Accuracy of diagnosis on death certificates for underlying causes of death in a long-term autopsy-based population study in Hisayama, Japan; with special reference to cardiovascular diseases

Progress in life expectancy and the growing number of people living to old age intensify the phenomenon of multi-morbidity, defined as the coexistence of several chronic diseases. By exploiting all the medical information in death certificates, the multiple causes of death (MCoD) approach serves to investigate complex pathological processes that lead eventually to death. This is the first MCoD analysis for Poland and its objective is twofold: to examine the quality of information on contributing causes of death, in particular in the regional dimension, and to assess the scale of multi-morbidity involving conditions that are becoming more and more frequent in ageing populations. The analysis is carried out for all deaths that took place in Poland in 2013. The results show that medical doctors issuing death certificates often define contributing causes of death, but a large part of this information includes unknown or ill-defined conditions. Several conditions favour the certification of well-defined contributing causes: when death occurs in hospital, or is due to underlying causes other than cardiovascular, the number of contributing conditions is higher. Important regional differences are observed in this regard. The analysis highlights the importance of diseases that are rarely certified as the underlying causes, but often contribute to mortal conditions, such as diseases of the blood and the blood-forming organs, diseases of the skin and subcutaneous tissue, diseases of the genitourinary system or mental and behavioural disorders.

Download Full-text

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

JMIR Medical Informatics ◽

10.2196/17125 ◽

2020 ◽

Vol 8 (4) ◽

pp. e17125 ◽

Cited By ~ 1

Author(s):

Louis Falissard ◽

Claire Morgand ◽

Sylvie Roussel ◽

Claire Imbaud ◽

Walid Ghosn ◽

...

Keyword(s):

Neural Network ◽

Cause Of Death ◽

Deep Neural Network ◽

Causes Of Death ◽

State Of The Art ◽

Death Certificates ◽

Current State ◽

Underlying Causes ◽

Artificial Neural ◽

Underlying Cause Of Death

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.

Download Full-text

Geochemical Risk Factors for Mental Functioning, Based on the Ontario Longitudinal Study of Aging (LSA) V. Comparisons of the Results, relevant to Aluminum Water Concentrations, obtained from the LSA and from Death Certificates mentioning Dementia

Canadian Journal on Aging / La Revue canadienne du vieillissement ◽

10.1017/s0714980800016378 ◽

1995 ◽

Vol 14 (4) ◽

pp. 642-656 ◽

Cited By ~ 15

Author(s):

W. F. Forbes ◽

S. Lessard ◽

J. F. Gentleman

Keyword(s):

Causes Of Death ◽

Outcome Variable ◽

Death Certificates ◽

Mental Functioning ◽

Mental Impairment ◽

Presenile Dementia ◽

Water Characteristics ◽

Underlying Causes ◽

Trial Basis ◽

Underlying Cause Of Death

AbstractPrevious studies in this series of papers investigated the associations between aluminum (Al) water concentrations and relatively high risks of a measure of mental impairment and also various possible other drinking water characteristics, particularly pH, turbidity, fluoride and silica. The results were based on one measure of mental impairment, which would not be expected to give the same results as the more definitive endpoint (outcome variable) of a record of Alzheimer's Disease (AD) as the underlying cause of death on a death certificate. The present paper therefore investigates the relevant associations, based both on the measure of mental impairment and on death certificates in which AD and presenile dementia are listed as the underlying causes of death. As expected, the associations were not identical, but they were similar. More specifically, Al water concentrations were strongly associated with the recording of AD on death certificates, as were pH, fluoride, and silica concentrations. The implications of these results are discussed, and it is suggested that the evidence is sufficiently strong for methods of water purification to be modified, at least on a trial basis, because of the likelihood that this will reduce the incidence of AD.

