ASA Connect

Dear Giants of Statistics,

I would appreciate your statistical input to me, an epidemiologist.

   A few days ago Trump indicated that there were only 64 deaths my Maria in Puerto Rico, despite two studies suggesting 2715 (GW) and 4645 (Harvard).   I became interested in this huge variation.  Quite clearly the number of deaths exceed 64.  However, as I dug deeper in the other studies, it seemed that the GW and Harvard estimates were very high.  I wanted to lay out my finding to see what you think, and what type of studies could be done to get better estimates.  What I discovered is that the estimates of "Death by Hurricane" are extremely variable.  Here are the GW and Harvard estimates. Both have problems

https://publichealth.gwu.edu/sites/default/files/downloads/projects/PRstudy/Acertainment%20of%20the%20Estimated%20Excess%20Mortality%20from%20Hurricane%20Maria%20in%20Puerto%20Rico.pdf

https://www.nejm.org/doi/full/10.1056/NEJMsa1803972

Here is what stimulated my interest

Hurricane          # Deaths

Puerto RIco          4645

Puerto RIco           2715

Katrina                   283-1836

Audrey                    418

Sandy                      230

Harvey                     82

Hugo                        34

florence                     5

I find it very hard to believe that maria was 2-3 times as dangerous a Katrina.  

2.  Previous studies mentioned that 40 to 70% of the deaths from Hurricanes are due to drowning due to the storm surges.  The numbers of deaths for drowning is very low in Puerto  rico, I could not find information find what happened after the hurricane, but it is hear where most of the excess deaths would be seen. Even if there was a 10 fold increase in drowning death, this would not bring the estimates even close to ~3000

3.  the GW study postulates that the excess deaths were not found because of poor death certificate coding.  This is difficult to believe, as the increase in deaths would be located with drowning and injuries, both of which would be picked up. Also, for this was the cause, then 99% of the deaths due to the hurricane would have to be miscoded to account for the excess. The PR health department of doctors would never be this bad.  Finally, if there were 3000-5000  deaths during this period it would be obvious as as people would be "dying in the streets", something we saw with Katrina.  THis really does not make sense

4. A primary reason for this is that it is difficult to classify a hurricane related death.  This is an interesting article indicating that we still do not have a firm figure about deaths .  https://fivethirtyeight.com/features/we-still-dont-know-how-many-people-died-because-of-katrina/?ex_cid=538twitter  Much of the variation can be explained through this.

6.  The approached used in the GW study are suspicious for me as an epidemiologist. They estimate the rates and numbers of total mortality before the hurricane, and then during it.  This type of analysis is call an ecologic analysis, and quite problematic for epidemiologist like me.   THe reason is that this is a pure correlation analysis and not adequate for testing specific hypotheses.    Other things could be changing rapidly as well, e.g. the opiate epidemic, rapidly changing economic conditions.  I will not feel comfortable until I see the changes in the cause specific deaths, especially the change in drowning death.

Another issue is that there was a suggestion of increased mortality due to diabetes.  I am one of the world's experts on diabetes epidemiology, and seriously doubt that the storm could speed the death of people with diabetes. I do not know of any data to support this.

Thus I agree that the number of deaths is greater than 64.  However, I really find it hard to believe that this was 2-3 times as deadly at Katrina

I would love your thoughts

ron

Dear Giants of Statistics,

I would appreciate your statistical input to me, an epidemiologist.

   A few days ago Trump indicated that there were only 64 deaths my Maria in Puerto Rico, despite two studies suggesting 2715 (GW) and 4645 (Harvard).   I became interested in this huge variation.  Quite clearly the number of deaths exceed 64.  However, as I dug deeper in the other studies, it seemed that the GW and Harvard estimates were very high.  I wanted to lay out my finding to see what you think, and what type of studies could be done to get better estimates.  What I discovered is that the estimates of "Death by Hurricane" are extremely variable.  Here are the GW and Harvard estimates. Both have problems

https://publichealth.gwu.edu/sites/default/files/downloads/projects/PRstudy/Acertainment%20of%20the%20Estimated%20Excess%20Mortality%20from%20Hurricane%20Maria%20in%20Puerto%20Rico.pdf

https://www.nejm.org/doi/full/10.1056/NEJMsa1803972

Here is what stimulated my interest

Hurricane          # Deaths

Puerto RIco          4645

Puerto RIco           2715

Katrina                   283-1836

Audrey                    418

Sandy                      230

Harvey                     82

Hugo                        34

florence                     5

I find it very hard to believe that maria was 2-3 times as dangerous a Katrina.  

2.  Previous studies mentioned that 40 to 70% of the deaths from Hurricanes are due to drowning due to the storm surges.  The numbers of deaths for drowning is very low in Puerto  rico, I could not find information find what happened after the hurricane, but it is hear where most of the excess deaths would be seen. Even if there was a 10 fold increase in drowning death, this would not bring the estimates even close to ~3000

3.  the GW study postulates that the excess deaths were not found because of poor death certificate coding.  This is difficult to believe, as the increase in deaths would be located with drowning and injuries, both of which would be picked up. Also, for this was the cause, then 99% of the deaths due to the hurricane would have to be miscoded to account for the excess. The PR health department of doctors would never be this bad.  Finally, if there were 3000-5000  deaths during this period it would be obvious as as people would be "dying in the streets", something we saw with Katrina.  THis really does not make sense

4. A primary reason for this is that it is difficult to classify a hurricane related death.  This is an interesting article indicating that we still do not have a firm figure about deaths .  https://fivethirtyeight.com/features/we-still-dont-know-how-many-people-died-because-of-katrina/?ex_cid=538twitter  Much of the variation can be explained through this.

