A Flexible Statistical Framework for Estimating Excess Mortality

Abstract

Quantifying the impact of natural disasters or epidemics is critical for guiding policy decisions and interventions. When the effects of an event are long-lasting and difficult to detect in the short term, the accumulated effects can be devastating. Mortality is one of the most reliably measured health outcomes, partly due to its unambiguous definition. As a result, excess mortality estimates are an increasingly effective approach for quantifying the effect of an event. However, the fact that indirect effects are often characterized by small, but enduring, increases in mortality rates present a statistical challenge. This is compounded by sources of variability introduced by demographic changes, secular trends, seasonal and day of the week effects, and natural variation. Here, we present a model that accounts for these sources of variability and characterizes concerning increases in mortality rates with smooth functions of time that provide statistical power. The model permits discontinuities in the smooth functions to model sudden increases due to direct effects. We implement a flexible estimation approach that permits both surveillance of concerning increases in mortality rates and careful characterization of the effect of a past event. We demonstrate our tools' utility by estimating excess mortality after hurricanes in the United States and Puerto Rico. We use Hurricane Maria as a case study to show appealing properties that are unique to our method compared with current approaches. Finally, we show the flexibility of our approach by detecting and quantifying the 2014 Chikungunya outbreak in Puerto Rico and the COVID-19 pandemic in the United States. We make our tools available through the excessmort R package available from https://cran.r-project.org/web/packages/excessmort/.

Comparing methods to predict baseline mortality for excess mortality calculations

“…Four methods will be used for predicting mortality, including the WHO’s method, an advanced alternative method that also uses splines, developed by Acosta and Irizarry in 2020 [34], and two naive methods as a comparison. These cover the widely used, classical statistical methods used for predicting baseline mortality in excess mortality studies.…”

“…Acosta-Irizarry (AI) method: the method described in [34] using their reference implementation. Parameters: starting year (from which the model is fitted) and tkpy , the number of trend knots per year; other parameters are left on their default values (i.e., two harmonic term is used).…”

“…Parameters: starting year (from which the model is fitted) and k, the dimension of the basis of the spline used for capturing the long-term trend. • Acosta-Irizarry method: the method described in [34] using their reference implementation. Parameters: starting year (from which the model is fitted) and tkpy, the number of trend knots per year; other parameters are left on their default values (i.e., two harmonic term is used).…”

“…Using splines [32,33]. The method of Acosta and Irizarry [34,35] is based on splines, just as many other custom implementation [16,36], which, crucially, includes the model used by the World Health Organization (WHO) [37].…”

All-cause excess mortality across 90 municipalities in Gujarat, India, during the COVID-19 pandemic (March 2020-April 2021)

“…In the absence of reliable death registration data in the aftermath of disasters and public health emergencies, scientists have relied on alternative methods to estimate deaths, including household based surveys, crematoria and funeral home body counts, or verbal autopsies [12][13][14][15]. In many countries, the estimation of all-cause mortality has provided an alternative proxy for the underestimation of COVID-19 attributable deaths in official statistics, whereby total observed deaths are compared to expected deaths computed from historical baselines [16][17][18][19][20][21]. These estimates include directly attributable deaths (those that died from SARS-Cov-2 and its complications) and indirectly attributable deaths (those that died from the indirect impacts of the pandemic, like disruptions in access to livelihoods, food security, public assistance, preventive health interventions, or medical care) [22,23].…”