reduce cold-related mortality.

Exposure to low and high temperatures increases the risk of morbidity and mortality from several cardiovascular and respiratory diseases, especially among the elderly. Increasing access to heating could substantially reduce cold-related mortality burden.

Ambient temperature and seasonal variation in inpatient mortality from respiratory diseases: a retrospective observational study.

Background: The seasonal fluctuation in mortality and hospital admissions from respiratory diseases, with a winter peak and a summer trough, is widely recognized in extratropical countries. However, little is known about the seasonality of inpatient mortality and the role of ambient temperature remains uncertain. We aimed to analyse the association between ambient temperature and in-hospital mortality from respiratory diseases in the provinces of Madrid and Barcelona, Spain. Methods: We used data on daily hospitalisations, weather (ie, temperature and relative humidity) and air pollutants (ie, PM2.5, PM10, NO2 and O3) for the Spanish provinces of Madrid and Barcelona during 2006-2019. We applied a daily time-series quasi-Poisson regression in combination with distributed lag non-linear models (DLNM) to assess, on the one hand, the seasonal variation in fatal hospitalisations and the contribution of ambient temperature, and on the other hand, the day-to-day association between temperature and fatal hospital admissions. The analyses were stratified by sex, age and primary diagnostic of hospitalisation. Findings: The study analysed 1 710 012 emergency hospital admissions for respiratory diseases (mean [SD] age, 60.4 [31.0] years; 44.2% women), from which 103 845 resulted in in-hospital death (81.4 [12.3] years; 45.1%). We found a strong seasonal fluctuation in in-hospital mortality from respiratory diseases. While hospital admissions were higher during the cold season, the maximum incidence of inpatient mortality was during the summer and was strongly related to high temperatures. When analysing the day-to-day association between temperature and in-hospital mortality, we only found an effect for high temperatures. The relative risk (RR) of fatal hospitalisation at the 99th percentile of the distribution of daily temperatures vs the minimum mortality temperature (MMT) was 1.395 (95% eCI: 1.211-1.606) in Madrid and 1.612 (1.379-1.885) in Barcelona. In terms of attributable burden, summer temperatures (June-September) were responsible for 16.2% (8.8-23.3) and 22.3% (15.4-29.2) of overall fatal hospitalisations from respiratory diseases in Madrid and Barcelona, respectively. Women were more vulnerable to heat than men, whereas the results by diagnostic of admission showed heat effects for acute bronchitis and bronchiolitis, pneumonia and respiratory failure. Interpretation: Unless effective adaptation measures are taken in hospital facilities, climate warming could exacerbate the burden of inpatient mortality from respiratory diseases during the warm season. Funding: European Research Council Consolidator Grant EARLY-ADAPT, European Research Council Proof-of-Concept Grants HHS-EWS and FORECAST-AIR.

Collider and reporting biases involved in the analyses of cause of death associations in death certificates: an illustration with cancer and suicide.

BACKGROUND: Mortality data obtained from death certificates have been studied to explore causal associations between diseases. However, these analyses are subject to collider and reporting biases (selection and information biases, respectively). We aimed to assess to what extent associations of causes of death estimated from individual mortality data can be extrapolated as associations of disease states in the general population. METHODS: We used a multistate model to generate populations of individuals and simulate their health states up to death from national health statistics and artificially replicate collider bias. Associations between health states can then be estimated from such simulated deaths by logistic regression and the magnitude of collider bias assessed. Reporting bias can be approximated by comparing the estimates obtained from the observed death certificates (subject to collider and reporting biases) with those obtained from the simulated deaths (subject to collider bias only). As an illustrative example, we estimated the association between cancer and suicide in French death certificates and found that cancer was negatively associated with suicide. Collider bias, due to conditioning inclusion in the study population on death, increasingly downwarded the associations with cancer site lethality. Reporting bias was much stronger than collider bias and depended on the cancer site, but not prognosis. RESULTS: The magnitude of the biases ranged from 1.7 to 9.3 for collider bias, and from 4.7 to 64 for reporting bias. CONCLUSIONS: These results argue for an assessment of the magnitude of both collider and reporting biases before performing analyses of cause of death associations exclusively from mortality data. If these biases cannot be corrected, results from these analyses should not be extrapolated to the general population.

