INTRODUCTION: The Covid-19 pandemic has caused significant increases in mortality in the United States, and nursing homes were particularly impacted early in the pandemic. With questions around underreporting, limited testing, and indirect effects, it is difficult to understand the true impact of the pandemic on US mortality while only examining the deaths attributed to Covid-19. Prior analyses have looked at excess mortality, the increase in mortality over what would have been expected in 2020 if the Covid-19 pandemic had not occurred, focusing on temporal and spatial relationships between excess mortality and direct Covid-19 attributed mortality. The true impact of Covid-19 by place of death remains to be understood. Recent historical trends in mortality by place of death have shown a decreasing share of deaths occurring in hospitals while deaths in homes have increased and deaths in nursing homes have not changed significantly. OBJECTIVE: This observational study aims to characterize the impact of the Covid-19 pandemic on nursing homes in the US by examining direct Covid-19 mortality, excess mortality, and the relationship between direct and excess mortality by place of death at a state level. METHODS: Vital statistics data around mortality by place of death from CDC WONDER and the NVSS Provisional Covid-19 Deaths dataset were used along with US Census data to create a time series for US mortality by place of death from 2013 to 2020. The analysis was restricted to individuals above the age of 65 to limit fields with missing or suppressed data and stratified by 10-year age category. 2020 mortality in the absence of Covid-19 was estimated using the historical average mortality and the simple linear extrapolation of historical mortality within each age group, place of death, and state. Excess deaths were divided into those assigned to Covid-19 and those not assigned to Covid-19 and compared by place of death, age category, and state. RESULTS: 26.2% of direct Covid-19 deaths were found to occur in nursing homes, while 63.1% of Covid-19 deaths occurred in hospitals and 5.3% occurred at home. The excess mortality rate was found to be the highest at home, with 1.78 more deaths per thousand person-years occurring in 2020 in the US than would have been expected in the absence of Covid-19, despite a low direct Covid-19 mortality rate of only 0.162 deaths per thousand person-years. Excess mortality rates in nursing homes across the US were relatively low at 0.296 deaths per thousand person-years, with a direct Covid-19 mortality rate (1.29 deaths per thousand person-years) that was higher than the estimated excess mortality. Despite the high direct Covid-19 mortality compared to excess mortality in nursing homes, a regression model examining the extent to which Covid-19 mortality and historical mortality predicted 2020 mortality in nursing homes suggested that for every 100 deaths assigned to Covid-19, there were 107 more all-cause deaths in 2020. Nursing home excess mortality was found to be highest in Utah, and lowest in North Carolina and New York. CONCLUSION: This work suggests that direct Covid-19 mortality captures most of the impact of Covid-19 on mortality in US nursing homes in individuals over the age of 65. A significant difference was discovered between direct Covid-19 mortality and excess mortality in decedent’s homes, which warrants additional study. / 2022-11-23T00:00:00Z

Abstract Objectives: To identify a simple and effective means for public health professionals in rural communities to identify excess mortality using publicly available data. Methods: Online data from four rural funeral homes, and obituary data from the most widely circulated newspaper in the same region were collected from January 2017 through December 2020. A three-year monthly average of death listings was created for 2017-2019 and compared, month-by-month with the amount of 2020 death listings. Results: The four funeral homes reported a total of 3,957 deaths, and there were 7,623 newspaper obituaries published between January 2017 and December 2020. In the five-month period following the first COVID-19 death in the region on July 28, 2020, funeral home reports and newspaper obituaries reported a 20.2% and a 14.5% increase in deaths, respectively, for 2020 compared to the prior three-year average. Conclusion: During the five months following the first death attributed to COVID-19, funeral home reports and newspaper obituaries both identified a significant increase in deaths over the monthly average death listings of the three years prior. Policy Implications: Local public health officials may be able to use a multi-year, month-by-month summary of deaths, as reported by funeral homes and/or newspaper obituaries, to provide an “early indicator” of excess mortality in rural areas.

This thesis analyses the COVID-19 pandemic’s effects on Swedish mortality during 2020 by investigating whether it has resulted in excess mortality. This is done using a stochastic mortality projection model from the Lee-Carter framework and by assuming the number of deaths follows a Poisson distribution. Due to the few confirmed COVID-19 deaths at younger ages, the decision is made to only include 50-to-100-year-olds in the analysis. Models in the Lee-Carter framework are fitted on historical data from 1993–2019 collected from Human Mortality Database and Statistiska Centralbyrån. After evaluating the models, inter alia using residual analysis and backtesting, we ascertain that the classical Lee-Carter model accomplishes a wanted level of fit and forecast accuracy. During the morality projection with the Lee-Carter model, three different sources of uncertainty are accounted for by constructing prediction intervals using bootstrap. The results show that the large age group 67–94-year-olds have suffered from statistically significant excess mortality during 2020. The level of excess mortality differs between ages, with the ages 70–90-year-olds having the highest number of excess deaths. Comparing the number of confirmed COVID-19 deaths to our forecasted number of excess deaths indicates the COVID-19 virus likely caused the surge in deaths.

