1. Automated Monitoring of Injuries Due to Falls Using the BioSense System
Achintya N. Dey, MA1, Jerome I Tokars, MD MPH1, Peter Hicks, MPH2, Matthew Miller, BA2 and Roseanne English,BS1
1 Division of Emergency Preparedness and Response
National Center for Public Health Informatics
Centers for Disease Control and Prevention
2 Constella Group, LLC – an SRA International company
The findings and conclusions in this presentation are those of the authors and do not necessary represent the views of the Centers for Disease Control and Prevention

2. BioSense System
National real-time biosurveillance system to enable public health situational awareness
Current data sources: 569 hospitals, >1100 ambulatory care Departments of Defense (DoD) and Veterans Affairs (VA) medical facilities, 2 large national laboratories
Data stored, analyzed, and visualized in a secure web-based application
BioSense is an innovative biosurveillance program which is increasing the nation’s emergency preparedness through the development of a national network for real time outbreak detection, monitoring, and health situational awareness.
The system receives both chief complaint and diagnosis data from over 569 non-federal hospitals and over 1100 Department of Defense (DoD) and Veterans Affairs (VA) ambulatory care facilities.

3. Background
Falls are the leading cause of nonfatal medically attended injuries in the U.S.
Falls accounted for 21% of injury-related visits to U.S. emergency departments (ED) in 2005
Falls are one of 11 injury-related sub-syndromes currently tracked by BioSense
Falls are one of the leading causes of nonfatal injuries in the US
Recent data from National Hospitals Ambulatory Medical Care Survey showed that falls accounted for 21% of all injury related visits to EDs.
Falls are one of 11 injury related subsyndromes currently tracked by BioSense
Although fall related injuries occur throughout the year, only a few studies have examined it’s association with weather.

4. Objectives
Identify and characterize clusters of falls in metropolitan areas during the winter season
Assess association between falls clusters and severe weather
Our objectives for this inquiry were to:
Identify and characterize clusters of falls in metropolitan areas during the winter season and to assess the association between falls clusters and severe weather.

5. Methods
Studied chief complaints of fall in 19 metropolitan areas with ≥2 participating ED facilities
Study period October 1, 2007 – March 31, 2008
We studied chief complaints related to falls in 19 metropolitan areas with at least 2 or more ED facilities.
The study period was from October 1st thru March 31st, 2008

6. Methods (cont..) Identified clusters of falls based on:
Time series analysis: modified EARS C2 algorithm
Recurrence interval ≥ 500 days (p<0.002)
Observed/expected ≥2
Observed-expected (excess visits) per day ≥10
Identified falls related to snow or ice by searching chief complaints for “fell on ice,” “fell due to ice,” “trip on ice”
Identified associated fractures (ICD-9 codes )
Falls clusters were identified based on:
A time series analysis which is a modified C2 = (Observed – Expected) / SD
A recurrence interval of 500 days and over
The ratio of observed and expected value at 2 or more
And 10 or more excess visits per day

7. Fall Anomaly Clusters Cluster # Metro Area Region Anomaly Date(s) 1 A
South-Atlantic
12/6/07 2
02/12/08 – 2/13/08 3 B
New England
12/10/07 4
12/17/07 5 C
East North Central
12/9/07 – 12/10/07 6
02/9/08 7 D 8 E 9
02/17/08 10 F
West North Central
12/8/07 – 12/9/07 11 G
11/10/07 12 H 13 I 14
We found 14 clusters of falls in 10 metropolitan areas.
5 of the clusters were multiday
9 of the clusters occurred during Dec 6-18, 2007, which was a period of heavy snowfall in the East and Midwest regions.
4 of the clusters occurred during February 9-17, 2008, and were coincident with freezing rain conditions

8. Fall Anomaly Clusters (December 2007)
Metro Area
Region
Anomaly Date(s) 1 A
South-Atlantic
12/6/07 2
02/12/08 – 2/13/08 3 B
New England
12/10/07 4
12/17/07 5 C
East North Central
12/9/07 – 12/10/07 6
02/9/08 7 D 8 E 9
02/17/08 10 F
West North Central
12/8/07 – 12/9/07 11 G
11/10/07 12 H 13 I 14
We found 14 clusters of falls in 10 metropolitan areas.
5 of the clusters were multiday
9 of the clusters occurred during Dec 6-18, 2007, which was a period of heavy snowfall in the East and Midwest
4 of the clusters occurred during February 9-17, 2008, and were coincident with freezing rain conditions

9. Fall Anomaly Clusters (February 2008)
Metro Area
Region
Anomaly Date(s) 1 A
South-Atlantic
12/6/07 2
02/12/08 – 2/13/08 3 B
New England
12/10/07 4
12/17/07 5 C
East North Central
12/9/07 – 12/10/07 6
02/9/08 7 D 8 E 9
02/17/08 10 F
West North Central
12/8/07 – 12/9/07 11 G
11/10/07 12 H 13 I 14
We found 14 clusters of falls in 10 metropolitan areas.
5 of the clusters were multiday
9 of the clusters occurred during Dec 6-18, 2007, which was a period of heavy snowfall in the East and Midwest
4 of the clusters occurred during February 9-17, 2008, and were coincident with freezing rain conditions

