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.

Associations of cause-specific mortality with area level deprivation and travel time to health care in France from 1990 to 2007, a multilevel analysis.

BACKGROUND: It is now widely accepted that social and physical environment participate in shaping health. While mortality is used to guide public health policies and is considered as a synthetic measure of population health, few studies deals with the contextual features potentially associated with mortality in a representative sample of an entire country. This paper investigates the possible role of area deprivation (FDep99) and travel time to health care on French cause-specific mortality in a proper multilevel setting. METHODS: The study population was a 1% sample representative of the French population aged from 30 to 79 years in 1990 and followed up until 2007. A frailty Cox model was used to measure individual, contextual effects and spatial variances for several causes of death. The chosen contextual scale was the Zone d'Emploi of 1994 (348 units) which delimits the daily commute of people. The geographical accessibility to health care score was constructed with principal component analysis, using 40 variables of hospital specialties and health practitioners' travel time. RESULTS: The outcomes highlight a positive and significant association between area deprivation and mortality for all causes (HR = 1.24), cancers, cerebrovascular diseases, ischemic heart diseases, and preventable and amenable diseases (HR from 1.14 to 1.29). These contextual associations exhibit no substantial differences by sex except for premature ischemic heart diseases mortality which was much greater in women. Unexpectedly, mortality decreased as the time to reach health care resources increased. Only geographical disparities in cerebrovascular and ischemic heart diseases mortality were explained by compositional and contextual effects. DISCUSSION: The findings suggest the presence of confounding factors in the association between mortality and travel time to health care, possibly owing to population density and health-selected migration. Although the spatial scale considered to define the context of residence was relatively large, the associations with area deprivation were strong in comparison to the existing literature and significant for almost all the causes of deaths investigated. CONCLUSION: The broad spectrum of diseases associated with area deprivation and individual education support the idea of a need for a global health policy targeting both individual and territories to reduce social and socio-spatial inequalities.

Hospital comparisons based on mortality: revisiting the choice of postadmission timeframe and evaluating the contribution of cause-of-death data, France, 2009.

BACKGROUND: In-hospital mortality is widely used to judge the quality of hospital care, but is biased by discharge patterns. Fixed-timeframe indicators have thus been recommended. However, the 30-day postadmission indicator may underestimate hospital-wide mortality, as patients dying in hospital >30 days after admission are considered as survivors. OBJECTIVES: To identify the most relevant timeframes and to assess the contribution of cause-of-death data. METHODS: The 2009 French hospital discharge database was linked to vital status records and to the causes of death register for 11.5 million hospital stays by beneficiaries of French general health insurance. Correlations and agreements between the 30-day hospital standardized mortality ratio (HSMR) and the in-hospital, 60-, 90-, 180-, and 365-day postadmission HSMRs were estimated. RESULTS: A total of 7.8%, 1.5%, and 0.5% of patients who died during their hospital stay were considered as survivors by the 30-, 60-, and 90-day HSMRs, respectively. The 30-day HSMR correlated strongly with the 60-day HSMR (Pearson coefficient=0.92), and their agreement on outlier status was excellent (κ coefficient=0.80). The association remained substantial at 90 days, but weakened at 180 days and even more so at 365 days. Regardless of the timeframe, exclusion of deaths likely due to independent causes barely modified the indicators. CONCLUSIONS: This nationwide study shows that 60- and 90-day HSMRs encompass in-hospital deaths better than the 30-day HSMR, while capturing the same interhospital variations. They should thus be preferred. The contribution of cause-of-death data to hospital-wide indicators seems negligible.

Estimating the number of homeless deaths in France, 2008-2010.

BACKGROUND: The homeless population of France has increased by 50% over the last 10 years. Studies have shown that homelessness is associated with a high risk of premature death. The aim of this study was to estimate the number of homeless deaths in France between 2008 and 2010, using a reproducible method. METHODS: We used the capture-recapture method to estimate the number of homeless deaths in France using two independent sources. An associative register of homeless deaths was matched with the national exhaustive database of the medical causes of death, using several matching approaches based on various combinations of the following variables: gender, age, place of death, date of death. RESULTS: The estimated number of homeless deaths between 2008 and 2010 was 6730 (95% CI: [4381-9079]), a number greatly underestimated by the two sources considered separately (less than 20%). CONCLUSIONS: In the absence of a register of the homeless deaths, the capture-recapture method provides an order of magnitude for evaluation of the resources that may be allocated by policy makers to manage the issue. Based on common and routinely produced databases, this estimate may therefore be used to monitor the mortality of the homeless population. Further studies about homeless mortality, particularly on the lead causes of deaths, are needed to manage this issue and to implement strategy to decrease the number of homeless deaths.

Trends in geographic mortality inequalities and their association with population changes in France, 1975-2006.