Artificial intelligence exceeds humans in epidemiological job coding.

BACKGROUND: Work circumstances can substantially negatively impact health. To explore this, large occupational cohorts of free-text job descriptions are manually coded and linked to exposure. Although several automatic coding tools have been developed, accurate exposure assessment is only feasible with human intervention. METHODS: We developed OPERAS, a customizable decision support system for epidemiological job coding. Using 812,522 entries, we developed and tested classification models for the Professions et Catégories Socioprofessionnelles (PCS)2003, Nomenclature d'Activités Française (NAF)2008, International Standard Classifications of Occupation (ISCO)-88, and ISCO-68. Each code comes with an estimated correctness measure to identify instances potentially requiring expert review. Here, OPERAS' decision support enables an increase in efficiency and accuracy of the coding process through code suggestions. Using the Formaldehyde, Silica, ALOHA, and DOM job-exposure matrices, we assessed the classification models' exposure assessment accuracy. RESULTS: We show that, using expert-coded job descriptions as gold standard, OPERAS realized a 0.66-0.84, 0.62-0.81, 0.60-0.79, and 0.57-0.78 inter-coder reliability (in Cohen's Kappa) on the first, second, third, and fourth coding levels, respectively. These exceed the respective inter-coder reliability of expert coders ranging 0.59-0.76, 0.56-0.71, 0.46-0.63, 0.40-0.56 on the same levels, enabling a 75.0-98.4% exposure assessment accuracy and an estimated 19.7-55.7% minimum workload reduction. CONCLUSIONS: OPERAS secures a high degree of accuracy in occupational classification and exposure assessment of free-text job descriptions, substantially reducing workload. As such, OPERAS significantly outperforms both expert coders and other current coding tools. This enables large-scale, efficient, and effective exposure assessment securing healthy work conditions.

Work can expose us to health risks, such as asbestos and constant noise. To study these risks, job descriptions are collected and classified by experts to standard codes. This is time-consuming, expensive, and requires expert knowledge. To improve this coding, we created computer code based on Artificial Intelligence that can both automate this process and suggest codes to experts, who can then check and change it manually if needed. Our system outperforms both expert coders and other available tools. This system could make studying occupational health risks more efficient and accurate, resulting in safer work environments.

Drivers of the time-varying heat-cold-mortality association in Spain: A longitudinal observational study.

BACKGROUND: A number of studies have reported reductions in mortality risk due to heat and cold over time. However, questions remain about the drivers of these adaptation processes to ambient temperatures. We aimed to analyse the demographic and socioeconomic drivers of the downward trends in vulnerability to heat- and cold-related mortality observed in Spain during recent decades (1980-2018). METHODS: We collected data on all-cause mortality, temperature and relevant contextual indicators for 48 provinces in mainland Spain and the Balearic Islands between Jan 1, 1980, and Dec 31, 2018. Fourteen contextual indicators were analysed representing ageing, isolation, urbanicity, heating, air conditioning (AC), house antiquity and ownership, education, life expectancy, macroeconomics, socioeconomics, and health investment. The statistical analysis was separately performed for the range of months mostly causing heat- (June-September) and cold- (October-May) related mortality. We first applied a quasi-Poisson generalised linear regression in combination with distributed lag non-linear models (DLNM) to estimate province-specific temperature-mortality associations for different periods, and then we fitted univariable and multivariable multilevel spatiotemporal meta-regression models to evaluate the effect modification of the contextual characteristics on heat- and cold-related mortality risks over time. FINDINGS: The average annual mean temperature has risen at an average rate of 0·36 °C per decade in Spain over 1980-2012, although the increase in temperature has been more pronounced in summer (0·40 °C per decade in June-September) than during the rest of the year (0·33 °C per decade). This warming has been observed, however, in parallel with a progressive reduction in the mortality risk associated to both hot and cold temperatures. We found independent associations for AC with heat-related mortality, and heating with cold-related mortality. AC was responsible for about 28·6% (31·5%) of the decrease in deaths due to heat (extreme heat) between 1989 and 1993 and 2009-2013, and heating for about 38·3% (50·8%) of the reductions in deaths due to cold (extreme cold) temperatures. Ageing (ie, proportion of population over 64 years) attenuated the decrease in cold-related mortality. INTERPRETATION: AC and heating are effective societal adaptive measures to heat and cold temperatures. This evidence holds important implications for climate change health adaptation policies, and for the projections of climate change impacts on human health.

