37
Superstorm Sandy Mortality Surveillance in New York City Elizabeth “Beth” Begier, MD, MPH Assistant Commissioner, Bureau of Vital Statistics NYC Department of Health and Mental Hygiene

Superstorm Sandy Mortality Surveillance in New York City

  • Upload
    mervyn

  • View
    50

  • Download
    2

Embed Size (px)

DESCRIPTION

Superstorm Sandy Mortality Surveillance in New York City. Elizabeth “Beth” Begier, MD, MPH Assistant Commissioner, Bureau of Vital Statistics NYC Department of Health and Mental Hygiene. Overview. Sandy Impact and Public Health Threats NYC Death Registration System - PowerPoint PPT Presentation

Citation preview

Page 1: Superstorm Sandy  Mortality Surveillance in New York City

Superstorm Sandy Mortality Surveillance

in New York City

Elizabeth “Beth” Begier, MD, MPHAssistant Commissioner, Bureau of Vital StatisticsNYC Department of Health and Mental Hygiene

Page 2: Superstorm Sandy  Mortality Surveillance in New York City

Overview

• Sandy Impact and Public Health Threats• NYC Death Registration System• Mortality Surveillance Activities/Findings• Challenges and Lessons Learned• Next Steps

Page 3: Superstorm Sandy  Mortality Surveillance in New York City

3

Hurricane Sandy

• October 29, 2012 post-tropical cyclone hit 100 miles south of NYC

• Record storm surge flooding (4 to 11 ft. predicted vs. 14 actual)

• Wind-downed trees and power lines caused fatal drowning and injuries (direct mortality)

Page 4: Superstorm Sandy  Mortality Surveillance in New York City

NYC Sandy Impacts and Hazards• Widespread power and heat outages resulting cold

stress and lack of refrigeration (some prolonged building-related issues)

• High rise buildings lost elevators and running water– Mobility-impaired stranded

• Transit disruption– Total subway/transit shutdown followed by slow

phased resumption– Bridge and tunnel closures– Gasoline shortages

4

Page 5: Superstorm Sandy  Mortality Surveillance in New York City

5

NYC Sandy Impacts and Hazards• Evacuation of 4 hospitals (1,262 pts.) and 17 nursing

homes (2507 residents)• Disruption of ambulatory health care and pharmacy

access• Respiratory hazards and concerns– Indoor demolition and flooded building cleanup– Outdoor dust from debris movement– Emissions from temporary generators and boilers

• Mental health stress and loss of 300 psychiatric inpatient beds due to evacuations

• Injuries relating to rebuilding efforts

Page 6: Superstorm Sandy  Mortality Surveillance in New York City

Routine Death Registration in NYC• NYC Health Code Death Reporting Requirements:– 24 or less hours for certification• Clinician's report of decedent's name, sex,

time/date/place of death, and free-text cause of death• Medical providers start “case” in EDRS

– 72 hours or less for FD completes and sends for registration • Decedent's address, DOB, and other demographics

completed by funeral director– After FD completes, “case” goes electronically to death

registration staff for review and registration• 94% of deaths reported fully electronically reported via EDRS

Page 7: Superstorm Sandy  Mortality Surveillance in New York City

Death Registration during/after Sandy • 1 registration staff member stayed at Manhattan

office through storm (Steve S relieved him morning after)• Bureau of Vital Statistics’ Office building in lower

Manhattan closed for ~1 week due to power and heat outage that began overnight during storm

• Relocated Registration Unit morning after hurricane to alternate site in Brooklyn (Thanks Flor Betancourt!)

• EDRS up during/after storm; no prolonged outages• A few hospitals initially with internet connectivity

issues preventing use of EDRS

Page 8: Superstorm Sandy  Mortality Surveillance in New York City

Tracking: Deaths directly related to Hurricane

• Most of such deaths initially registered by MEs with cause pending

• ME notified us via email of deaths believed to be Hurricane-related after preliminary investigation

• All CODs reviewed in real-time by registration unit for possible ME cases (per routine)

• Added pop-up to EDRS cause-of-death page to prompt ME referrals for any deaths related to storm

• Conducted free-text searches of DC text fields post-registration for injury deaths as double check