10. ED Chief Complaint of Fall, 26 Facilities, Cluster 5, Metro Area: C
In the next few slides you will see screen shots of the BioSense application, demonstrating the large clusters of falls we identified.
On December 9th and 10th in metropolitan area C, we identified very large clusters of falls. On December 9th there were 233 excess visits to the ED; and on Dec 10th there were 136 excess visits.
Source: BioSense Application
Excess visits for falls 233, 136

11. ED Chief Complaints of Fall, 17 Facilities, Cluster 8, Metro Area: E
Similarly, in December 9th and 10th in metropolitan area E, we also identified large clusters of falls.
Both of these clusters of falls coincide with heavy snowfall in the East, North Central area of the country.
Excess visits for falls: 137, 197

12. ED Chief Complaints of Fall, 17 Facilities, Cluster 9, Metro Area: E
This cluster of falls was seen on Feb 17th in metropolitan area E
Excess visits for falls: 100

13. ED Chief Complaint of Fall, 2 Facilities, Cluster 14, Metro Area: I
These 2 clusters of falls were detected on Feb 12th and 13th in metropolitan area I.
Both of these clusters coincide with the freezing rain in the South-Atlantic and East North Central regions.
Excess visits for falls 40, 39

14. ED Chief Complaint of Fall, 24 Facilities, Cluster 12, Metro Area: H
This is another cluster of falls detected on Dec 10th in metropolitan area H of the West North central region.
Excess visits for falls: 117

15. Metro Area: E, Cluster 7, by Age Group
Here we see a falls cluster in metropolitan area E, stratified by age group.
As you can see, a significant increase of falls injuries among the age group
Source: BioSense Application

16. Patient List, BioSense Application
This is a screen shot of the patients list module in the BioSense application.
This column shows the chief complaint and this other column shows the final diagnosis.

17. The Chief Complaint shown is a broken arm due to falling on ice
The Chief Complaint shown is a broken arm due to falling on ice. The final dx confirms a fracture.

18. Summary of Falls Clusters, 2007-2008
14 clusters of falls in 9 metro areas
Total ED visits for falls: 2,394
Excess ED visits for falls: 1,593
Time period
Dec 6-18, 2007, 9 clusters (winter storm)
February 9-17, 2008, 4 clusters (freezing rain
During the winter of , the BioSense application identified 14 clusters of falls in 9 metropolitan areas. Total ED visits for falls was 2,394 and excess visits for falls was 1,593.
9 of the clusters occurred during Dec 6-18, 2007, which coincided with the winter storm.
4 of the clusters occurred during Feb 9-17, 2008 and coincided with freezing rain

19. Characteristics of ED Visits for Falls
Demographics
Mean age: 47 years
57% were women
Mention of “ice” or “snow” in chief complaint: 9%
Associated fracture
33% had a final ICD-9 coded diagnosis
15% of these had an ICD-9 code for fracture
Mean age of patients was 47 years; 57% were women
9% mentioned ice or snow in their chief complaint
33% had a final ICD-9 coded diagnosis and 15% of these had an ICD-9 code for fracture

20. Percentage of Visits for Falls, by Age and Time Period
This bar graph shows the percentage of visits for falls by age group and time period.
Here we compared number of falls in cluster days, with the baseline. Baseline is identified as the 28 days before the clusters occurred.
Baseline ED visits for falls was highest among the 50 years and over age group.
However in cluster days, falls increased for all age groups.
Baseline=28 days before clusters
Cluster=Day of falls clusters

21. Relative Risk* of Falls by Age Group
This slide represents the ratio of percent of visits with falls on cluster days to base line days.
Relative risk of falls is broken up by age group.
As you can see, the highest ratio of falls occurred among the age group 20 to 49, followed by years old.
*Percent of visits with fall on cluster days/baseline days

22. This is a news paper article related to the ice storm in one of the Metropolitan areas during February 2008, which coincides with 4 falls clusters we detected during that same time period.

23. Discussion Limitations of geographic coverage and data availability
Identified several large clusters of ED visits due to falls associated with severe weather
Increase in falls visits highest for years
Several earlier studies have shown similar results
A limitation of our study is that we included only metropolitan areas with facilities sending data to BioSense. Currently BioSense receives data from 12% of the EDs in the US.
Inspite of this, our automated surveillance system was able to identify several large clusters of ED visits due to falls associated with severe winter weather primarily affecting year olds.
Several earlier studies, such as the (1) NIH study on injuries associated with falls during ice storms in 1994; and the (2) study in North Carolina on injuries from the 2002 ice storm, have shown a significant increase of fall injuries related with the winter storm. However both of these studies were based on retrospective chart reviews.

24. Discussion, Utility Surveillance for falls
Data available in near-real time
Permits rapid detection of increases in falls injuries
May be helpful in determining the effectiveness of public health campaigns
May be helpful in prevention programs
Surveillance for injuries due to natural disasters such as hurricanes in real time
Provide descriptive data
Track geographically
The advantage of using BioSense is that data was available in real time, and permitted rapid detection of increased fall injuries during the time of disaster.
It may be helpful in determining the effectiveness of public health campaigns and also may be helpful in prevention programs.
During a natural disaster, such as a hurricane or ice storm, BioSense can provide real time descriptive data on injuries and track the data geographically.

25. Thank You BioSenseHelp@cdc.gov25