6.  The approached used in the GW study are suspicious for me as an epidemiologist. They estimate the rates and numbers of total mortality before the hurricane, and then during it.  This type of analysis is call an ecologic analysis, and quite problematic for epidemiologist like me.   THe reason is that this is a pure correlation analysis and not adequate for testing specific hypotheses.    Other things could be changing rapidly as well, e.g. the opiate epidemic, rapidly changing economic conditions.  I will not feel comfortable until I see the changes in the cause specific deaths, especially the change in drowning death.

Another issue is that there was a suggestion of increased mortality due to diabetes.  I am one of the world's experts on diabetes epidemiology, and seriously doubt that the storm could speed the death of people with diabetes. I do not know of any data to support this.

Thus I agree that the number of deaths is greater than 64.  However, I really find it hard to believe that this was 2-3 times as deadly at Katrina

I would love your thoughts

ron

Dear Giants of Statistics,

I would appreciate your statistical input to me, an epidemiologist.

   A few days ago Trump indicated that there were only 64 deaths my Maria in Puerto Rico, despite two studies suggesting 2715 (GW) and 4645 (Harvard).   I became interested in this huge variation.  Quite clearly the number of deaths exceed 64.  However, as I dug deeper in the other studies, it seemed that the GW and Harvard estimates were very high.  I wanted to lay out my finding to see what you think, and what type of studies could be done to get better estimates.  What I discovered is that the estimates of "Death by Hurricane" are extremely variable.  Here are the GW and Harvard estimates. Both have problems

https://publichealth.gwu.edu/sites/default/files/downloads/projects/PRstudy/Acertainment%20of%20the%20Estimated%20Excess%20Mortality%20from%20Hurricane%20Maria%20in%20Puerto%20Rico.pdf

https://www.nejm.org/doi/full/10.1056/NEJMsa1803972

Here is what stimulated my interest

Hurricane          # Deaths

Puerto RIco          4645

Puerto RIco           2715

Katrina                   283-1836

Audrey                    418

Sandy                      230

Harvey                     82

Hugo                        34

florence                     5

I find it very hard to believe that maria was 2-3 times as dangerous a Katrina.  

2.  Previous studies mentioned that 40 to 70% of the deaths from Hurricanes are due to drowning due to the storm surges.  The numbers of deaths for drowning is very low in Puerto  rico, I could not find information find what happened after the hurricane, but it is hear where most of the excess deaths would be seen. Even if there was a 10 fold increase in drowning death, this would not bring the estimates even close to ~3000

3.  the GW study postulates that the excess deaths were not found because of poor death certificate coding.  This is difficult to believe, as the increase in deaths would be located with drowning and injuries, both of which would be picked up. Also, for this was the cause, then 99% of the deaths due to the hurricane would have to be miscoded to account for the excess. The PR health department of doctors would never be this bad.  Finally, if there were 3000-5000  deaths during this period it would be obvious as as people would be "dying in the streets", something we saw with Katrina.  THis really does not make sense

4. A primary reason for this is that it is difficult to classify a hurricane related death.  This is an interesting article indicating that we still do not have a firm figure about deaths .  https://fivethirtyeight.com/features/we-still-dont-know-how-many-people-died-because-of-katrina/?ex_cid=538twitter  Much of the variation can be explained through this.

6.  The approached used in the GW study are suspicious for me as an epidemiologist. They estimate the rates and numbers of total mortality before the hurricane, and then during it.  This type of analysis is call an ecologic analysis, and quite problematic for epidemiologist like me.   THe reason is that this is a pure correlation analysis and not adequate for testing specific hypotheses.    Other things could be changing rapidly as well, e.g. the opiate epidemic, rapidly changing economic conditions.  I will not feel comfortable until I see the changes in the cause specific deaths, especially the change in drowning death.

Another issue is that there was a suggestion of increased mortality due to diabetes.  I am one of the world's experts on diabetes epidemiology, and seriously doubt that the storm could speed the death of people with diabetes. I do not know of any data to support this.

Thus I agree that the number of deaths is greater than 64.  However, I really find it hard to believe that this was 2-3 times as deadly at Katrina

I would love your thoughts

ron

Dear Giants of Statistics,

I would appreciate your statistical input to me, an epidemiologist.

   A few days ago Trump indicated that there were only 64 deaths my Maria in Puerto Rico, despite two studies suggesting 2715 (GW) and 4645 (Harvard).   I became interested in this huge variation.  Quite clearly the number of deaths exceed 64.  However, as I dug deeper in the other studies, it seemed that the GW and Harvard estimates were very high.  I wanted to lay out my finding to see what you think, and what type of studies could be done to get better estimates.  What I discovered is that the estimates of "Death by Hurricane" are extremely variable.  Here are the GW and Harvard estimates. Both have problems

https://publichealth.gwu.edu/sites/default/files/downloads/projects/PRstudy/Acertainment%20of%20the%20Estimated%20Excess%20Mortality%20from%20Hurricane%20Maria%20in%20Puerto%20Rico.pdf

https://www.nejm.org/doi/full/10.1056/NEJMsa1803972

Here is what stimulated my interest

Hurricane          # Deaths

Puerto RIco          4645

Puerto RIco           2715

Katrina                   283-1836

Audrey                    418

Sandy                      230

Harvey                     82

Hugo                        34

florence                     5

I find it very hard to believe that maria was 2-3 times as dangerous a Katrina.  

