[Source: Journal of American Medical Association, full page: (LINK). Abstract, edited.]
Research Letter / August 2, 2018
Use of Death Counts From Vital Statistics to Calculate Excess Deaths in Puerto Rico Following Hurricane Maria
Alexis R. Santos-Lozada, PhD1; Jeffrey T. Howard, PhD2
Author Affiliations: 1 Department of Human Development and Family Studies, Pennsylvania State University, University Park; 2 Department of Kinesiology, Health and Nutrition, University of Texas at San Antonio
JAMA. Published online August 2, 2018. doi:10.1001/jama.2018.10929
The official death toll for Hurricane Maria, which devastated Puerto Rico on September 20, 2017, has remained at 64 since December 29, 2017. Accurate estimates of deaths from environmental disasters are important for informing rescue, recovery, and policy decisions.
Using preliminary death counts through October 2017, excess deaths related to the hurricane were estimated at 1085.1 However, other estimates suggest that the number of excess deaths may be as high as 4645.2 The variance in estimates is due to differences in methodology. The official government death toll includes only deaths in which documentation of “hurricane-related” as the cause of death appears on the individual’s death certificate and does not account for indirect deaths, including from infectious disease outbreaks or lack of services (such as electricity, water, and medical care). Estimates of excess deaths address both direct and indirect deaths and typically use either death counts from government agencies1 or surveys,2 which are susceptible to larger margins of error. We calculated the number of excess deaths following Hurricane Maria through December 2017, using death counts from vital statistics records, updating a previous estimate.1
Monthly death counts, from January 2010 through December 2017, including previously unavailable death counts for January through December 2017, were obtained from the Puerto Rico vital statistics system to calculate excess deaths in Puerto Rico following Hurricane Maria; this system has a 99% coverage rate based on previous analyses.3 Because these data are deidentified aggregate counts of deceased individuals, this study is considered to be research not involving human subjects as defined by US regulation (45 CFR 46.102[d]).
Consistent with prior studies,4,5 death counts from vital records from 2010 through 2016 were used to establish expected monthly deaths (mean), and historical ranges of natural variability (95% CIs). For September through December 2017, we used the difference between number of deaths from vital statistics records and the upper 95% CI bound as a measure of excess mortality. The upper 95% CI limits were calculated as: … where x̅ is the mean value, t is the 2-sided value of the student t-distribution, α is the alpha level (.05), n is the number of observations (n = 7), and s is the standard error of the mean. The 1-sample t test was used to calculate 95% CIs around the excess death estimates. Analyses were conducted using SAS (SAS Institute), version 9.4.
The expected numbers of deaths were 2396 in September, 2407 in October, 2416 in November, and 2697 in December (Table). The upper bounds (95% CIs) for the same months were 2469, 2476, 2555, and 2824, respectively, whereas the actual numbers of deaths were 2928, 3040, 2671, and 2820. There were 1139 (95% CI, 1006-1272) excess deaths: 459 (95% CI, 425-593) in September, 564 (95% CI, 531-597) in October, and 116 (95% CI, 50-182) in November. Although August and July experienced lower numbers of deaths than expected, September and October had higher numbers of deaths than expected, the number of deaths decreased in November, and by December had returned to a level within historical variation (Figure).
Based on death records following Hurricane Maria, the estimated hurricane-related mortality burden of 1139 excess deaths through December 2017 was higher than the official death toll of 64. The estimate is conservative, because the expected number of deaths used the upper 95% CI and did not consider the population denominators, which were decreasing. The strength of the present approach is that it is based on death counts from vital statistics records and is consistent with previous estimates1 and methods.4,5 The primary limitation of the study is that the specific cause of each individual death is not known; thus only an aggregate number of deaths in excess of historical variation can be estimated. Another recent study2 suggested that there were 4645 excess deaths (95% CI, 783-8498), but it was based on a survey that underestimated prehurricane mortality, overestimated posthurricane mortality, and had a large CI, indicating a high level of uncertainty. Future studies would benefit from careful analysis of deaths from vital records rather than surveys.
Section Editor: Jody W. Zylke, MD, Deputy Editor.
Keywords: Puerto Rico; Public Health; Hurricanes; Environmental Disasters.