Excess Mortality and Associated Risk Factors Related to Hurricane Maria in Puerto Rico

Publication Date: 2021

Executive Summary

Overview

Hurricane Maria, a category four hurricane, hit Puerto Rico on September 20, 2017. In the six months that followed, the number of people who died from the hurricane became a point of debate as the death toll was much higher than was initially reported. A number of studies were conducted to estimate how many people were killed, and the governor changed the official death toll from 64 to 2,975 deaths. The number of people killed by indirect causes related to the hurricane has been difficult to determine. People living with chronic diseases experienced difficulty accessing appropriate medical care particularly those who needed dialysis. Disruption of health care services and displacement of people whose homes had been destroyed by the hurricane led to the exacerbation of pre-existing chronic diseases, including diabetes, which is prevalent in Puerto Rico. Little has been done to identify which causes of death were most common or in excess following the hurricane. In this study we obtained death certificates from Puerto Rico’s Department of Health Demographic Registry to ascertain how many deaths occurred as a result of the hurricane during the six months following the disaster.

Research Questions and Design

Our study addresses the following research questions:

1. How many excess deaths occurred in Puerto Rico following Hurricane Maria from September 2017 to March 2018 that would not have happened if there had been no hurricane?
2. What sociodemographic factors are associated with higher risk of death in the six months following Hurricane Maria in Puerto Rico?
3. What were the most common causes of death in the six months that followed Hurricane Maria in Puerto Rico?
4. Were there geographical clusters of excess mortality in the six months following Hurricane Maria?

This study used death certificate data along with US Census Population Data and the CDC Social Vulnerability Index to answer these questions. Mortality data from September 13, 2016 to March 31, 2017 was compared to deaths that occurred from September 13, 2017 to March 31, 2018. We also obtained death certificates for the period one week prior to the hurricane to account for deaths due to injuries or other external causes. We grouped deaths into chronic disease-related categories to examine trends within larger categories of cause of death. Additionally, Poisson regression was used to compare differences between death rates at the municipality level controlling for month of death. The data analysis application SAS 9.4 was used to clean data, collect statistics on our population, and conduct Poisson regression to compare differences in death rates.

Preliminary Findings

Our study found that 18,510 death certificates were recorded in Puerto Rico from September 13th 2017 to March 31st 2018. The majority of deaths (54.8%) were among men. Most deaths were among those who were 65 years of age or older (76.5%) with only 0.6% of deaths occurring among people under 18 years of age. Most deaths were recorded as natural causes (88.8%), followed by undetermined cause (4.9%), followed by homicide (2.3%), accidents (3.2%), and suicide (0.8%). The most frequent underlying cause of death in our study was Alzheimer’s disease, followed by diabetes, acute myocardial infarction, heart disease, and sepsis.

Conclusions

Moving forward, we plan to examine mortality trends by geographical location and social vulnerability at the community level using municipality as the unit of analysis. We will work with colleagues at Ponce Health Sciences University and Puerto Rico Hurricane Response Hub to disseminate the findings of this study. Our final results will be shared with Puerto Rico’s health authorities for consideration in preparation for future hurricanes in Puerto Rico and to better protect the island’s most vulnerable populations.

Introduction

Hurricane Maria hit Puerto Rico on September 20th, 2017 and in the year that followed determining the number of people who died directly or indirectly from the hurricane was the subject of fierce debate, as the death toll was much higher than had been initally reported. Several studies were conducted to estimate the number of deaths, and as a result the governor changed the official death toll from 64 to 2,975 deaths (Florido, 2018¹; Santiago et al., 2018²). Many people living with chronic diseases experienced difficulty accessing appropriate medical care, particularly those who needed dialysis. There is a gap in the literature when it comes to quantifying deaths attributable to Hurricane Maria using actual death certificate data. Chronic disease deaths that were indirectly caused by the hurricane and deaths of people who were displaced from their homes due to hurricane related damage were particularly hard to identify.

The initial official report of 64 deaths following Hurricane Maria in Puerto Rico was shown to be a gross underestimate since the use of CDC’s certification of deaths related to disasters protocol was not implemented in Puerto Rico prior to Hurricane Maria (Centers for Disease Control and Prevention (CDC), 2017³). A study commissioned by the Puerto Rican government to the George Washington University Miliken Institute School of Public Health which led to the death toll being revised upward to 2,975 deaths attributable to Hurricane Maria (Florido, 2018; Santiago et al., 2018). Depending on the socioeconomic status and underlying health issues of people living in the areas most affected by a storm, hurricanes can lead to large increases in morbidity and mortality associated with infectious diseases, environmental hazards, injuries, and chronic diseases (Bourque et al., 2006⁴; Diaz, 2004⁵; Miller & Arquilla, 2008⁶). In Puerto Rico almost half of the population is covered by Medicaid (Macpac, 2019⁷) and many of them are living with at least one chronic disease (Division & Secretariat for Health Promotion, 2020⁸). The severe infrastructure damage caused by the hurricane eliminated access to the care that many patients needed to maintain control of their chronic diseases. In the months following Hurricane Maria, access to care continued to be limited and environmental hazards such as flooding and mold growth may have also contributed to death and disease among Puerto Ricans.