2.  Previous studies mentioned that 40 to 70% of the deaths from Hurricanes are due to drowning due to the storm surges.  The numbers of deaths for drowning is very low in Puerto  rico, I could not find information find what happened after the hurricane, but it is hear where most of the excess deaths would be seen. Even if there was a 10 fold increase in drowning death, this would not bring the estimates even close to ~3000

3.  the GW study postulates that the excess deaths were not found because of poor death certificate coding.  This is difficult to believe, as the increase in deaths would be located with drowning and injuries, both of which would be picked up. Also, for this was the cause, then 99% of the deaths due to the hurricane would have to be miscoded to account for the excess. The PR health department of doctors would never be this bad.  Finally, if there were 3000-5000  deaths during this period it would be obvious as as people would be "dying in the streets", something we saw with Katrina.  THis really does not make sense

4. A primary reason for this is that it is difficult to classify a hurricane related death.  This is an interesting article indicating that we still do not have a firm figure about deaths .  https://fivethirtyeight.com/features/we-still-dont-know-how-many-people-died-because-of-katrina/?ex_cid=538twitter  Much of the variation can be explained through this.

6.  The approached used in the GW study are suspicious for me as an epidemiologist. They estimate the rates and numbers of total mortality before the hurricane, and then during it.  This type of analysis is call an ecologic analysis, and quite problematic for epidemiologist like me.   THe reason is that this is a pure correlation analysis and not adequate for testing specific hypotheses.    Other things could be changing rapidly as well, e.g. the opiate epidemic, rapidly changing economic conditions.  I will not feel comfortable until I see the changes in the cause specific deaths, especially the change in drowning death.

Another issue is that there was a suggestion of increased mortality due to diabetes.  I am one of the world's experts on diabetes epidemiology, and seriously doubt that the storm could speed the death of people with diabetes. I do not know of any data to support this.

Thus I agree that the number of deaths is greater than 64.  However, I really find it hard to believe that this was 2-3 times as deadly at Katrina

I would love your thoughts

ron

Dear Giants of Statistics,

I would appreciate your statistical input to me, an epidemiologist.

   A few days ago Trump indicated that there were only 64 deaths my Maria in Puerto Rico, despite two studies suggesting 2715 (GW) and 4645 (Harvard).   I became interested in this huge variation.  Quite clearly the number of deaths exceed 64.  However, as I dug deeper in the other studies, it seemed that the GW and Harvard estimates were very high.  I wanted to lay out my finding to see what you think, and what type of studies could be done to get better estimates.  What I discovered is that the estimates of "Death by Hurricane" are extremely variable.  Here are the GW and Harvard estimates. Both have problems

https://publichealth.gwu.edu/sites/default/files/downloads/projects/PRstudy/Acertainment%20of%20the%20Estimated%20Excess%20Mortality%20from%20Hurricane%20Maria%20in%20Puerto%20Rico.pdf

https://www.nejm.org/doi/full/10.1056/NEJMsa1803972

Here is what stimulated my interest

Hurricane          # Deaths

Puerto RIco          4645

Puerto RIco           2715

Katrina                   283-1836

Audrey                    418

Sandy                      230

Harvey                     82

Hugo                        34

florence                     5

I find it very hard to believe that maria was 2-3 times as dangerous a Katrina.  

2.  Previous studies mentioned that 40 to 70% of the deaths from Hurricanes are due to drowning due to the storm surges.  The numbers of deaths for drowning is very low in Puerto  rico, I could not find information find what happened after the hurricane, but it is hear where most of the excess deaths would be seen. Even if there was a 10 fold increase in drowning death, this would not bring the estimates even close to ~3000

3.  the GW study postulates that the excess deaths were not found because of poor death certificate coding.  This is difficult to believe, as the increase in deaths would be located with drowning and injuries, both of which would be picked up. Also, for this was the cause, then 99% of the deaths due to the hurricane would have to be miscoded to account for the excess. The PR health department of doctors would never be this bad.  Finally, if there were 3000-5000  deaths during this period it would be obvious as as people would be "dying in the streets", something we saw with Katrina.  THis really does not make sense

4. A primary reason for this is that it is difficult to classify a hurricane related death.  This is an interesting article indicating that we still do not have a firm figure about deaths .  https://fivethirtyeight.com/features/we-still-dont-know-how-many-people-died-because-of-katrina/?ex_cid=538twitter  Much of the variation can be explained through this.

6.  The approached used in the GW study are suspicious for me as an epidemiologist. They estimate the rates and numbers of total mortality before the hurricane, and then during it.  This type of analysis is call an ecologic analysis, and quite problematic for epidemiologist like me.   THe reason is that this is a pure correlation analysis and not adequate for testing specific hypotheses.    Other things could be changing rapidly as well, e.g. the opiate epidemic, rapidly changing economic conditions.  I will not feel comfortable until I see the changes in the cause specific deaths, especially the change in drowning death.

Another issue is that there was a suggestion of increased mortality due to diabetes.  I am one of the world's experts on diabetes epidemiology, and seriously doubt that the storm could speed the death of people with diabetes. I do not know of any data to support this.

Thus I agree that the number of deaths is greater than 64.  However, I really find it hard to believe that this was 2-3 times as deadly at Katrina

I would love your thoughts

ron

Dear Giants of Statistics,

I would appreciate your statistical input to me, an epidemiologist.

   A few days ago Trump indicated that there were only 64 deaths my Maria in Puerto Rico, despite two studies suggesting 2715 (GW) and 4645 (Harvard).   I became interested in this huge variation.  Quite clearly the number of deaths exceed 64.  However, as I dug deeper in the other studies, it seemed that the GW and Harvard estimates were very high.  I wanted to lay out my finding to see what you think, and what type of studies could be done to get better estimates.  What I discovered is that the estimates of "Death by Hurricane" are extremely variable.  Here are the GW and Harvard estimates. Both have problems

https://publichealth.gwu.edu/sites/default/files/downloads/projects/PRstudy/Acertainment%20of%20the%20Estimated%20Excess%20Mortality%20from%20Hurricane%20Maria%20in%20Puerto%20Rico.pdf

https://www.nejm.org/doi/full/10.1056/NEJMsa1803972

Here is what stimulated my interest

Hurricane          # Deaths

Puerto RIco          4645

Puerto RIco           2715

Katrina                   283-1836

Audrey                    418

Sandy                      230

Harvey                     82

Hugo                        34

florence                     5

I find it very hard to believe that maria was 2-3 times as dangerous a Katrina.  