Page 9: Superstorm Sandy  Mortality Surveillance in New York City

Tracking: Indirectly Related Deaths • Health Commissioner very concerned about

unrecognized morbidity and mortality increases associated with extensive environmental hazards and stresses in aftermath – Syndromic surveillance for morbidity– Initiated mortality surveillance

• Timeframes– Categorized 10 days following hurricane as

immediate aftermath (1/1–10)– Rest of time examined as rolling post-aftermath

Page 10: Superstorm Sandy  Mortality Surveillance in New York City

Mortality Surveillance – ApproachAll-cause mortality counts by date to identify excess mortality of any cause: primary focus post-event– Used pre-registration certified deaths to enhance

timeliness of close to complete counts• Concerned about decreased registration

timeliness 2° to storm-related disruptions – Adjusted prior year comparison counts for

“reporting lag”, i.e., limited to deaths certified by same date in prior years• Had to add certification date to stats file (only

had registration date)

Page 11: Superstorm Sandy  Mortality Surveillance in New York City

Mortality Surveillance – Approach• Categorized deaths by age and cause with post-

registration data– Cause-of-death routinely ICD-10 coded locally by

NYC nosologists 1st business day post-registration• Seemed too complicated to change routine to

pre-registration on fly– Date of birth not on certification – provided by FD

Page 12: Superstorm Sandy  Mortality Surveillance in New York City

Mortality Surveillance:Neighborhood Flood Levels

• Categorized deaths by decedent residence neighborhood flood levels with post-registration data– Quickly geocoded residence address to census tract

level, including quickly resolving manual rejects• Quickly recoded prior years to 2010 census for

accurate comparisons– Do not get decedent address until registration (from

funeral directors)– Excluded nursing home deaths to focus on deaths

occurring in community• Separate analysis done for evacuated nursing

homes

Page 13: Superstorm Sandy  Mortality Surveillance in New York City

Mortality – Other investigations• Initial assessment of Hospital evacuations– Calculated mortality counts at receiving hospitals in

10 days following storm– Added expected death counts for evacuated hospitals

to receiving hospitals’ baselines based on count of evacuees received to improve comparability

– Did not have names or acuity of transferees (ICU, etc)• Initial look at nursing home evacuations– Identified decadents with evacuated nursing homes

as their residence address– Compared counts for 4 weeks post-sandy to 2 prior

years

Page 14: Superstorm Sandy  Mortality Surveillance in New York City

RESULTS

Page 15: Superstorm Sandy  Mortality Surveillance in New York City

Results: Timeliness

• Time to certification & registration increased during/after storm compared to current and prior year baselines

Page 16: Superstorm Sandy  Mortality Surveillance in New York City

Timeliness by Date– Certification:Deaths Certified in <24 hours in Blue

10/15/2012

10/17/2012

10/19/2012

10/21/2012

10/23/2012

10/25/2012

10/27/2012

10/29/2012

10/31/2012

11/2/2012

11/4/2012

11/6/2012

11/8/2012

11/10/2012

11/12/2012

11/14/2012

11/16/2012

11/18/2012

11/20/2012

11/22/2012

11/24/2012

11/26/2012

11/28/2012

11/30/20120

10

20

30

40

50

60

70

80

90

100

> 5 days4-5 days3-4 days2-3 days1-2 days<1 days

Perc

ent o

f Reg

istra

tions

Page 17: Superstorm Sandy  Mortality Surveillance in New York City

Timeliness by Date– Registration:Day 1 Blue, Day 2 Red, Day 3 Green

10/15/2012

10/17/2012

10/19/2012

10/21/2012

10/23/2012

10/25/2012

10/27/2012

10/29/2012

10/31/2012

11/2/2012

11/4/2012

11/6/2012

11/8/2012

11/10/2012

11/12/2012

11/14/2012

11/16/2012

11/18/2012

11/20/2012

11/22/2012

11/24/2012

11/26/2012

11/28/2012

11/30/2012

0

10

20

30

40

50

60

70

80

90

100

> 5 days4-5 days3-4 days2-3 days1-2 days<1 days

Perc

ent o

f Reg

istra

tions

Page 18: Superstorm Sandy  Mortality Surveillance in New York City

Median Days to Registration from Death by Percent of Census Tract Flooded and Year: Oct 29–Nov 10