BACKGROUND: Although some studies have reported that population change is associated with spatial mortality inequalities, few of them have tried to take a dynamic approach to the association. The aim of this study was to explore and interpret the ecological association between the change in cause-specific mortality inequalities and population change over a 30-year period in areas exhibiting different deprivation and urbanization levels in France. METHODS: The French communes were classified by category of demographic change during the period 1962-2006. The changes in standardized mortality ratios were analysed by category over 5 inter-censal periods, taking into account degree of urbanization and deprivation quintile. The magnitude and significance of the associations for various causes of death were estimated using a Generalised Estimating Equation Poisson model. RESULTS: Overall, the change in relative mortality was negatively associated with population growth. For a compound annual population growth rate of 1% in 1990-99, the standardized mortality ratios decreased, on average, by 2.1% (95% confidence interval: -1.45 to -2.72). The association was stronger in urban areas, and reversed in the least deprived areas. The association was stronger and more significant for men, subjects aged less than 65 years and alcohol-related and violent deaths. CONCLUSION: This study highlights the significance of dynamic approaches. Population growth was associated with a decrease in relative mortality level; the direction and strength of the association varied depending on the socio-territorial characteristics. As is the case for English-speaking countries, in France, population growth may be considered a component of current social dynamics that are not measured by usual indicators.

Impact of the COVID-19 crisis on the mortality profiles of the foreign-born in France during the first pandemic wave.

BACKGROUND: Immigrants in Western countries have been particularly affected by the COVID-19 crisis. OBJECTIVE: We analysed excess mortality rates among the foreign-born population and changes in their distinctive mortality profiles ("migrant mortality advantage") during the first pandemic wave in France. DATA AND METHODS: Deaths from all causes in metropolitan France from March 18 to May 19, 2020 were used, with information on sex, age, region of residence and country of birth. Similar data from 2016 through 2019 were used for comparisons. RESULTS: During the pre-pandemic period (2016-2019), immigrant populations (except those from Central and Eastern Europe) had lower standardized mortality rates than the native-born population, with a particularly large advantage for immigrants from sub-Saharan Africa. In the regions most affected by COVID-19 (Grand-Est and Île-de-France), the differences in excess mortality by country of birth were large, especially in the working-age groups (40-69 years), with rates 8 to 9 times higher for immigrants from sub-Saharan Africa, and about 3 to 4 times higher for immigrants from North Africa, from the Americas and from Asia and Oceania relative to the native-born population. The relative overall mortality risk for men born in sub-Saharan Africa compared to native-born men, which was 0.8 before the pandemic, shifted to 1.8 during the first wave (0.9 to 1.5 for women). It also shifted from 0.8 to 1.1 for men from North Africa (0.9 to 1.1 for women), 0.7 to 1.0 for men from the Americas (0.9 to 1.3 for women), and 0.7 to 1.2 for men from Asia and Oceania (0.9 to 1.3 for women). CONCLUSION: Our findings shed light on the disproportionate impact of the first wave of the pandemic on the mortality of populations born outside Europe, with a specific burden of excess mortality within the working-age range, and a complete reversal of their mortality advantage.

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.

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.

Impact of unemployment variations on suicide mortality in Western European countries (2000-2010).

BACKGROUND: A scientific debate is currently taking place on whether the 2008 economic crisis caused an increase in suicide rates. Our main objective was to assess the impact of unemployment rate on suicide rate in Western European countries between 2000 and 2010. We then tried to estimate the excess number of suicides attributable to the increase of unemployment during the 2008-2010 economic crisis. METHODS: The yearly suicide rates were modelled using a quasi-Poisson model, controlling for sex, age, country and a linear time trend. For each country, the unemployment-suicide association was assessed, and the excess number of suicides attributable to the increase of unemployment was estimated. Sensitivity analyses were performed, notably in order to evaluate whether the unemployment-suicide association found was biased by a confounding context effect ('crisis effect'). RESULTS: A significant 0.3% overall increase in suicide rate for a 10% increase in unemployment rate (95% CI 0.1% to 0.5%) was highlighted. This association was significant in three countries: 0.7% (95% CI 0.0% to 1.4%) in the Netherlands, 1.0% (95% CI 0.2% to 1.8%) in the UK and 1.9% (95% CI 0.8% to 2.9%) in France, with a significant excess number of suicides attributable to unemployment variations between 2008 and 2010 (respectively 57, 456 and 564). The association was modified inconsistently when adding a 'crisis effect' into the model. CONCLUSIONS: Unemployment and suicide rates are globally statistically associated in the investigated countries. However, this association is weak, and its amplitude and sensitivity to the 'crisis effect' vary across countries. This inconsistency provides arguments against its causal interpretation.

Spatial interactions between urban areas and cause-specific mortality differentials in France.

Spatial interactions constitute a challenging but promising approach for investigation of spatial mortality inequalities. Among spatial interactions measures, between-spatial unit migration differentials are a marker of socioeconomic imbalance, but also reflect discrepancies due to other factors. Specifically, this paper asks whether population exchange intensities measure differentials or similarities that are not captured by usual socioeconomic indicators. Urban areas were grouped pairwise by the intensity of connection estimated from a gravity model. The mortality differences for several causes of death were observed to be significantly smaller for strongly connected pairs than for weakly connected pairs even after adjustment on deprivation.