2.  Previous studies mentioned that 40 to 70% of the deaths from Hurricanes are due to drowning due to the storm surges.  The numbers of deaths for drowning is very low in Puerto  rico, I could not find information find what happened after the hurricane, but it is hear where most of the excess deaths would be seen. Even if there was a 10 fold increase in drowning death, this would not bring the estimates even close to ~3000

3.  the GW study postulates that the excess deaths were not found because of poor death certificate coding.  This is difficult to believe, as the increase in deaths would be located with drowning and injuries, both of which would be picked up. Also, for this was the cause, then 99% of the deaths due to the hurricane would have to be miscoded to account for the excess. The PR health department of doctors would never be this bad.  Finally, if there were 3000-5000  deaths during this period it would be obvious as as people would be "dying in the streets", something we saw with Katrina.  THis really does not make sense

4. A primary reason for this is that it is difficult to classify a hurricane related death.  This is an interesting article indicating that we still do not have a firm figure about deaths .  https://fivethirtyeight.com/features/we-still-dont-know-how-many-people-died-because-of-katrina/?ex_cid=538twitter  Much of the variation can be explained through this.

6.  The approached used in the GW study are suspicious for me as an epidemiologist. They estimate the rates and numbers of total mortality before the hurricane, and then during it.  This type of analysis is call an ecologic analysis, and quite problematic for epidemiologist like me.   THe reason is that this is a pure correlation analysis and not adequate for testing specific hypotheses.    Other things could be changing rapidly as well, e.g. the opiate epidemic, rapidly changing economic conditions.  I will not feel comfortable until I see the changes in the cause specific deaths, especially the change in drowning death.

Another issue is that there was a suggestion of increased mortality due to diabetes.  I am one of the world's experts on diabetes epidemiology, and seriously doubt that the storm could speed the death of people with diabetes. I do not know of any data to support this.

Thus I agree that the number of deaths is greater than 64.  However, I really find it hard to believe that this was 2-3 times as deadly at Katrina

I would love your thoughts

ron

Dear Giants of Statistics,

I would appreciate your statistical input to me, an epidemiologist.

   A few days ago Trump indicated that there were only 64 deaths my Maria in Puerto Rico, despite two studies suggesting 2715 (GW) and 4645 (Harvard).   I became interested in this huge variation.  Quite clearly the number of deaths exceed 64.  However, as I dug deeper in the other studies, it seemed that the GW and Harvard estimates were very high.  I wanted to lay out my finding to see what you think, and what type of studies could be done to get better estimates.  What I discovered is that the estimates of "Death by Hurricane" are extremely variable.  Here are the GW and Harvard estimates. Both have problems

https://publichealth.gwu.edu/sites/default/files/downloads/projects/PRstudy/Acertainment%20of%20the%20Estimated%20Excess%20Mortality%20from%20Hurricane%20Maria%20in%20Puerto%20Rico.pdf

https://www.nejm.org/doi/full/10.1056/NEJMsa1803972

Here is what stimulated my interest

Hurricane          # Deaths

Puerto RIco          4645

Puerto RIco           2715

Katrina                   283-1836

Audrey                    418

Sandy                      230

Harvey                     82

Hugo                        34

florence                     5

I find it very hard to believe that maria was 2-3 times as dangerous a Katrina.  

2.  Previous studies mentioned that 40 to 70% of the deaths from Hurricanes are due to drowning due to the storm surges.  The numbers of deaths for drowning is very low in Puerto  rico, I could not find information find what happened after the hurricane, but it is hear where most of the excess deaths would be seen. Even if there was a 10 fold increase in drowning death, this would not bring the estimates even close to ~3000

3.  the GW study postulates that the excess deaths were not found because of poor death certificate coding.  This is difficult to believe, as the increase in deaths would be located with drowning and injuries, both of which would be picked up. Also, for this was the cause, then 99% of the deaths due to the hurricane would have to be miscoded to account for the excess. The PR health department of doctors would never be this bad.  Finally, if there were 3000-5000  deaths during this period it would be obvious as as people would be "dying in the streets", something we saw with Katrina.  THis really does not make sense

4. A primary reason for this is that it is difficult to classify a hurricane related death.  This is an interesting article indicating that we still do not have a firm figure about deaths .  https://fivethirtyeight.com/features/we-still-dont-know-how-many-people-died-because-of-katrina/?ex_cid=538twitter  Much of the variation can be explained through this.

6.  The approached used in the GW study are suspicious for me as an epidemiologist. They estimate the rates and numbers of total mortality before the hurricane, and then during it.  This type of analysis is call an ecologic analysis, and quite problematic for epidemiologist like me.   THe reason is that this is a pure correlation analysis and not adequate for testing specific hypotheses.    Other things could be changing rapidly as well, e.g. the opiate epidemic, rapidly changing economic conditions.  I will not feel comfortable until I see the changes in the cause specific deaths, especially the change in drowning death.

Another issue is that there was a suggestion of increased mortality due to diabetes.  I am one of the world's experts on diabetes epidemiology, and seriously doubt that the storm could speed the death of people with diabetes. I do not know of any data to support this.

Thus I agree that the number of deaths is greater than 64.  However, I really find it hard to believe that this was 2-3 times as deadly at Katrina

I would love your thoughts

ron

Dear Giants of Statistics,

I would appreciate your statistical input to me, an epidemiologist.