Suicide and All-Cause Mortality Within 1 Year After a Suicide Attempt in the VigilanS Cohort.

Objective: Obtaining better knowledge on the outcomes of patients who attempt suicide is crucial for suicide prevention. The aim of our study was to determine the causes of death 1 year after a suicide attempt (SA) in the VigilanS program, mortality rates, and risk factors associated with any cause of death and suicide.Methods: A prospective cohort of 7,406 people who had attempted suicide between January 1, 2017, and December 31, 2018, was included in the study. The vital status of each participant was sought, and the cause of death was established through a phone call to their general practitioner or psychiatrist. Second, the relationship between sociodemographic and clinical factors and death by suicide within 1 year of an SA was assessed using a multivariable Cox model.Results: At 1 year, 125 (1.7%) participants had died, 77 of whom died by suicide. Half of the deaths occurred within the first 4 months after an SA. Hanging (20.3%; 24/125) and self-poisoning (19.5%; 23/125) were the methods the most often used for suicide. We demonstrated that male sex (HR = 1.79 [1.13-2.82], P = .01) and being 45 years of age or older (between 45 and 64 years old, HR = 2.08 [1.21-3.56], P < .01; 65 years or older, HR = 5.36 [2.72-10.54], P < .01) were associated with a higher risk of death by suicide 1 year after an SA and that being younger than 25 years was associated with a lower risk (HR = 0.22 [0.07-0.76], P = .02).Conclusions: One out of 100 people who attempted suicide died by suicide within 1 year after an SA. Greater vigilance is required in the first months following an SA, especially for males older than 45 years.Trial Registration: ClinicalTrials.gov identifier: NCT03134885.

Reactive surveillance of suicides during the COVID-19 pandemic in France, 2020 to March 2022.

AIMS: Mitigation actions during the COVID-19 pandemic may impact mental health and suicide in general populations. We aimed to analyse the evolution in suicide deaths from 2020 to March 2022 in France. METHODS: Using free-text medical causes in death certificates, we built an algorithm, which aimed to identify suicide deaths. We measured its retrospective performances by comparing suicide deaths identified using the algorithm with deaths which had either a Tenth revision of the International Classification of Diseases (ICD-10) code for 'intentional self-harm' or for 'external cause of undetermined intent' as the underlying cause. The number of suicide deaths from January 2020 to March 2022 was then compared with the expected number estimated using a generalized additive model. The difference and the ratio between the observed and expected number of suicide deaths were calculated on the three lockdown periods and for periods between lockdowns and after the third one. The analysis was stratified by age group and gender. RESULTS: The free-text algorithm demonstrated high performances. From January 2020 to mid-2021, suicide mortality declined during France's three lockdowns, particularly in men. During the periods between and after the two first lockdowns, suicide mortality remained comparable to the expected values, except for men over 85 years old and in 65-84 year-old age group, where a small number of excess deaths was observed in the weeks following the end of first lockdown, and for men aged 45-64 years old, where the decline continued after the second lockdown ended. After the third lockdown until March 2022, an increase in suicide mortality was observed in 18-24 year-old age group for both genders and in men aged 65-84 years old, while a decrease was observed in the 25-44 year-old age group. CONCLUSIONS: This study highlighted the absence of an increase in suicide mortality during France's COVID-19 pandemic and a substantial decline during lockdown periods, something already observed in other countries. The increase in suicide mortality observed in 18-24 year-old age group and in men aged 65-84 years old from mid-2021 to March 2022 suggests a prolonged impact of COVID-19 on mental health, also described on self-harm hospitalizations and emergency department's attendances in France. Further studies are required to explain the factors for this change. Reactive monitoring of suicide mortality needs to be continued since mental health consequences and the increase in suicide mortality may be continued in the future with the international context.

Daylight saving time affects European mortality patterns.