There are a wide range of estimates about how bad the impact of Hurricane Maria would be on death rates in the aftermath of the storm. Various methods were used to estimate mortality in the short and long term following Maria (George Washington University Miliken Institute School of Public Health, 2018⁹; Kishore et al., 2018¹⁰; Santos-Burgoa et al., 2018¹¹). One study that summarized all of the estimates of post-Maria mortality in Puerto Rico stated that estimates of excess mortality for September and October 2017 ranges from 1023 to 1077 more deaths than normally expected for that time period (Sandberg et al., 2019¹²). Little has been done to identify mortality factors following Maria and common causes of death. Measuring excess mortality can be an effective way to assess disaster recovery and to identify the health risks that municipalities face following a hurricane. It is also useful in preparing for future hurricanes. This study seeks to create a more accurate measurement of mortality in Puerto Rico in the six months following Hurricane Maria.

Methods

Research Questions

Our study addresses the following research questions:

1. How many excess deaths occurred in Puerto Rico following Hurricane Maria from September 2017 to March 2018 that would not have happened if there had been no hurricane?
2. What sociodemographic factors are associated with higher risk of death in the six months following Hurricane Maria in Puerto Rico?
3. What were the most common causes of death in the six months that followed Hurricane Maria in Puerto Rico?
4. Were there geographical clusters of excess mortality in the six months following Hurricane Maria?

Study Site Description

This study used data from Puerto Rico’s Department of Health Demographic Registry. Puerto Rico is a U.S. territory that was impacted by Hurricanes Maria and Irma in September 2017. Following Hurricane Maria, there were discrepancies in the number of deaths that were attributed to the hurricanes. The Centers for Disease Control and Prevention provides guidance for properly using death certificates to document deaths attributed to a disaster. A study conducted after Hurricane Maria found that many physicians were not aware of appropriate post-natural disaster death certification practices (George Washington University Miliken Institute School of Public Health, 2018). Lack of death certificate documentation made data access more difficult in Puerto Rico than in other locations. The Hurricane Response Hub and the Office of Vital Statistics worked with us to get the data. Due to these difficulties, we only gained access to the death certificates near the end of the time period for conducting the study. The data, however, had been uploaded into a database making it much easier to use than expected and scanning the death certificates was not required.

Data, Methods, and Procedures

We used the death certificate database kept by Puerto Rico’s Department of Health Demographic Registry. Mortality data from September 13, 2016 to March 31, 2017 was compared to deaths that occurred from September 13, 2017 to March 31, 2018. We also obtained death certificates for the period one week prior to the hurricane to account for deaths due to injuries or other external causes. The death certificate database includes sociodemographic information such as age, sex, race, marital status, place of residence, residential area, place of death, education, occupation, type of death, and Hispanic origin among others. In addition, the database contains important information for public health officials such as the variables underlying the immediate cause of death.

Using this information we developed a specific variable that regrouped the causes of deaths due to the most common chronic health conditions prevalent in Puerto Rico. We developed another specific variable that regrouped all causes of death using a list developed by the National Center for Health Statistics (NCHS). We also created independent variables using the information collected in part two of the death certificate where all additional information on the significant health conditions of the deceased is recorded.

After obtaining data from the same months for the year prior to Hurricane Maria, Poisson regression was used to compare differences between death rates at the municipality level controlling for month of death. Over the coming months, we will use publicly available U.S. Census Population data and Social Vulnerability Index data at the census tract level to further investigate how social vulnerability affected death rates in the period after hurricane Maria.

Sample Size and Participants

All deaths that occurred in Puerto Rico from September 13, 2016 to March 31, 2017 and from September 13, 2017 to March 31, 2017 were included in our dataset. Our final dataset included 35,370 deaths.

Data Analysis

The data analysis application SAS 9.4 was used to clean data, collect statistics on our population, and conduct Poisson regression to compare differences in death counts. Further analysis is planned for the coming months that will include examining sociodemographic factors and geographic clusters associated with higher risk of death following Hurricane Maria.

Researcher Positionality, Reciprocity, and Other Ethical Considerations

As researchers, we are influenced by present and past experiences and recognize that our positionality impacts the way we conduct research. Many dimensions of our identities grant us privileges in life that others do not have. Our privilege has shaped how we view experiences that affect communities of color, those with disabilities, aging populations, and those who practice other religions. With this being said, we have worked in disaster epidemiology and have seen the way that disasters disproportionately affect socially vulnerable communities. It is important to us to investigate how social vulnerabilities put communities at risk of suffering higher death rates following a hurricane. This project is not as community-based as other social sciences research because we are only viewing de-identified death certificates. As we continue our analysis, we intend to work with the Hurricane Response Hub to distribute our research findings to others working to prepare Puerto Rico for hurricanes so that future deaths can be prevented.