   A few days ago Trump indicated that there were only 64 deaths my Maria in Puerto Rico, despite two studies suggesting 2715 (GW) and 4645 (Harvard).   I became interested in this huge variation.  Quite clearly the number of deaths exceed 64.  However, as I dug deeper in the other studies, it seemed that the GW and Harvard estimates were very high.  I wanted to lay out my finding to see what you think, and what type of studies could be done to get better estimates.  What I discovered is that the estimates of "Death by Hurricane" are extremely variable.  Here are the GW and Harvard estimates. Both have problems

https://publichealth.gwu.edu/sites/default/files/downloads/projects/PRstudy/Acertainment%20of%20the%20Estimated%20Excess%20Mortality%20from%20Hurricane%20Maria%20in%20Puerto%20Rico.pdf

https://www.nejm.org/doi/full/10.1056/NEJMsa1803972

Here is what stimulated my interest

Hurricane          # Deaths

Puerto RIco          4645

Puerto RIco           2715

Katrina                   283-1836

Audrey                    418

Sandy                      230

Harvey                     82

Hugo                        34

florence                     5

I find it very hard to believe that maria was 2-3 times as dangerous a Katrina.  

2.  Previous studies mentioned that 40 to 70% of the deaths from Hurricanes are due to drowning due to the storm surges.  The numbers of deaths for drowning is very low in Puerto  rico, I could not find information find what happened after the hurricane, but it is hear where most of the excess deaths would be seen. Even if there was a 10 fold increase in drowning death, this would not bring the estimates even close to ~3000

3.  the GW study postulates that the excess deaths were not found because of poor death certificate coding.  This is difficult to believe, as the increase in deaths would be located with drowning and injuries, both of which would be picked up. Also, for this was the cause, then 99% of the deaths due to the hurricane would have to be miscoded to account for the excess. The PR health department of doctors would never be this bad.  Finally, if there were 3000-5000  deaths during this period it would be obvious as as people would be "dying in the streets", something we saw with Katrina.  THis really does not make sense

4. A primary reason for this is that it is difficult to classify a hurricane related death.  This is an interesting article indicating that we still do not have a firm figure about deaths .  https://fivethirtyeight.com/features/we-still-dont-know-how-many-people-died-because-of-katrina/?ex_cid=538twitter  Much of the variation can be explained through this.

6.  The approached used in the GW study are suspicious for me as an epidemiologist. They estimate the rates and numbers of total mortality before the hurricane, and then during it.  This type of analysis is call an ecologic analysis, and quite problematic for epidemiologist like me.   THe reason is that this is a pure correlation analysis and not adequate for testing specific hypotheses.    Other things could be changing rapidly as well, e.g. the opiate epidemic, rapidly changing economic conditions.  I will not feel comfortable until I see the changes in the cause specific deaths, especially the change in drowning death.

Another issue is that there was a suggestion of increased mortality due to diabetes.  I am one of the world's experts on diabetes epidemiology, and seriously doubt that the storm could speed the death of people with diabetes. I do not know of any data to support this.

Thus I agree that the number of deaths is greater than 64.  However, I really find it hard to believe that this was 2-3 times as deadly at Katrina

I would love your thoughts

ron

Dear Giants of Statistics,

I would appreciate your statistical input to me, an epidemiologist.

   A few days ago Trump indicated that there were only 64 deaths my Maria in Puerto Rico, despite two studies suggesting 2715 (GW) and 4645 (Harvard).   I became interested in this huge variation.  Quite clearly the number of deaths exceed 64.  However, as I dug deeper in the other studies, it seemed that the GW and Harvard estimates were very high.  I wanted to lay out my finding to see what you think, and what type of studies could be done to get better estimates.  What I discovered is that the estimates of "Death by Hurricane" are extremely variable.  Here are the GW and Harvard estimates. Both have problems

https://publichealth.gwu.edu/sites/default/files/downloads/projects/PRstudy/Acertainment%20of%20the%20Estimated%20Excess%20Mortality%20from%20Hurricane%20Maria%20in%20Puerto%20Rico.pdf

https://www.nejm.org/doi/full/10.1056/NEJMsa1803972

Here is what stimulated my interest

Hurricane          # Deaths

Puerto RIco          4645

Puerto RIco           2715

Katrina                   283-1836

Audrey                    418

Sandy                      230

Harvey                     82

Hugo                        34

florence                     5

I find it very hard to believe that maria was 2-3 times as dangerous a Katrina.  

2.  Previous studies mentioned that 40 to 70% of the deaths from Hurricanes are due to drowning due to the storm surges.  The numbers of deaths for drowning is very low in Puerto  rico, I could not find information find what happened after the hurricane, but it is hear where most of the excess deaths would be seen. Even if there was a 10 fold increase in drowning death, this would not bring the estimates even close to ~3000

3.  the GW study postulates that the excess deaths were not found because of poor death certificate coding.  This is difficult to believe, as the increase in deaths would be located with drowning and injuries, both of which would be picked up. Also, for this was the cause, then 99% of the deaths due to the hurricane would have to be miscoded to account for the excess. The PR health department of doctors would never be this bad.  Finally, if there were 3000-5000  deaths during this period it would be obvious as as people would be "dying in the streets", something we saw with Katrina.  THis really does not make sense

4. A primary reason for this is that it is difficult to classify a hurricane related death.  This is an interesting article indicating that we still do not have a firm figure about deaths .  https://fivethirtyeight.com/features/we-still-dont-know-how-many-people-died-because-of-katrina/?ex_cid=538twitter  Much of the variation can be explained through this.