Daylight saving time (DST) consists in a one-hour advancement of legal time in spring offset by a backward transition of the same magnitude in fall. It creates a minimal circadian misalignment that could disrupt sleep and homoeostasis in susceptible individuals and lead to an increased incidence of pathologies and accidents during the weeks immediately following both transitions. How this shift affects mortality dynamics on a large population scale remains, however, unknown. This study examines the impact of DST on all-cause mortality in 16 European countries for the period 1998-2012. It shows that mortality decreases in spring and increases in fall during the first two weeks following each DST transition. Moreover, the alignment of time data around DST transition dates revealed a septadian mortality pattern (lowest on Sundays, highest on Mondays) that persists all-year round, irrespective of seasonal variations, in men and women aged above 40.

Association of Daily Temperature With Suicide Mortality: A Comparison With Other Causes of Death and Characterization of Possible Attenuation Across 5 Decades.

Suicide is one of the leading causes of death in young adults in many Western countries. We examined the short-term association of temperature with cause-specific mortality, comparing suicide with other causes of death and describing possible attenuation of associations with temperature across decades. We considered all deaths that occurred in France between 1968 and 2016. For each cause of death, we conducted a 2-stage meta-analysis of associations with daily temperature. We stratified the association across time periods. A total of 502,017 deaths by suicide were recorded over 49 years. Temperature was monotonically associated with suicide mortality. The strongest association was found at lag 0 days. The relative risk of suicide mortality at the 99th (compared with the 1st) temperature percentile was 1.54 (95% confidence interval, 1.46, 1.63). Among all causes of death, suicide was the only cause displaying a monotonic trend with temperature and ranked seventh for heat-related mortality; 2 other causes of death implying the nervous system ranked third and fourth. Associations with temperature attenuated between the 1968-1984 and 1985-2000 periods for all-cause mortality and suicide mortality, without clear further attenuation in the 2001-2016 period. The robust short-term monotonic association between temperature and suicide risk could be considered in heat effects- and suicide-related prevention campaigns.

Neural Translation and Automated Recognition of ICD-10 Medical Entities From Natural Language: Model Development and Performance Assessment.

BACKGROUND: The recognition of medical entities from natural language is a ubiquitous problem in the medical field, with applications ranging from medical coding to the analysis of electronic health data for public health. It is, however, a complex task usually requiring human expert intervention, thus making it expansive and time-consuming. Recent advances in artificial intelligence, specifically the rise of deep learning methods, have enabled computers to make efficient decisions on a number of complex problems, with the notable example of neural sequence models and their powerful applications in natural language processing. However, they require a considerable amount of data to learn from, which is typically their main limiting factor. The Centre for Epidemiology on Medical Causes of Death (CépiDc) stores an exhaustive database of death certificates at the French national scale, amounting to several millions of natural language examples provided with their associated human-coded medical entities available to the machine learning practitioner. OBJECTIVE: The aim of this paper was to investigate the application of deep neural sequence models to the problem of medical entity recognition from natural language. METHODS: The investigated data set included every French death certificate from 2011 to 2016. These certificates contain information such as the subject's age, the subject's gender, and the chain of events leading to his or her death, both in French and encoded as International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) medical entities, for a total of around 3 million observations in the data set. The task of automatically recognizing ICD-10 medical entities from the French natural language-based chain of events leading to death was then formulated as a type of predictive modeling problem known as a sequence-to-sequence modeling problem. A deep neural network-based model, known as the Transformer, was then slightly adapted and fit to the data set. Its performance was then assessed on an external data set and compared to the current state-of-the-art approach. CIs for derived measurements were estimated via bootstrapping. RESULTS: The proposed approach resulted in an F-measure value of 0.952 (95% CI 0.946-0.957), which constitutes a significant improvement over the current state-of-the-art approach and its previously reported F-measure value of 0.825 as assessed on a comparable data set. Such an improvement makes possible a whole field of new applications, from nosologist-level automated coding to temporal harmonization of death statistics. CONCLUSIONS: This paper shows that a deep artificial neural network can directly learn from voluminous data sets in order to identify complex relationships between natural language and medical entities, without any explicit prior knowledge. Although not entirely free from mistakes, the derived model constitutes a powerful tool for automated coding of medical entities from medical language with promising potential applications.

Recent historic increase of infant mortality in France: A time-series analysis, 2001 to 2019.