Download Full-text

Sickle Cell Disease Mortality in California and Georgia 2004-2008

Blood ◽

10.1182/blood.v124.21.439.439 ◽

2014 ◽

Vol 124 (21) ◽

pp. 439-439

Author(s):

Susan Paulukonis ◽

Todd Griffin ◽

Mei Zhou ◽

James R. Eckman ◽

Robert Hagar ◽

...

Keyword(s):

African American ◽

Cause Of Death ◽

Sickle Cell ◽

Causes Of Death ◽

Mortality Rates ◽

Population Based ◽

Age Group ◽

Death Certificates ◽

Underlying Causes ◽

Underlying Cause Of Death

Abstract On-going public health surveillance efforts in sickle cell disease (SCD) are critical for understanding the course and outcomes of this disease over time. Once nearly universally fatal by adolescence, many patients are living well into adulthood and sometimes into retirement years. Previous SCD mortality estimates have relied on data from death certificates alone or from deaths of patients receiving care in high volume hematology clinics, resulting in gaps in reporting and potentially biased conclusions. The Registry and Surveillance System for Hemoglobinopathies (RuSH) project collected and linked population-based surveillance data on SCD in California and Georgia from a variety of sources for years 2004-2008. These data sources included administrative records, newborn screening reports and health insurance claims as well as case reports of adult and pediatric patients receiving care in the following large specialty treatment centers: Georgia Comprehensive Sickle Cell Center, Georgia Regents University, Georgia Comprehensive Sickle Cell Center at Grady Health Systems and Children's Healthcare of Atlanta in Georgia, and Children's Hospital Los Angeles and UCSF Benioff Children's Hospital Oakland in California. Cases identified from these combined data sources were linked to death certificates in CA and GA for the same years. Among 12,143 identified SCD cases, 640 were linked to death certificates. Combined SCD mortality rates by age group at time of death are compared to combined mortality rates for all African Americans living in CA and GA. (Figure 1). SCD death rates among children up to age 14 and among adults 65 and older were very similar to those of the overall African American population. In contrast, death rates from young adulthood to midlife were substantially higher in the SCD population. Overall, only 55% of death certificates linked to the SCD cases had SCD listed in any of the cause of death fields. Thirty-four percent (CA) and 37% (GA) had SCD as the underlying cause of death. An additional 22% and 20% (CA and GA, respectively) had underlying causes of death that were not unexpected for SCD patients, including related infections such as septicemia, pulmonary/cardiac causes of death, renal failure and stroke. The remaining 44% (CA) and 43% (GA) had underlying causes of death that were either not related to SCD (e.g., malignancies, trauma) or too vague to be associated with SCD (e.g., generalized pulmonary or cardiac causes of death. Figure 2 shows the number of deaths by state, age group at death and whether the underlying cause of death was SCD specific, potentially related to SCD or not clearly related to SCD. While the number of deaths was too small to use for life expectancy calculations, there were more deaths over age 40 than under age 40 during this five year period. This effort represents a novel, population-based approach to examine mortality in SCD patients. These data suggest that the use of death certificates alone to identify deceased cases may not capture all-cause mortality among all SCD patients. Additional years of surveillance are needed to provide better estimates of current life expectancy and the ability to track and monitor changes in mortality over time. On-going surveillance of the SCD population is required to monitor changes in mortality and other outcomes in response to changes in treatments, standards of care and healthcare policy and inform advocacy efforts. This work was supported by the US Centers for Disease Control and Prevention and the National Heart, Lung and Blood Institute, cooperative agreement numbers U50DD000568 and U50DD001008. Figure 1: SCD-Specific & Overall African American Mortality Rates in CA and GA, 2004 – 2008. Figure 1:. SCD-Specific & Overall African American Mortality Rates in CA and GA, 2004 – 2008. Figure 2: Deaths (Count) Among Individuals with SCD in CA and GA, by Age Group and Underlying Cause of Death, 2004-2008 (N=615) Figure 2:. Deaths (Count) Among Individuals with SCD in CA and GA, by Age Group and Underlying Cause of Death, 2004-2008 (N=615) Disclosures No relevant conflicts of interest to declare.