6.  The approached used in the GW study are suspicious for me as an epidemiologist. They estimate the rates and numbers of total mortality before the hurricane, and then during it.  This type of analysis is call an ecologic analysis, and quite problematic for epidemiologist like me.   THe reason is that this is a pure correlation analysis and not adequate for testing specific hypotheses.    Other things could be changing rapidly as well, e.g. the opiate epidemic, rapidly changing economic conditions.  I will not feel comfortable until I see the changes in the cause specific deaths, especially the change in drowning death.

Another issue is that there was a suggestion of increased mortality due to diabetes.  I am one of the world's experts on diabetes epidemiology, and seriously doubt that the storm could speed the death of people with diabetes. I do not know of any data to support this.

Thus I agree that the number of deaths is greater than 64.  However, I really find it hard to believe that this was 2-3 times as deadly at Katrina

I would love your thoughts

ron

Dear Giants of Statistics,

I would appreciate your statistical input to me, an epidemiologist.

   A few days ago Trump indicated that there were only 64 deaths my Maria in Puerto Rico, despite two studies suggesting 2715 (GW) and 4645 (Harvard).   I became interested in this huge variation.  Quite clearly the number of deaths exceed 64.  However, as I dug deeper in the other studies, it seemed that the GW and Harvard estimates were very high.  I wanted to lay out my finding to see what you think, and what type of studies could be done to get better estimates.  What I discovered is that the estimates of "Death by Hurricane" are extremely variable.  Here are the GW and Harvard estimates. Both have problems

https://publichealth.gwu.edu/sites/default/files/downloads/projects/PRstudy/Acertainment%20of%20the%20Estimated%20Excess%20Mortality%20from%20Hurricane%20Maria%20in%20Puerto%20Rico.pdf

https://www.nejm.org/doi/full/10.1056/NEJMsa1803972

Here is what stimulated my interest

Hurricane          # Deaths

Puerto RIco          4645

Puerto RIco           2715

Katrina                   283-1836

Audrey                    418

Sandy                      230

Harvey                     82

Hugo                        34

florence                     5

I find it very hard to believe that maria was 2-3 times as dangerous a Katrina.  

2.  Previous studies mentioned that 40 to 70% of the deaths from Hurricanes are due to drowning due to the storm surges.  The numbers of deaths for drowning is very low in Puerto  rico, I could not find information find what happened after the hurricane, but it is hear where most of the excess deaths would be seen. Even if there was a 10 fold increase in drowning death, this would not bring the estimates even close to ~3000

3.  the GW study postulates that the excess deaths were not found because of poor death certificate coding.  This is difficult to believe, as the increase in deaths would be located with drowning and injuries, both of which would be picked up. Also, for this was the cause, then 99% of the deaths due to the hurricane would have to be miscoded to account for the excess. The PR health department of doctors would never be this bad.  Finally, if there were 3000-5000  deaths during this period it would be obvious as as people would be "dying in the streets", something we saw with Katrina.  THis really does not make sense

4. A primary reason for this is that it is difficult to classify a hurricane related death.  This is an interesting article indicating that we still do not have a firm figure about deaths .  https://fivethirtyeight.com/features/we-still-dont-know-how-many-people-died-because-of-katrina/?ex_cid=538twitter  Much of the variation can be explained through this.

6.  The approached used in the GW study are suspicious for me as an epidemiologist. They estimate the rates and numbers of total mortality before the hurricane, and then during it.  This type of analysis is call an ecologic analysis, and quite problematic for epidemiologist like me.   THe reason is that this is a pure correlation analysis and not adequate for testing specific hypotheses.    Other things could be changing rapidly as well, e.g. the opiate epidemic, rapidly changing economic conditions.  I will not feel comfortable until I see the changes in the cause specific deaths, especially the change in drowning death.

Another issue is that there was a suggestion of increased mortality due to diabetes.  I am one of the world's experts on diabetes epidemiology, and seriously doubt that the storm could speed the death of people with diabetes. I do not know of any data to support this.

Thus I agree that the number of deaths is greater than 64.  However, I really find it hard to believe that this was 2-3 times as deadly at Katrina

I would love your thoughts

ron

Dear Giants of Statistics,

I would appreciate your statistical input to me, an epidemiologist.

   A few days ago Trump indicated that there were only 64 deaths my Maria in Puerto Rico, despite two studies suggesting 2715 (GW) and 4645 (Harvard).   I became interested in this huge variation.  Quite clearly the number of deaths exceed 64.  However, as I dug deeper in the other studies, it seemed that the GW and Harvard estimates were very high.  I wanted to lay out my finding to see what you think, and what type of studies could be done to get better estimates.  What I discovered is that the estimates of "Death by Hurricane" are extremely variable.  Here are the GW and Harvard estimates. Both have problems

https://publichealth.gwu.edu/sites/default/files/downloads/projects/PRstudy/Acertainment%20of%20the%20Estimated%20Excess%20Mortality%20from%20Hurricane%20Maria%20in%20Puerto%20Rico.pdf

https://www.nejm.org/doi/full/10.1056/NEJMsa1803972

Here is what stimulated my interest

Hurricane          # Deaths

Puerto RIco          4645

Puerto RIco           2715

Katrina                   283-1836

Audrey                    418

Sandy                      230

Harvey                     82

Hugo                        34

florence                     5

I find it very hard to believe that maria was 2-3 times as dangerous a Katrina.  