BACKGROUND: The infant mortality rate (IMR) serves as a key indicator of population health. METHODS: We used data from the French National Institute of Statistics and Economic Studies on births and deaths during the first year of life from 2001 to 2019 to calculate IMR aggregated by month. We ran joinpoint regressions to identify inflection points and assess the linear trend of each segment. Exploratory analyses were performed for overall IMR, as well as by age at death subgroups (early neonatal [D0-D6], late neonatal [D7-27], and post-neonatal [D28-364]), and by sex. We performed sensitivity analyses by excluding deaths at D0 and using other time-series modeling strategies. RESULTS: Over the 19-year study period, 53,077 infant deaths occurred, for an average IMR of 3·63/1000 (4·00 in male, 3·25 in female); 24·4% of these deaths occurred during the first day of life and 47·8% during the early neonatal period. Joinpoint analysis identified two inflection points in 2005 and 2012. The IMR decreased sharply from 2001 to 2005 (slope: -0·0167 deaths/1000 live births/month; 95%CI: -0·0219 to -0·0116) and then decreased slowly between 2005 and 2012 (slope: -0·0041; 95%CI: -0·0065 to -0·0016). From 2012 onwards, a significant increase in IMR was observed (slope: 0·0033; 95%CI: 0·0011 to 0·0056). Subgroup analyses indicated that these trends were driven notably by an increase in the early neonatal period. Sensitivity analyses provided consistent results. INTERPRETATION: The recent historic increase in IMR since 2012 in France should prompt urgent in-depth investigation to understand the causes and prepare corrective actions. FUNDING: No financial relationships with any organizations that might have an interest in the submitted work in the previous three years, no other relationships or activities that could appear to have influenced the submitted work.

Time trend in excess mortality in children with type 1 diabetes from 1987 to 2016 in mainland France.

BACKGROUND: Mortality risk for children with type 1 diabetes (T1D) is unknown in France and their causes of death are not well documented. AIM: To determine the standardized mortality ratios (SMRs) and causes of death in children aged 1-14 years with T1D from 1987 to 2016. METHODS: The French Center for Epidemiology on Medical Causes of Death collected all death certificates in mainland France. SMRs, corrected SMRs (accounting for missing cases of deaths unrelated to diabetes), and 95% confidence intervals were calculated. RESULTS: Of 146 deaths with the contribution of diabetes, 97 were due to T1D. Mean age at death of the subjects with T1D was 8.8 ± 4.1 years (54% males). The cause of death was diabetic ketoacidosis (DKA) in 58% of the cases (70% in subjects 1-4 years), hypoglycemia or dead-in-bed syndrome in 4%, related to diabetes but not described in 24%, and unrelated to diabetes in 14%. The SMRs showed a significant decrease across the years, except for the 1-4 age group. In the last decade (2007-2016), the crude and corrected SMRs were significantly different from 1 in the 1-4 age group (5.4 [2.3; 10.7] and 6.1 [2.8; 11.5]), no longer significant in the 5-9 age group (1.7 [0.6; 4.0] and 2.1 [0.8; 4.5]) and borderline significant in the 10-14 age group (1.7 [0.8; 3.2] and 2.3 [1.2; 4.0]). CONCLUSIONS: Children with T1D aged 1-4 years still had a high mortality rate. Their needs for early recognition and safe management of diabetes are not being met.

National Variations in Recent Trends of Sudden Unexpected Infant Death Rate in Western Europe.

OBJECTIVE: To study recent epidemiologic trends of sudden unexpected death in infancy (SUDI) in Western Europe. STUDY DESIGN: Annual national statistics of death causes for 14 Western European countries from 2005 to 2015 were analyzed. SUDI cases were defined as infants younger than 1 year with the underlying cause of death classified as "sudden infant death syndrome," "unknown/unattended/unspecified cause," or "accidental threats to breathing." Poisson regression models were used to study temporal trends of SUDI rates and source of variation. RESULTS: From 2005 to 2015, SUDI accounted for 15 617 deaths, for an SUDI rate of 34.9 per 100 000 live births. SUDI was the second most common cause of death after the neonatal period (22.2%) except in Belgium, Finland, France, and the UK, where it ranked first. The overall SUDI rate significantly decreased from 40.2 to 29.9 per 100 000, with a significant rate reduction experienced for 6 countries, no significant evolution for 7 countries, and a significant increase for Denmark. The sudden infant death syndrome/SUDI ratio was 56.7%, with a significant decrease from 64.9% to 49.7% during the study period, and ranged from 6.1% in Portugal to 97.8% in Ireland. We observed between-country variations in SUDI and sudden infant death syndrome sex ratios. CONCLUSIONS: In studied countries, SUDI decreased during the study period but remained a major cause of infant deaths, with marked between-country variations in rates, trends, and components. Standardization is needed to allow for comparing data to improve the implementation of risk-reduction strategies.