Download Full-text

Excess Mortality Among 10-Year Survivors of Classical Hodgkin Lymphoma in Adolescents and Young Adults

Blood ◽

10.1182/blood.v128.22.694.694 ◽

2016 ◽

Vol 128 (22) ◽

pp. 694-694

Author(s):

Ana Xavier ◽

Luciano J. Costa

Keyword(s):

Cardiovascular Disease ◽

Young Adults ◽

Cardiovascular Diseases ◽

Hodgkin Lymphoma ◽

Causes Of Death ◽

Excess Mortality ◽

Mortality Rates ◽

Classical Hodgkin Lymphoma

Background: Adolescents and young adults (AYA) surviving classical Hodgkin Lymphoma (cHL) commonly face significant treatment-related morbidity that can result in premature death. It is not known if changes in therapy in recent years have resulted in reduction of excess mortality among long-term survivors of AYA-cHL. Methods: We used data from the National Cancer Institute's Surveillance Epidemiology and End Results program (SEER-18) to determine the excess mortality rate (EMR) for 10-year survivors of cHL. We included patients age 15-39 years diagnosed with cHL as first malignant neoplasm from 1993 until 2003. Cases reported from death certificates or autopsy only were excluded. Follow up was updated at the end of 2013. EMR was calculated using the difference between the observed mortality in the cohort of interest and the adjusted expected mortality among age, gender and race-matched individuals in the general population. Results: A total of 6,480 cases of cHL were included in the analysis with median follow up of 42.5 months. Median age of patients was 27 years at the time of diagnosis, 3,172 (48.9%) were male, 4,405 (68%) had stage I or II, 1,750 (27%) had stage III or IV, and in 325 (5%) cases stage was not available. Most patients were non-Hispanic white (4,783, 73.8%), 761 (11.7%) Hispanic, 615 (9.5%) non-Hispanic black, 276 (4.3%) other ethnicity, and 45 (0.7%) unknown. Five thousand and sixty eight (78.2%) had nodular sclerosis cHL, 616 (9.5%) had mixed-cellularity cHL, 606 (9.4%) had cHL non-otherwise specified, 153 (2.4%) had lymphocyte-rich cHL, and 37 (0.57%) had lymphocyte depleted cHL. The 15-year, 18-year, and 20-year EMR for 10-year survivors was 2.19% (95% C.I, 1.69%-2.86%), 3.48% (95% C.I. 2.57%-4.64%) and 4.07% (95% C.I. 2.53%-6.52%), respectively. EMR among 10 year AYA cHL survivors with diagnosis between 1993 and 2003 was substantially improved when compared to a similar cohort of 5,870 survivors with diagnosis between 1973 and 1992 (Figure 1). EMR at 15 and 18 years of diagnosis was higher among survivors of stage III-IV cHL than among survivors of stages I-II (Figure 2a), and higher at 15 years among males than female survivors (Figure 2b). Use of radiation therapy for early stage disease did not seem to affect the risk of excess mortality among 10-year survivors (Table). Causes of death were similar for stages I-II vs. stages III-IV. Most common causes of death were second malignancy (27.2%), cardiovascular disease (19%), and Hodgkin lymphoma (18.5%). Among male survivors there was a higher proportion of deaths due to cardiovascular disease (23.7% vs. 11.5%, p <0.038). Conclusion: Excess mortality rates for 10-year survivors of AYA-cHL has decreased with adoption of less toxic therapies. However, mortality rates continue high for several years for long term AYA HL survivors, mainly due to second malignancies and cardiovascular diseases. Less toxic therapies, control of cardiovascular diseases and implementation of cancer prevention programs for survivors of AYA-cHL are needed. Table. Table. Disclosures Costa: Sanofi: Honoraria, Research Funding.

Download Full-text