c2-d2 / pr_mort_official

Data and pipeline for Mortality in Puerto Rico after Hurricane Maria

Introduction

Here we provide the data and pipeline for: Mortality in Puerto Rico after Hurricane Maria

Citation

Kishore N, Marqués D, Mahmud A, et al. Mortality in Puerto Rico after Hurricane Maria. N Engl J Med. DOI: 10.1056/NEJMsa1803972

Additional Resources

1. FAQs [PDF] — (Spanish [PDF])
2. Responses to Inquiries [PDF]
3. Technical FAQ [PDF]

Repository at time of publication

This repository is constantly being updated in response to feedback and inquiries; however, all code will remain entirely reproducible at any point in the commit history.

For full transparency, we wanted to note what the repository looked like before we made additional changes. Thus, the paper release is the version of the repository that existed at the original time of publication. You can get this release by downloading it or using git checkout paper in your local repository.

Abstract

Background: Quantifying the effect on society of natural disasters is critical for recovery of public health services and infrastructure. The death toll can be difficult to assess in the aftermath of a major disaster. In September 2017, Hurricane Maria caused massive infrastructural damage to Puerto Rico, but its effect on mortality remains contentious.

Methods: Using a representative, stratified sample, we surveyed 3299 randomly chosen households across Puerto Rico to produce an independent estimate of all-cause mortality after the hurricane. Respondents were asked about displacement, infrastructure loss, and causes of death. We calculated excess deaths by comparing our estimated post-hurricane mortality rate with official rates for the same period in 2016.

Results: From the survey data, we estimated a mortality rate of 14.3 deaths (95% confidence interval [CI], 9.8 to 18.9) per 1000 persons from September 20 through December 31, 2017. This rate yielded a total of 4645 excess deaths during this period (95% CI, 793 to 8498), equivalent to a 62% increase in the mortality rate as compared with the same period in 2016. However, this number is likely to be an underestimate because of survivor bias. The mortality rate remained high through the end of December 2017, and one third of the deaths were attributed to delayed or interrupted health care. Hurricane-related migration was substantial.

Conclusions: The official estimate of 64 deaths attributed to Hurricane Maria in Puerto Rico is a substantial underestimate. This survey, based on community sampling, indicated that the number of excess deaths is likely to be more than 70 times the official estimate. (Funded by the Harvard T.H. Chan School of Public Health and others.)

a) A comparison of excess death estimates from official reports, press/academic reports, and our survey. b) Reported deaths per month in the survey, categorized by reported cause of death. Two individuals who died of similar causes are superimposed in December who died at the same age resulting in a count of 37 points representing 38 deaths after the hurricane.

Organization

• code — Scripts and output for figures included in the manuscript and supplement
• data — Initial data resources and survey data (more information in the folder README )
• figures — Final figures included in manuscript
• ref — pipelines used to clean raw information and generate .RDS files found in /data/rdata/
• misc — Folder with a Spanish version of the paper as well as an xml file with the survey instrument used to gather data.

Use

• master is locked
• We have tagged the version of this repository at the time of publication under the paper release
• Feel free to create a new branch for further incorporation and analysis
• All geospatial data has been stripped; by using this dataset, you agree to not undertake any steps to identify any respondents or their families
• More information in data

Correspondence

For any issues with anonymization or major issues with the functionality of the script please create an issue.

License

The data collected and presented is licensed under the Creative Commons Attribution 3.0 license, and the underlying code used to format, analyze and display that content is licensed under the MIT license.

pr_mort_official's Issues

The deaths_official data frame has "Pop" as a character variable

I don't think it hurts anything, but this is probably not what you want. The mistake probably originates in /resources/deaths_official.csv, which I can't see. (And that is OK! I assume that the resources directory includes non-public data.)

adjusted_rates and adjusted_rates2 are identical

all.equal(adjusted_rates, adjusted_rates2)
[1] TRUE

I was not sure if this was intentional. Feel free to close if it is.

The state of the repository *at publication*

This repository is changing constantly in response to questions and feedback from others; however, I think it's worth documenting (somewhere) the commit at the time of publication is: e43ffc4

You can browse the files of this repository and all its code/data/etc in this "original state" via
https://github.com/c2-d2/pr_mort_official/tree/e43ffc43c2d1844a91e2570b1cb570cc27a6262d

Population Size Discrepancies

Hello,

I would like to know why on page 7 of the appendix, it says that the number used as a population size (3,030,307) to estimate the excess death, but the excess_est.R file, on line 39 uses 3,405,520. This yields different results for the excess death number.

Thank You

Is the code to replicate Table S4 available?

The R scripts in the code/ directory allows me to replicate Tables S1, S2 and S3. But I don't see anything for Table S4. Perhaps I am looking in the wrong place? I am most interested in replicating the 14.31 [9.76, 18.86] estimate for the annual mortality rate post-Maria. My attempts to do so "by hand" have been unsuccessful so far.

Confidence interval definition

I do believe the definition of a confidence interval is not technically correct: "has a 95% chance of including the actual death count." This particular CI either contains the population value or it does not. If the survey were repeated 100 times, 95 of those would contain the population value. So we may say we can be 95% confident the interval contains the population value. An excellent visualization of the issue around interpreting CIs at http://rpsychologist.com/d3/CI/

Impossibly low pre-Maria mortality rate

I calculate a mortality rate of 2.6 deaths (95% confidence interval [CI], 1.4 to 3.8)) per 1000 persons from January 1 through September 19, 2017. This mortality rate is inconsistent with the rate calculated from the official monthly statistics for 2010 through 2017: 8.3 deaths (95% CI, 8.2 to 8.4) per 1000 persons. It is, statistically, almost impossible for there to have been only 18 deaths in the 3299 households prior to Maria. The problem remains even with the authors’ later calculations adjusting for household size.

Two possible errors in code

To get the number of deaths in 2016 Line 45 of excess_est reads

mutate(Sep = Sep*(1/3)) %>% sum

But to get the deaths in 2016 from September 20 to December 31, 11 days of September should be considered, not 10. Thus the line should be

mutate(Sep = Sep*(11/30)) %>% sum

Also, all 102/365 should be 103/365. The general conclusion of the article will not change, but point estimates and interval bounds will be shifted down by 70.