Contacts with Health Services During the Year Prior to Suicide Death and Prevalent Conditions A Nationwide Study.

BACKGROUND: This study was designed to describe contacts with health services during the year before suicide death in France, and prevalent mental and physical conditions. METHODS: Data were extracted from the French National Health Data System (SNDS), which comprises comprehensive claims data for inpatient and outpatient care linked to the national causes-of-death registry. Individuals aged ≥15 years who died from suicide in France in 2013-2015 were included. Medical consultations, emergency room visits, and hospitalisations during the year preceding death were collected. Conditions were identified, and standardised prevalence ratios (SPRs) were estimated to compare prevalence rates in suicide decedents with those of the general population. RESULTS: The study included 19,144 individuals. Overall, 8.5% of suicide decedents consulted a physician or attended an emergency room on the day of death, 34.1% during the week before death, 60.9% during the month before death. Most contacts involved a general practitioner or an emergency room. During the month preceding suicide, 24.4% of individuals were hospitalised at least once. Mental conditions (36.8% of cases) were 7.9-fold more prevalent in suicide decedents than in the general population. The highest SPRs among physical conditions were for liver/pancreatic diseases (SPR=3.3) and epilepsy (SPR=2.7). LIMITATIONS: The study population was restricted to national health insurance general scheme beneficiaries (76% of the population living in France). CONCLUSIONS: Suicide decedents have frequent contacts with general practitioners and emergency departments during the last weeks before death. Improving suicide risk identification and prevention in these somatic healthcare settings is needed.

A Deep Artificial Neural Network-Based Model for Prediction of Underlying Cause of Death From Death Certificates: Algorithm Development and Validation.

BACKGROUND: Coding of underlying causes of death from death certificates is a process that is nowadays undertaken mostly by humans with potential assistance from expert systems, such as the Iris software. It is, consequently, an expensive process that can, in addition, suffer from geospatial discrepancies, thus severely impairing the comparability of death statistics at the international level. The recent advances in artificial intelligence, specifically the rise of deep learning methods, has enabled computers to make efficient decisions on a number of complex problems that were typically considered out of reach without human assistance; they require a considerable amount of data to learn from, which is typically their main limiting factor. However, the CépiDc (Centre d'épidémiologie sur les causes médicales de Décès) stores an exhaustive database of death certificates at the French national scale, amounting to several millions of training examples available for the machine learning practitioner. OBJECTIVE: This article investigates the application of deep neural network methods to coding underlying causes of death. METHODS: The investigated dataset was based on data contained from every French death certificate from 2000 to 2015, containing information such as the subject's age and gender, as well as the chain of events leading to his or her death, for a total of around 8 million observations. The task of automatically coding the subject's underlying cause of death was then formulated as a predictive modelling problem. A deep neural network-based model was then designed and fit to the dataset. Its error rate was then assessed on an exterior test dataset and compared to the current state-of-the-art (ie, the Iris software). Statistical significance of the proposed approach's superiority was assessed via bootstrap. RESULTS: The proposed approach resulted in a test accuracy of 97.8% (95% CI 97.7-97.9), which constitutes a significant improvement over the current state-of-the-art and its accuracy of 74.5% (95% CI 74.0-75.0) assessed on the same test example. Such an improvement opens up a whole field of new applications, from nosologist-level batch-automated coding to international and temporal harmonization of cause of death statistics. A typical example of such an application is demonstrated by recoding French overdose-related deaths from 2000 to 2010. CONCLUSIONS: This article shows that deep artificial neural networks are perfectly suited to the analysis of electronic health records and can learn a complex set of medical rules directly from voluminous datasets, without any explicit prior knowledge. Although not entirely free from mistakes, the derived algorithm constitutes a powerful decision-making tool that is able to handle structured medical data with an unprecedented performance. We strongly believe that the methods developed in this article are highly reusable in a variety of settings related to epidemiology, biostatistics, and the medical sciences in general.

