Analysing Cause-Specific Mortality Trends using Compositional Functional Data Analysis

Abstract

Overall mortality trends may be partially explained by cause-specific data. A recent example is provided by Woolf and Schoomaker (2019) who try to shed light on the decreasing trend of US life expectancy inspecting mortality by cause, finding that midlife mortality caused by drug overdoses, alcohol abuse, suicides and a diverse list of organ system diseases have particularly increased in the

Functional concurrent regression with compositional covariates and its application to the time-varying effect of causes of death on human longevity

“…Age classes Congenital anomalies (CONG) 0-4 Infancy related causes, excluded congenital anomalies (INFA) 0-4 Certain infectious and parasitic diseases (INFE) 0-4 5-39 40-64 65+ Neoplasms (NEOP) 0-4 5-39 40-64 65+ Respiratory diseases (RESP) 0-4 5-39 40-64 65+ External causes of death (EXT) 0-4 5-39 40-64 65+ Diseases of nervous system (NERV) 0-4 5-39 40-64 65+ Digestive system diseases (DIG) 5-39 40-64 65+ Mental disorders (MENT) 5-39 40-64 65+ Endocrine, nutritional and metabolic diseases (END) 5-39 40-64 65+ Circulatory system diseases (CIRC) 5-39 40-64 65+ Diseases of urogenital system (UROG) 40-64 65+ Lung cancer (LUNG) 40-64 65+ Diseases of skin, musculoskeletal system and connective tissue system (SKIN) 40-64 65+ We suggest that such an analysis can be performed by regressing the evolution of overall mortality (measured in terms of life expectancy at birth) with cause-of-death composition of mortality as defined by Stefanucci and Mazzuco (2022). Sun et al (2020) have recently proposed a log-contrast regression model with functional compositional covariates.…”

“…A primary issue is that the International Classification of Diseases (ICD) has changed significantly over the years, determining potentially biased results. Following Canudas-Romo, Adair and Mazzuco (2020) and Stefanucci and Mazzuco (2022), we use broad categories of causes, which are minimally affected by the classification revisions. The categories considered are shown in Table 1: the number of causes is higher with respect to Stefanucci and Mazzuco (2022), who limit their analysis to age group 40-64.…”

“…Following Canudas-Romo, Adair and Mazzuco (2020) and Stefanucci and Mazzuco (2022), we use broad categories of causes, which are minimally affected by the classification revisions. The categories considered are shown in Table 1: the number of causes is higher with respect to Stefanucci and Mazzuco (2022), who limit their analysis to age group 40-64. Here, we also consider causes that are specific to infant ages (e. g., congenital anomalies) and senescent ones (e. g., mental disorders, including dementia and Alzheimer's disease).…”

“…This kind of studies are paramount in public health, but handling causes of death is challenging, as in a competing risk perspective, cause-specific mortality can decline because there has been a significant improvement in treatment and/or prevention of that disease or just because other causes have grown meanwhile. This has recently been considered by Stefanucci and Mazzuco (2022), who propose a combination of Functional Data Analysis (FDA) and Compositional Data Analysis (CDA) to analyze the cause-of-death time pattern, limiting to mortality at age 40-64. Although the study by Stefanucci and Mazzuco (2022) provides useful insights on the evolution of cause-specific mortality, it remains of descriptive nature, while it might be of interest to measure if and to what extent different cause-of-death compositions drove the evolution of overall mortality in recent years.…”

“…This has recently been considered by Stefanucci and Mazzuco (2022), who propose a combination of Functional Data Analysis (FDA) and Compositional Data Analysis (CDA) to analyze the cause-of-death time pattern, limiting to mortality at age 40-64. Although the study by Stefanucci and Mazzuco (2022) provides useful insights on the evolution of cause-specific mortality, it remains of descriptive nature, while it might be of interest to measure if and to what extent different cause-of-death compositions drove the evolution of overall mortality in recent years.…”