... skating>, fell on thumb.' 'Right thumb' 'Broken

1
Measuring Injury Using the National Health
Interview Survey

• Margaret Warner, PhD
• Lois A. Fingerhut, MA
• Pat Barnes, MA

2
Location of injury and poisoning data

• Injury Episode file
• Injury Verbatim file
• Poison Episode file
• Person file

3
Injury episode file

• For each person
• Up to 4 injury episodes
• For each episode
• Up to 4 ICD-9-CM diagnosis codes
• Up to 3 ICD-9-CM external cause codes

4
Available data in the Injury Episode file

• External cause (ICD-9-CM)
• Injury diagnoses (ICD-9-CM)
• Place of occurrence
• Activity
• Resulting limitations
• Detail on certain causes (e.g. mv crashes, falls)

5
Injury episode file
ICD-9-CM diagnosis codes and E-codes are based on
information supplied by respondents, not
medically trained personnel.
6
Poisoning episode file

• Contains no ICD-9-CM codes or E-codes
• Contains 47 episodes coded as Allergic/adverse
  reaction to medical or other substance or
  Something else not poisoning

7
Available data in the Poisoning Episode file

• Cause of poisoning
• Call to poison control center

8
Available data in both injury and poisoning files

• Date of episode
• Hospitalization
• Days out of school
• Days out of work
• Time elapsed between episode and interview

9
Injury sectionPerson file

• Contains summary information about injury and
  poisoning episodes
• Annual estimates can only be made on variables
  INJCT and POICT

10
How do I put the data together?
11
Injury and poisoning episode, and injury
verbatim files

• Contain only injury or poisoning information
  and unique identifiers
• Must be linked to the person file to obtain
  sociodemographic variables, design variables,
  etc.

12
Episode and person based files
Episode based A person will appear in these
files as many times as he/she has a unique
episodes Person based A person will appear in
this file only once
13
Concatenating injury and poisoning episode files

Injury
Poisoning
Injury
Poisoning
Missing data, questions not found in both files
14
Merging injury/poisoning episode file and person
file

Injury
Person
Poisoning
Missing data, questions not found in both files
15
Merged injury/poisoning episode file and person
file
Injury
Poisoning
Person
Missing data, questions not found in both files
16
Episodes and Conditions

• Episode Event that caused the condition (e.g.
  fall down stairs)
• Conditions physical harm resulting from the
  injury episode (e.g. broken leg, bruised arm)

17
Creating condition filesInjury
Injury Episode with ICD9_1 ICD9_2 ICD9_3 ICD9_4
ICD9_1
ICD9_2
ICD9_3
ICD9_4
18
Creating condition filesPoisoning
Poisoning episode
Poisoning condition

19
Concatenating injury and poisoning condition files

Injury conditions
Poisoning
Injury conditions
Poisoning conditions
Missing data, questions not found in both files
20
What Can I Do With the Data?
21
Handout

• Using SAS to manipulate the injury and poisoning
  data
• Example SAS programs for concatenating and
  merging the files

22
Analyzing the injury and poisoning data

• Estimates of frequency
• Percent distributions
• Rates
• Standard errors

23
Calculating annual estimates

• Annual est (variable)(4)(WTFA)
• Annual estimates of episodes and conditions can
  be calculated.
• Annual estimates of the number of persons
  injured cannot be calculated due to the 3 month
  reference period.

24
Standard Errors (SE)

• SE of the injury/poisoning rate for males and
  females
• SE of the injury/poisoning rate for specified
  racial groups

25
Estimates of frequencyExamples

• Number of episodes resulting from specific
  causes
• Number of injury/poisoning episodes by sex
• Number of injury/poisoning episodes by mechanism

26
Results
27
Rates Examples

• Annual injury/poisoning rates for males and
  females
• Annual injury/poisoning rates by mechanism
• Annual rate of falls for females over 65 years
  of age

28
Injury and poisoning episode rates
Age in years
95 CI
Episodes per 1,000 population
29
Age adjusted injury episode rates by mechanism of
injury
30
Age adjusted condition rates by type of injury,
1997
31
PercentagesExamples

• Percent of episodes resulting from specific
  causes
• Percent of poisoning episodes in which a poison
  control center was called
• Percent of injury episodes that occurred in the
  home

32
Percent of injury episodes by place of injury,
1997
Percent of episodes
Persons could indicate up to two places per
episode.
33
Work-related injuries

• Injury episode file
• Activity at the time of injury
• Working at a paid job
• Sample person file
• Occupation and industry

34
Percent of injury episodes by activity, 1997
Percent of episodes
Persons could indicate up to two places per
episode.
35
Injury verbatim file

• Responses recorded by the interviewer
• How the injury occurred
• Body part injured
• Nature of injury
• Text file linkable to other files

36
Available data in injury verbatim

• How the injury occurred (up to 336
  characters)
• Up to 4 body parts injured (up to 34
  characters each)
• Up to 4 nature of injuries (up to 44
  characters each)

37
Injury verbatim file

• Edited only for confidentiality
• Grammatical and/or spelling errors were not
  corrected

38
Uses of injury verbatim

• NCHS used to code ICD-9-CM diagnoses and
  external causes
• Code according to other classification schemes
  (e.g. BLS, ICECI, etc)
• Text search for words of interest
• In-depth analysis of ICD-9-CM categories

39
Check sample size
Analysts should be cautioned against making
estimates based on small numbers of
observations.
40
Verbatim textExample

• Cause of injury .
• Part of body
• Type of injury

• While ltin-line skatinggt, fell on thumb.
• Right thumb
• Broken

41
Examples of verbatimMachinery related injuries
(E919)

• He was at home working on his table saw and
  accidentally ran his left thumb through the
  machine lengthwise. (table saw, n8)
• Forklift extension came loose dropping on
  right foot at work (forklift, n3)

42
Contact for questions related to injury data or
injury classification issuesMWarner_at_cdc.gov