0% 0-10% 10-75% >75%0

0.5

1

1.5

2

2.5

3

3.5

4

201020112012

Page 19: Superstorm Sandy  Mortality Surveillance in New York City

Mortality Surveillance - Findings

• Relied on ME email notification to BVS for timely notification of preliminary Sandy-related deaths – N=43; Major causes drowning (81%), blunt

trauma (16%)– >50% on Staten Island, nearly half aged 65+– No Sandy-related deaths found through free-text

search of post-registration data for injury terms that had not already sent by ME

Page 20: Superstorm Sandy  Mortality Surveillance in New York City

Sample Early Table: Number of Certified Daily Deaths from All Causes, Nov 1–19, 2010-2012, NYC, Reported As of 21NOV2012

(Deaths Certified Through 11/20 of Each Year)

Date of Death 2010 2011 2012

Absolute difference between 2012 and

average of 2010-2011

Percent difference between 2012 and

average of 2010-201111012012 146 151 137 -11.5 -8%11022012 127 130 146 17.5 14%11032012 134 135 157 22.5 17%11042012 153 149 162 11 7%11052012 129 122 161 35.5 28%11062012 133 135 163 29 22%11072012 155 156 161 5.5 4%11082012 168 159 164 0.5 0%11092012 129 150 173 33.5 24%11102012 152 149 185 34.5 23%11112012 140 158 155 6 4%11122012 152 143 154 6.5 4%11132012 136 145 163 22.5 16%11142012 141 162 145 -6.5 -4%11152012 123 140 145 13.5 10%11162012 155 147 154 3 2%11172012 143 122 136 3.5 3%11182012 127 127 119 -8 -6%11192012 101 111 111 5 5%Mean 139.2 141.6 152.2 11.8 8%

Median 140.0 145.0 155.0 12.5 9%Total 2,644 2,691 2,891 223.5 8%

Please note additional deaths are expected to be registered ongoing on all these dates.

Page 21: Superstorm Sandy  Mortality Surveillance in New York City

Sample Early Table: T Number of Certified Daily Deaths from All Causes, Nov 1–19, 2010-2012, NYC, Reported As of 21NOV2012

(Deaths Certified Through 11/20 of Each Year)

Date of Death 2010 2011 2012

Absolute difference between 2012 and

average of 2010-2011

Percent difference between 2012 and

average of 2010-201111012012 146 151 137 -11.5 -8%11022012 127 130 146 17.5 14%11032012 134 135 157 22.5 17%11042012 153 149 162 11 7%11052012 129 122 161 35.5 28%11062012 133 135 163 29 22%11072012 155 156 161 5.5 4%11082012 168 159 164 0.5 0%11092012 129 150 173 33.5 24%11102012 152 149 185 34.5 23%11112012 140 158 155 6 4%11122012 152 143 154 6.5 4%11132012 136 145 163 22.5 16%11142012 141 162 145 -6.5 -4%11152012 123 140 145 13.5 10%11162012 155 147 154 3 2%11172012 143 122 136 3.5 3%11182012 127 127 119 -8 -6%11192012 101 111 111 5 5%Mean 139.2 141.6 152.2 11.8 8%

Median 140.0 145.0 155.0 12.5 9%Total 2,644 2,691 2,891 223.5 8%

Please note additional deaths are expected to be registered ongoing on all these dates.

• 11/19/2012 on 11/21 report: • Deaths certified by 11/20: 111• Adjusted baseline 101 and 111 deaths indicated not beyond historical norms• That day would eventually get to 138 deaths

Page 22: Superstorm Sandy  Mortality Surveillance in New York City

100

110

120

130

140

150

160

170

180

190

200

New York City All-Cause Daily Deaths Counts, October 20–December 31, 2012

Date of Death

Num

ber o

f Dea

ths

October November December

Hurricane Sandy10 day aftermathDeaths up 11%

Influenza Season Starts

Page 23: Superstorm Sandy  Mortality Surveillance in New York City

Sample Table: Analysis by Flooding Level

N ROW% N ROW% N ROW% N ROW% N ROW% N COL%CLRD 35 73 0 . 6 13 7 15 0 . 48 4CVD 303 72 19 5 47 11 48 12 2 1 419 36P_I 58 80 2 3 10 14 3 4 0 . 73 6

