Hurricane Maria killed more people than did Hurricane Katrina.
Although government officials initially reported 64 deaths due to the hurricane, independent analyses have estimated that Maria claimed thousands of victims. The discrepancies in the estimates can be traced to the methods used. Poor estimates have serious ramifications such as insufficient aid for disaster relief.
This post is based on the CBS News article, Hurricane Maria Caused an Estimated 2,975 Deaths in Puerto Rico, New Study Finds, by S. Lynch Baldwin and D. Begnaud, August 28, 2018. The videos shared in the spotlight are embedded in this article. Image source: Stocktrek Images/Getty Images.
1. What is a forecast? Would you define the estimate of 2,975 deaths as a forecast?
Guidance: Review the definition of a forecast. Since the deaths have already occurred, the estimate is not a forecast. However, the estimate is based on a prediction of the number of deaths if the hurricane had not struck the island. This prediction is derived using conventional forecasting techniques.
2. Why was the initial report of 64 deaths so inaccurate?
Guidance: The initial data only accounted for deaths directly attributable to the storm itself, not its aftermath. Many people died due to a lack of basic services such as electricity, drinkable water, and delayed or prevented access to healthcare (most hospitals did not have electricity and enough basic medical supplies). Moreover, the report relied on death certification by physicians and forensic physicians. They did not have adequate training and information about death certification processes in case of natural disasters. Finally, a lack of communication among local government agencies is also responsible for the inaccurate documentation of deaths. Bad data produce poor estimates (GIGO).
3. Besides the initial report of 64 deaths, the article cites three estimates of the number of deaths attributable to Hurricane Maria. Why do the estimates differ?
Guidance: Discuss how different data collection methods and predictive analytics can produce different estimates.
- The Harvard University study estimated more than 4,600 deaths. The estimate was generated using a stratified sample of 3,299 households, estimated the mortality rate, and compared it with official rates for the same period.
- Puerto Rico’s report to Congress provided an estimate of 1,427 deaths after Hurricanes Irma and Maria, but it cautioned that the deaths “may or may not be attributable to the hurricanes.” It is based on a comparison of the number of deaths before and after the hurricanes.
- The George Washington University’s Milken Institute School of Public Health report, featured in the article, analyzed mortality patterns from 2010 to 2017 to predict the mortality rate if Hurricane Maria had not occurred. Regression-type models were used. That number was subtracted from the observed mortality rate after Maria.
4. To estimate the number of deaths for the pre-hurricane period of 2010-2017, the George Washington University study relied on generalized linear models similar in concept to the multiple regression model described in your textbook. What independent variables are likely predictors of mortality?
Guidance: Variables may include population size, age, place of residence, socio-economic status, population displacement, and more.