La survie nette est un indicateur incontournable pour juger du control du cancer. Par définition, elle correspond à la survie que l’on observerait dans un monde hypothétique où le cancer étudié serait la seule cause possible de décès. L’objectif principal de cette thèse était de montrer l’intérêt de cet indicateur dans le cadre de la recherche clinique en prenant en compte quelques défis méthodologiques qui peuvent être rencontrés dans ce contexte. Nous avons présenté d’abord le concept de survie nette et ses méthodes d’estimation. Par la suite nous nous sommes intéressés à quelques problématiques rencontrées dans les essais cliniques à long terme lorsque l’on s’intéresse à l’estimation de la survie nette. Nous avons étudié également l’impact de l’utilisation de l’approche classique d’estimation de la survie nette dans les essais cliniques, i.e. la méthode cause-spécifique dans différentes configurations d’erreurs de classifications de la cause de décès. La deuxième problématique de cette thèse a porté sur la prise en compte du biais de sélection en termes de mortalité autres causes des patients. Nous avons proposé un modèle de mortalité en excès prenant en compte ce type de biais de sélection. Une troisième problématique qui est complémentaire à la deuxième est de prendre en compte inter-centres en même temps que le biais de sélection. Ce travail propose ainsi de nouveaux outils pouvant aider les spécialistes de la recherche clinique à évaluer de nouvelles stratégies thérapeutiques dans les essais cliniques en cancérologie, mais aussi dans d’autres domaines cliniques d’applications. / Net survival is a key indicator for measuring cancer control. By definition, it corresponds to the survival that would be observed in a hypothetical world where the cancer studied is the only possible cause of death. The main objective of this thesis was to show the interest of this indicator in the context of clinical research taking into account some methodological challenges that can be encountered. In this work, we have first presented the concept of net survival and its estimation methods. Subsequently, we were interested in some of the problems encountered in long-term clinical trials when the interest is in estimating net survival. We studied the impact of using the classic approach when estimating net survival in clinical trials, i.e. the cause-specific method in different configurations of misclassifications of the cause of death. The second objective of this thesis was to take into account the selection bias in terms of other causes mortality in the modeling of excess mortality, because of the noncomparability between patients from general population and those of clinical trials. We proposed an excess hazard model that corrects this type of selection bias. A third problem which is complementary to the second is to take into account the heterogeneity of patients in the different recruitment centers at the same time as the selection bias. This work proposes new tools which can help clinical research specialists to evaluate new therapeutic strategies in cancer clinical trials, but also in other areas of clinical application.

La survie nette, indicateur clé de l'efficacité des systèmes de soin dans la lutte contre le cancer, est un concept théorique représentant la survie que l'on observerait dans un monde hypothétique où le cancer étudié serait la seule cause de décès. En s'affranchissant de la mortalité due aux causes autres que ce cancer, elle permet des comparaisons entre populations. Dans cette thèse, après présentation du concept et des méthodes d'estimation de la survie nette quand la cause de décès est inconnue, nous étudions trois problématiques. La première porte sur les tables de mortalité utilisées pour estimer la survie nette. En France, ces tables sont stratifiées sur âge, sexe, année et département. Il serait intéressant d'utiliser des tables stratifiées sur d'autres facteurs impactant la mortalité. Nous étudions l'impact du manque de stratification sur les estimations des effets des facteurs pronostiques sur la mortalité en excès (celle due au cancer en l'absence des autres causes de décès) par des études de simulations et sur données réelles. La deuxième problématique porte sur la construction d'un test de type log-rank pour comparer des distributions de survie nette estimées par l'estimateur Pohar-Perme, estimateur non paramétrique consistant de la survie nette. Notre troisième problématique est de déterminer dans une aire géographique des zones différentes en termes de survie nette. Nous adaptons une méthode de détection de clusters à la survie nette en utilisant le test précédemment développé comme critère de découpage. Ce travail propose ainsi des développements et outils nouveaux pour étudier et améliorer la qualité de la prise en charge des patients atteints d'un cancer. / In cancer research, net survival is a key indicator of health care efficiency. This theoretical concept is the survival that would be observed in an hypothetical world where the disease under study would be the only possible cause of death. In population-based studies, where cause of death is unknown, net survival allows to compare net cancer survival between different groups by removing the effect of death from causes other than cancer. In this work, after presenting the concept and the estimation methods of net survival, we focus on three complementary issues. The first one is about the life tables used in the estimates of net survival. In France, these tables are stratified by age, sex, year and département. Other prognostic factors impact on mortality. So it would be interesting to use life tables stratified by some of these factors. We study the impact of the lack of stratification in life tables on the estimates of the effects of prognostic factors on excess mortality by simulations and real data studies. In 2012, the Pohar-Perme estimator was proposed. It is a consistent non parametric estimator of net survival. The second issue involves the building of a log-rank type test to compare distributions of net survival (estimated by the Pohar-Perme estimator) between several groups. Our third issue is to propose a method providing potential spatial clusters which could contain patients with similar net cancer survival rates. We adapt a clustering method using the test we have built as a splitting criterion. This work proposes new developments and new tools to study and improve the quality of care for cancer patients. These methods are suitable to other chronic diseases.