Automatic classification of free-text medical causes from death certificates for reactive mortality surveillance in France.

BACKGROUND: Mortality surveillance is of fundamental importance to public health surveillance. The real-time recording of death certificates, thanks to Electronic Death Registration System (EDRS), provides valuable data for reactive mortality surveillance based on medical causes of death in free-text format. Reactive mortality surveillance is based on the monitoring of mortality syndromic groups (MSGs). An MSG is a cluster of medical causes of death (pathologies, syndromes or symptoms) that meets the objectives of early detection and impact assessment of public health events. The aim of this study is to implement and measure the performance of a rule-based method and two supervised models for automatic free-text cause of death classification from death certificates in order to implement them for routine surveillance. METHOD: A rule-based method was implemented using four processing steps: standardization rules, splitting causes of death using delimiters, spelling corrections and dictionary projection. A supervised machine learning method using a linear Support Vector Machine (SVM) classifier was also implemented. Two models were produced using different features (SVM1 based solely on surface features and SVM2 combining surface features and MSGs classified by the rule-based method as feature vectors). The evaluation was conducted using an annotated subset of electronic death certificates received between 2012 and 2016. Classification performance was evaluated on seven MSGs (Influenza, Low respiratory diseases, Asphyxia/abnormal respiration, Acute respiratory disease, Sepsis, Chronic digestive diseases, and Chronic endocrine diseases). RESULTS: The rule-based method and the SVM2 model displayed a high performance with F-measures over 0.94 for all MSGs. Precision and recall were slightly higher for the rule-based method and the SVM2 model. An error-analysis shows that errors were not specific to an MSG. CONCLUSION: The high performance of the rule-based method and SVM2 model will allow us to set-up a reactive mortality surveillance system based on free-text death certificates. This surveillance will be an added-value for public health decision making.

A New Approach to Compare the Performance of Two Classification Methods of Causes of Death for Timely Surveillance in France.

Timely mortality surveillance in France is based on the monitoring of electronic death certificates to provide information to health authorities. This study aims to analyze the performance of a rule-based and a supervised machine learning method to classify medical causes of death into 60 mortality syndromic groups (MSGs). Performance was first measured on a test set. Then we compared the trends of the monthly numbers of deaths classified into MSGs from 2012 to 2016 using both methods. Among the 60 MSGs, 31 achieved recall and precision over 0.95 for either one or the other method on the test set. On the whole dataset, the correlation coefficient of the monthly numbers of deaths obtained by the two methods were close to 1 for 21 of the 31 MSGs. This approach is useful for analyzing a large number of categories or when annotated resources are limited.

Study of the Usability of an Automated Coding Software for Causes of Death in an African Context.

IRIS is an automated coding software for the causes of death. It is used in many European countries for the production of death statistics. The purpose of our work was to study the usability of this software in Africa where the quality of statistics is insufficient. For this, we have developed a device consisting of two software: "collector" and "encoder" cooperating via the same database.

A Note on the Measurement of Socioeconomic Inequalities in Life Years Lost by Cause of Death.

Quantifying socioeconomic inequalities in health in absolute terms is of prime interest for decision-making and for international comparisons. The Slope Index of Inequality (SII), an index that quantifies absolute socioeconomic inequalities, was recently formalized, particularly in the context of mortality differences measured in the rate or hazard scale. However, absolute inequalities using either rates or hazards do not translate into a time dimension, which makes their interpretation difficult for policymakers. We propose an extension of the (Equation is included in full-text article.)in terms of the expected number of life years lost before an upper age, as well as its decomposition by cause of death. The (Equation is included in full-text article.)in the life years lost metric quantifies the extent to which life expectancy is shortened when comparing the higher and lower ends of the socioeconomic scale. The methodology proposed builds on recent developments in survival analysis for decomposing the number of life years lost according to cause of death using a pseudo-value approach. We illustrate our proposal using a representative 1% sample of the French population. On average, the least educated men lost 7 years of life from age 30 up to age 90 compared to the most educated. The loss for women is twice as much with 3.5 years. The (Equation is included in full-text article.)in the life years lost metric is easily understood, and the decomposition of the all-cause mortality (Equation is included in full-text article.)into parts attributable to given causes provides a sound estimation of the burden of different causes of death on absolute socioeconomic inequalities in mortality.