Findings

Results

Overall, there were 1,650 more deaths in Puerto Rico during the six months following Hurricane Maria than during a similar time period a year before. We obtained a total of 16,860 death certificates for the period from September 13, 2016 to March 31, 2017 and 18,510 death certificates were recorded for the same time period in 2017–2018.

The majority of deaths (54.9%) were among men during the post-Maria period. Most of the dead were 65 years of age or older (76.5%). Only 0.6% of deaths occurred among people under 18 years of age. Most deaths were attributed to natural causes (88.9%), followed by undetermined cause (4.9%), homicide, (2.3%), accidents (3.2%), and suicide (0.8%) (Table 1). An increase of 100 deaths due to Alzheimer’s disease, diabetes mellitus, sepsis, chronic respiratory disease, and hypertension occurred after Maria. Minimal differences were found for asthma deaths while there were 126 fewer cancer deaths in the post-Maria period. A significant increase in the mortality rate after Hurricane Maria compared with before Hurricane Maria was seen for all disease categories examined excluding COPD, Mental Health Conditions, Asthma and cancer. (Table 2).

Table 3 shows these differences by major diasease categories in further detail broken down by month. We found greater differences in the two-to-three months immediately after Hurricane Maria compared with the pre-Maria period.

Using poisson regression to compare the rate of deaths within municiaplities before and after Hurricane Maria controlling for month of year, the rate of death per 100,000 was slightly higher post-Maria compared to before Hurricane Maria (risk ratio=1.14, 95% CI:1.00-1.17). This is similar to the results presented across all of Puerto Rico in table 2 showing slightly higher mortality rate after Hurricane Maria compared with before (risk ratio=1.13, 95% CI: (1.11, 1.15).

Discussion of Findings

The initial findings suggests an excess number of deaths occurred in the post-Maria period compared to a similar period a year before the hurricane. The number of excess deaths when compared by major disease categories warrants further analysis to determine if these deaths had a greater number of contributing factors (i.e. other chronic diseases or conditions) than deaths during the one-year period before Hurricane Maria.

In the future months, we plan to examine the distribution of deaths to determine if there are geographical clusters. We also intend to use CDC’s Social Vulnerability Index (Agency for Toxic Substances and Disease Registry (ATSDR, 2020¹³) to determine if there is an association between more vulnerable communities experiencing higher levels of mortality following Hurricane Maria.

Conclusions

Key Findings

Our key findings describe the post-hurricane death toll in Puerto Rico. In the six months examined during 2017–2018, there were 18,510 deaths recorded in Puerto Rico compared with 16,860 deaths in the pre-Maria period of 2016–2017. Our study showed that there is limited documentation in the post-disaster death records. Using the narrative information of the death certificate and underlying disease categorization we were better able to determine if deaths due to chronic disease could be indirectly attributed to Hurricane Maria.

Implications for Public Health Practice

Our final results will show which sociodemographic groups and geographical areas need to be targeted in hurricane preparedness plans. The results from this study can also inform efforts to better communicate proper post-disaster death certificate methods to medical professionals.

Dissemination of Findings

We plan to work with our colleagues at Ponce Health Sciences University and the Puerto Rico Hurricane Response Hub to present our findings to the Puerto Rico Department of Health and to help prepare for future hurricanes. Identifying the underlying conditions that may put individuals at high risk for illness or death following a hurricane will be useful in developing risk communications targeting these specific groups. We will also work with the Hurricane Response Hub to identify others involved in hurricane preparedness planning to make them aware of which geographical areas and sociodemographic groups are at the greatest risk of suffering higher death rates after a hurricane. Furthermore, we intend to publish findings from this study so that others who work in disaster epidemiology may replicate our methods.

Limitations

Due to a variety of issues that arose while we were forming an agreement with the Demographic Registry, it took longer than expected to obtain the mortality data.

Because the data is de-identified, our study does not involve any methods to collect more narrative information. Additionally, death certificate data is subject to the biases of whoever fills out the death certificate. It is possible that the increased workload and lack of medical staff led to death certificates being filled in with less detail in the months immediately following Hurricane Maria.

Another limitation of this study is that it does not capture deaths of people who left Puerto Rico after Hurricane Maria. It is possible that people with serious chronic diseases left Puerto Rico due to Hurricane Maria, and those that died outside of the territory would not be captured in this dataset.

Future Research Directions

We have many plans to continue our research in the coming months. Primarily, we have now obtained population data at the municipality level so we are comparing rates of specific causes of death within municipalities for more common causes of death. We would also like to match the death record data with social vulnerability data at the municipality level to learn more about the factors associated with excess mortality. We are still cross-referencing the census tract level social vulnerability index with the death certificate data. It is also possible that rates of excess mortality may cluster geographically based on the location of infrastructure damage. We plan to conduct geospatial analyses to determine whether geography was associated with mortality following Hurricane Maria. It would also be helpful to identify data sources for environmental exposures such as electrical outages, flooding, or mold growth determine whether these factors contributed to excess mortality following the storm.

References