OTHER 494 78 25 4 52 8 59 9 1 0 631 54All 890 76 46 4 115 10 117 10 3 0 1,171 100

CLRD 31 76 3 7 1 2 6 15 0 . 41 4CVD 321 78 12 3 39 10 39 10 0 . 411 36P_I 51 80 2 3 5 8 6 9 0 . 64 6

OTHER 473 75 32 5 63 10 59 9 2 0 629 55All 876 77 49 4 108 9 110 10 2 0 1,145 100

CLRD 31 74 2 5 4 10 4 10 1 2 42 3CVD 284 76 13 4 37 10 40 11 0 . 374 30P_I 69 84 1 1 7 9 5 6 0 . 82 6

OTHER 594 78 36 5 69 9 66 9 1 0 766 61All 978 77 52 4 117 9 115 9 2 0 1,264 100

CLRD 94% 133% 114% 62% -- 94%CVD 91% 84% 86% 92% 0% 90%P_I 127% 50% 93% 111% -- 120%

OTHER 123% 126% 120% 112% 67% 122%All 111% 109% 105% 101% 80% 109%

2010

2011

2012

YEAR/CAUSE Proportion of Census Tract in Inundation Zone All0% >0-10% >10-75% >75% Unknown

Percent Increase

Page 24: Superstorm Sandy  Mortality Surveillance in New York City

Sample Table: Analysis by Flooding Level November 1–10, 2010-2012

N ROW% N ROW% N ROW% N ROW% N ROW% N COL%CLRD 35 73 0 . 6 13 7 15 0 . 48 4CVD 303 72 19 5 47 11 48 12 2 1 419 36P_I 58 80 2 3 10 14 3 4 0 . 73 6

OTHER 494 78 25 4 52 8 59 9 1 0 631 54All 890 76 46 4 115 10 117 10 3 0 1,171 100

CLRD 31 76 3 7 1 2 6 15 0 . 41 4CVD 321 78 12 3 39 10 39 10 0 . 411 36P_I 51 80 2 3 5 8 6 9 0 . 64 6

OTHER 473 75 32 5 63 10 59 9 2 0 629 55All 876 77 49 4 108 9 110 10 2 0 1,145 100

CLRD 31 74 2 5 4 10 4 10 1 2 42 3CVD 284 76 13 4 37 10 40 11 0 . 374 30P_I 69 84 1 1 7 9 5 6 0 . 82 6

OTHER 594 78 36 5 69 9 66 9 1 0 766 61All 978 77 52 4 117 9 115 9 2 0 1,264 100

CLRD 94% 133% 114% 62% -- 94%CVD 91% 84% 86% 92% 0% 90%P_I 127% 50% 93% 111% -- 120%

OTHER 123% 126% 120% 112% 67% 122%All 111% 109% 105% 101% 80% 109%

2010

2011

2012

YEAR/CAUSE Proportion of Census Tract in Inundation Zone All0% >0-10% >10-75% >75% Unknown

Percent Increase

• Overall results:• 0% flooded: 111% of baseline• >0-10%: 109% of baseline• >10-75%: 105% of baseline • >75%: 101% of baseline

Page 25: Superstorm Sandy  Mortality Surveillance in New York City

Investigating Death Increase in Immediate post-Sandy Period

• Up 11% (156 deaths)• Not concentrated in flooded areas• Not concentrated in any specific age/cause group• Resolved after about 10 days as most of city

operations returned to normal outside of flooded areas

• No increase with past subway shutdowns seen• Later second rise in deaths found to related to

onset of influenza season

Page 26: Superstorm Sandy  Mortality Surveillance in New York City

Investigating Death Increase in Immediate Post-Sandy Period: Healthcare Facility Evacuations

• Deaths among evacuated nursing home residents up only 5-6 over prior years (preliminary)

• Deaths at receiving hospitals up ~20-25 deaths after adjustment of baselines (preliminary)– No adjustment made for medical acuity of

transferees– Unable to examine by transferring hospital

• No deaths reported to ME as transfer-related (would be reportable in NYC)