2.  Previous studies mentioned that 40 to 70% of the deaths from Hurricanes are due to drowning due to the storm surges.  The numbers of deaths for drowning is very low in Puerto  rico, I could not find information find what happened after the hurricane, but it is hear where most of the excess deaths would be seen. Even if there was a 10 fold increase in drowning death, this would not bring the estimates even close to ~3000

3.  the GW study postulates that the excess deaths were not found because of poor death certificate coding.  This is difficult to believe, as the increase in deaths would be located with drowning and injuries, both of which would be picked up. Also, for this was the cause, then 99% of the deaths due to the hurricane would have to be miscoded to account for the excess. The PR health department of doctors would never be this bad.  Finally, if there were 3000-5000  deaths during this period it would be obvious as as people would be "dying in the streets", something we saw with Katrina.  THis really does not make sense

4. A primary reason for this is that it is difficult to classify a hurricane related death.  This is an interesting article indicating that we still do not have a firm figure about deaths .  https://fivethirtyeight.com/features/we-still-dont-know-how-many-people-died-because-of-katrina/?ex_cid=538twitter  Much of the variation can be explained through this.

6.  The approached used in the GW study are suspicious for me as an epidemiologist. They estimate the rates and numbers of total mortality before the hurricane, and then during it.  This type of analysis is call an ecologic analysis, and quite problematic for epidemiologist like me.   THe reason is that this is a pure correlation analysis and not adequate for testing specific hypotheses.    Other things could be changing rapidly as well, e.g. the opiate epidemic, rapidly changing economic conditions.  I will not feel comfortable until I see the changes in the cause specific deaths, especially the change in drowning death.

Another issue is that there was a suggestion of increased mortality due to diabetes.  I am one of the world's experts on diabetes epidemiology, and seriously doubt that the storm could speed the death of people with diabetes. I do not know of any data to support this.

Thus I agree that the number of deaths is greater than 64.  However, I really find it hard to believe that this was 2-3 times as deadly at Katrina

I would love your thoughts

ron

Dear Giants of Statistics,

I would appreciate your statistical input to me, an epidemiologist.

   A few days ago Trump indicated that there were only 64 deaths my Maria in Puerto Rico, despite two studies suggesting 2715 (GW) and 4645 (Harvard).   I became interested in this huge variation.  Quite clearly the number of deaths exceed 64.  However, as I dug deeper in the other studies, it seemed that the GW and Harvard estimates were very high.  I wanted to lay out my finding to see what you think, and what type of studies could be done to get better estimates.  What I discovered is that the estimates of "Death by Hurricane" are extremely variable.  Here are the GW and Harvard estimates. Both have problems

https://publichealth.gwu.edu/sites/default/files/downloads/projects/PRstudy/Acertainment%20of%20the%20Estimated%20Excess%20Mortality%20from%20Hurricane%20Maria%20in%20Puerto%20Rico.pdf

https://www.nejm.org/doi/full/10.1056/NEJMsa1803972

Here is what stimulated my interest

Hurricane          # Deaths

Puerto RIco          4645

Puerto RIco           2715

Katrina                   283-1836

Audrey                    418

Sandy                      230

Harvey                     82

Hugo                        34

florence                     5

I find it very hard to believe that maria was 2-3 times as dangerous a Katrina.  

2.  Previous studies mentioned that 40 to 70% of the deaths from Hurricanes are due to drowning due to the storm surges.  The numbers of deaths for drowning is very low in Puerto  rico, I could not find information find what happened after the hurricane, but it is hear where most of the excess deaths would be seen. Even if there was a 10 fold increase in drowning death, this would not bring the estimates even close to ~3000

3.  the GW study postulates that the excess deaths were not found because of poor death certificate coding.  This is difficult to believe, as the increase in deaths would be located with drowning and injuries, both of which would be picked up. Also, for this was the cause, then 99% of the deaths due to the hurricane would have to be miscoded to account for the excess. The PR health department of doctors would never be this bad.  Finally, if there were 3000-5000  deaths during this period it would be obvious as as people would be "dying in the streets", something we saw with Katrina.  THis really does not make sense

4. A primary reason for this is that it is difficult to classify a hurricane related death.  This is an interesting article indicating that we still do not have a firm figure about deaths .  https://fivethirtyeight.com/features/we-still-dont-know-how-many-people-died-because-of-katrina/?ex_cid=538twitter  Much of the variation can be explained through this.

6.  The approached used in the GW study are suspicious for me as an epidemiologist. They estimate the rates and numbers of total mortality before the hurricane, and then during it.  This type of analysis is call an ecologic analysis, and quite problematic for epidemiologist like me.   THe reason is that this is a pure correlation analysis and not adequate for testing specific hypotheses.    Other things could be changing rapidly as well, e.g. the opiate epidemic, rapidly changing economic conditions.  I will not feel comfortable until I see the changes in the cause specific deaths, especially the change in drowning death.

Another issue is that there was a suggestion of increased mortality due to diabetes.  I am one of the world's experts on diabetes epidemiology, and seriously doubt that the storm could speed the death of people with diabetes. I do not know of any data to support this.

Thus I agree that the number of deaths is greater than 64.  However, I really find it hard to believe that this was 2-3 times as deadly at Katrina

I would love your thoughts

ron

Dear Giants of Statistics,

I would appreciate your statistical input to me, an epidemiologist.

   A few days ago Trump indicated that there were only 64 deaths my Maria in Puerto Rico, despite two studies suggesting 2715 (GW) and 4645 (Harvard).   I became interested in this huge variation.  Quite clearly the number of deaths exceed 64.  However, as I dug deeper in the other studies, it seemed that the GW and Harvard estimates were very high.  I wanted to lay out my finding to see what you think, and what type of studies could be done to get better estimates.  What I discovered is that the estimates of "Death by Hurricane" are extremely variable.  Here are the GW and Harvard estimates. Both have problems

https://publichealth.gwu.edu/sites/default/files/downloads/projects/PRstudy/Acertainment%20of%20the%20Estimated%20Excess%20Mortality%20from%20Hurricane%20Maria%20in%20Puerto%20Rico.pdf

https://www.nejm.org/doi/full/10.1056/NEJMsa1803972

Here is what stimulated my interest

Hurricane          # Deaths

Puerto RIco          4645

Puerto RIco           2715

Katrina                   283-1836

Audrey                    418

Sandy                      230

Harvey                     82

Hugo                        34

florence                     5

I find it very hard to believe that maria was 2-3 times as dangerous a Katrina.  