L'étude de la survie nette des patients atteints de cancer en population générale permet d'apprécier l'efficience globale du système de soin d'un pays. La survie nette se définit comme la survie qui serait observée si la seule cause de décès possible était le cancer. Ce concept est fondamental dans les comparaisons entre zones géographiques et/ou périodes de diagnostic dont l'intérêt est d'estimer les variations spécifiques de la mortalité due au cancer. Le concept de survie nette permet de prendre en compte les éventuelles différences de mortalité naturelle entre les groupes comparés. Actuellement, seuls deux outils estiment la survie nette sans biais : l'estimateur non paramétrique de Pohar-Perme et la modélisation paramétrique ajustée sur certaines covariables (essentiellement l'âge). Par ailleurs, les outils paramétriques s'étant perfectionnés, de nouveaux modèles flexibles permettent de modéliser les effets complexes des variables sur la mortalité. Ce travail repose sur la modélisation du taux de mortalité en excès à la suite d'un lymphome malin non hodgkinien, en se basant sur le modèle proposé par Remontet et al. et sur la nécessité de modéliser conjointement les effets complexes des covariables (telles que le temps de suivi, l'année de diagnostic et l'âge) sur la mortalité à l'aide d'une stratégie de modélisation adaptée. L'effet des variables est restitué sur la survie nette mais aussi sur le taux de mortalité en excès ce qui représente un élément nouveau dans les études de survie. Deux applications ont été menées sur des bases de données collaboratives de population : d'une part sur les données françaises du réseau FRANCIM à la suite d'un diagnostic de lymphome folliculaire entre 1995 et 2010 et, d'autre part, sur les données européennes d'EUROCARE-5 après un lymphome folliculaire ou un lymphome B diffus à grandes cellules diagnostiqué entre 1996 et 2004. Les résultats montrent que la dynamique du taux de mortalité en excès au cours du temps de suivi varie en fonction du sous-type de lymphome, de l'âge et de la zone géographique. Les tendances de cette dynamique en fonction de l'année de diagnostic sont également différentes / The net survival of cancer patients in population studies is the most relevant indicator to assess the overall efficiency of the healthcare system of a country. Net survival is defined as the survival that would be observed if the sole cause of death were cancer. This concept is crucial in comparative studies (between geographical areas and/or periods of diagnosis) that estimate specific variations of cancer-related deaths. Net survival takes into account potential differences in mortality patterns between groups. Currently, two methods provide unbiased estimations of net survival: the non-parametric estimator of Pohar-Perme and the parametric model adjusted on specific covariates (mainly, the age at diagnosis). Moreover, new improved parametric tools, such as flexible models, can model the complex covariate effects on mortality. In this work, we modeled the excess mortality rate after a non Hodgkin lymphoma diagnosis, with a model developed by Remontet et al. In addition, we used an appropriate model-building-strategy to model jointly the complex effects of some covariates (such as the time elapsed since diagnosis, the year of diagnosis, and age) on the excess mortality. Finally, this approach allowed for the covariate effects on the net survival and on the excess mortality rate. We applied this method to two different collaborative databases: first on the French database FRANCIM (1995 to 2010) to study the excess mortality after diagnosis of follicular lymphoma, then on the European data of EUROCARE-5 (1996 to 2004) to study the excess mortality after diagnosis of follicular lymphoma and diffuse large B-cell lymphoma. According to the results, the dynamics of the excess mortality rate varies over the time elapsed since diagnosis according to the lymphoma subtype, the age, and the geographical area. The trends of these dynamics over the years of diagnosis are different too