Page 27: Superstorm Sandy  Mortality Surveillance in New York City

Challenges I• Time-consuming ad hoc SQL queries required to extract

needed mortality data from EDRS to get certified data and additional data elements needed

• Needed to quickly resolve current year geocoding rejects and recode prior years to 2010 census

• SAS programs needed to be urgently developed for reports, delaying results despite timely data transfer from death reporters (ideally have canned reports/programs ready)

• De-duplicating pre-registration death certifications for analysis added extra step to analysis

• Difficult to account for evacuation impact

Page 28: Superstorm Sandy  Mortality Surveillance in New York City

Lessons Learned• EDRS effectively adapted to provide disaster-related mortality

surveillance• Pre-registration records increase timeliness but require staff with

SQL programming skills• Local MMDS software and nosologists essential to timely cause of

death coding• Zip code proved too coarse a measure in urban area to accurately

categorize areas by flood levels – Few zip codes with majority flooded and many with just small area

affected• Need to restricting comparison data by certification/registration

date 2°reporting lag• Post-registration free-text in injury searches not helpful in our

setting

Page 29: Superstorm Sandy  Mortality Surveillance in New York City

Next Steps I• Planned investigation for potential excess all-cause mortality

(some pending grant-related resources) – Time-series analysis of deaths counts to better account for

known environmental factors that influence daily death counts (EH)

– Use of hospital discharge data to characterize death increase – More detailed analysis of possible transfer related mortality

using linked hospital discharge data and mortality data (account for medical acuity)

– More detailed analysis of possible nursing home related death transfers with IJE data and possibly chart reviews given small counts

Page 30: Superstorm Sandy  Mortality Surveillance in New York City

Next Steps II• Consider adding DOB and address to DC

certification screens to allow for age and neighborhood analyses on more timely certified only data

• Consider operationalizing reports for expected emergencies (applied for staff as part syndromic surveillance program request)

• Advocate to NCHS for access to updated MMDS software to allow this kind of work for emergency response

Page 31: Superstorm Sandy  Mortality Surveillance in New York City

Food for Thought

• Most post-disaster mortality analyses not based on Vital Statistics– Recent MMWR used red cross data– Stand alone mortality reporting systems have

been used in US for other hurricanes• Issues are timeliness and lack of accurate capture of

disaster-related deaths in some areas (e.g., COD “fall” but does not mention Hurricane as cause)

Page 32: Superstorm Sandy  Mortality Surveillance in New York City

CoAuthors

• Renata Howland, MPH (CSTE epi fellow)• Wenhui Li, PhD (in attendance at NAPHSIS)• Ann Madsen, PhD, MPH• Howard Wong, MSc• Tara Das, PhD (in attendance at NAPHSIS)• Thomas Matte, MD, MPH• Catherine Stayton, DrPH, MPH

Page 33: Superstorm Sandy  Mortality Surveillance in New York City
Page 34: Superstorm Sandy  Mortality Surveillance in New York City
Page 35: Superstorm Sandy  Mortality Surveillance in New York City

All Cause Death Counts, Nov 1–10, 2010–2012

Date of Death 2010 2011 2012

Absolute difference

between 2012 and average of

2010-2011

Percent difference

between 2012 and average of

2010-201111/1/2012 150 153 134 -17.5 -12%11/2/2012 128 131 147 17.5 14%11/3/2012 135 140 157 19.5 14%11/4/2012 157 151 162 8 5%11/5/2012 130 125 161 33.5 26%11/6/2012 133 136 164 29.5 22%11/7/2012 156 159 161 3.5 2%11/8/2012 174 161 164 -3.5 -2%11/9/2012 130 153 172 30.5 22%

11/10/2012 153 150 186 34.5 23%Mean 144.6 145.9 160.8 15.6 11%

Median 142.5 150.5 161.5 15.0 10%Total 1,446 1,459 1,608 155.5 11%

Page 36: Superstorm Sandy  Mortality Surveillance in New York City

Death Registration

Page 37: Superstorm Sandy  Mortality Surveillance in New York City

Mortality Surveillance Findings

• Pre-registration certificates provided timely and relatively high quality data– Address most likely to be missing (80%), delaying

geographical analyses– Retrospective comparison between queried

records and registered records showed high agreement (94%) where fields were present