2.  Previous studies mentioned that 40 to 70% of the deaths from Hurricanes are due to drowning due to the storm surges.  The numbers of deaths for drowning is very low in Puerto  rico, I could not find information find what happened after the hurricane, but it is hear where most of the excess deaths would be seen. Even if there was a 10 fold increase in drowning death, this would not bring the estimates even close to ~3000

3.  the GW study postulates that the excess deaths were not found because of poor death certificate coding.  This is difficult to believe, as the increase in deaths would be located with drowning and injuries, both of which would be picked up. Also, for this was the cause, then 99% of the deaths due to the hurricane would have to be miscoded to account for the excess. The PR health department of doctors would never be this bad.  Finally, if there were 3000-5000  deaths during this period it would be obvious as as people would be "dying in the streets", something we saw with Katrina.  THis really does not make sense

4. A primary reason for this is that it is difficult to classify a hurricane related death.  This is an interesting article indicating that we still do not have a firm figure about deaths .  https://fivethirtyeight.com/features/we-still-dont-know-how-many-people-died-because-of-katrina/?ex_cid=538twitter  Much of the variation can be explained through this.

6.  The approached used in the GW study are suspicious for me as an epidemiologist. They estimate the rates and numbers of total mortality before the hurricane, and then during it.  This type of analysis is call an ecologic analysis, and quite problematic for epidemiologist like me.   THe reason is that this is a pure correlation analysis and not adequate for testing specific hypotheses.    Other things could be changing rapidly as well, e.g. the opiate epidemic, rapidly changing economic conditions.  I will not feel comfortable until I see the changes in the cause specific deaths, especially the change in drowning death.

Another issue is that there was a suggestion of increased mortality due to diabetes.  I am one of the world's experts on diabetes epidemiology, and seriously doubt that the storm could speed the death of people with diabetes. I do not know of any data to support this.

Thus I agree that the number of deaths is greater than 64.  However, I really find it hard to believe that this was 2-3 times as deadly at Katrina

I would love your thoughts

ron

Dear Giants of Statistics,

I would appreciate your statistical input to me, an epidemiologist.

   A few days ago Trump indicated that there were only 64 deaths my Maria in Puerto Rico, despite two studies suggesting 2715 (GW) and 4645 (Harvard).   I became interested in this huge variation.  Quite clearly the number of deaths exceed 64.  However, as I dug deeper in the other studies, it seemed that the GW and Harvard estimates were very high.  I wanted to lay out my finding to see what you think, and what type of studies could be done to get better estimates.  What I discovered is that the estimates of "Death by Hurricane" are extremely variable.  Here are the GW and Harvard estimates. Both have problems

https://publichealth.gwu.edu/sites/default/files/downloads/projects/PRstudy/Acertainment%20of%20the%20Estimated%20Excess%20Mortality%20from%20Hurricane%20Maria%20in%20Puerto%20Rico.pdf

https://www.nejm.org/doi/full/10.1056/NEJMsa1803972

Here is what stimulated my interest

Hurricane          # Deaths

Puerto RIco          4645

Puerto RIco           2715

Katrina                   283-1836

Audrey                    418

Sandy                      230

Harvey                     82

Hugo                        34

florence                     5

I find it very hard to believe that maria was 2-3 times as dangerous a Katrina.  

2.  Previous studies mentioned that 40 to 70% of the deaths from Hurricanes are due to drowning due to the storm surges.  The numbers of deaths for drowning is very low in Puerto  rico, I could not find information find what happened after the hurricane, but it is hear where most of the excess deaths would be seen. Even if there was a 10 fold increase in drowning death, this would not bring the estimates even close to ~3000

3.  the GW study postulates that the excess deaths were not found because of poor death certificate coding.  This is difficult to believe, as the increase in deaths would be located with drowning and injuries, both of which would be picked up. Also, for this was the cause, then 99% of the deaths due to the hurricane would have to be miscoded to account for the excess. The PR health department of doctors would never be this bad.  Finally, if there were 3000-5000  deaths during this period it would be obvious as as people would be "dying in the streets", something we saw with Katrina.  THis really does not make sense

4. A primary reason for this is that it is difficult to classify a hurricane related death.  This is an interesting article indicating that we still do not have a firm figure about deaths .  https://fivethirtyeight.com/features/we-still-dont-know-how-many-people-died-because-of-katrina/?ex_cid=538twitter  Much of the variation can be explained through this.

6.  The approached used in the GW study are suspicious for me as an epidemiologist. They estimate the rates and numbers of total mortality before the hurricane, and then during it.  This type of analysis is call an ecologic analysis, and quite problematic for epidemiologist like me.   THe reason is that this is a pure correlation analysis and not adequate for testing specific hypotheses.    Other things could be changing rapidly as well, e.g. the opiate epidemic, rapidly changing economic conditions.  I will not feel comfortable until I see the changes in the cause specific deaths, especially the change in drowning death.

Another issue is that there was a suggestion of increased mortality due to diabetes.  I am one of the world's experts on diabetes epidemiology, and seriously doubt that the storm could speed the death of people with diabetes. I do not know of any data to support this.

Thus I agree that the number of deaths is greater than 64.  However, I really find it hard to believe that this was 2-3 times as deadly at Katrina

I would love your thoughts

ron