||Contact | A-Z Health Index | Events & Meetings|
Traumatic Brain Injury Surveillance in Oklahoma (2011)
Traumatic brain injuries are a leading cause of death and disability in the U.S.1 Approximately 1.4 million people sustain a traumatic brain injury each year, resulting in 50,000 deaths, over one million emergency department visits, 235,000 hospitalizations, and 80,000-90,000 permanent severe neurological disabilities.1,2,3 Brain injuries are complex and only rarely are consequences limited to a single deficit. Many survivors with serious injuries experience a constellation of symptoms and impairments, such as physical, emotional, cognitive, and behavioral problems that may require months or years of rehabilitation.1 Injuries do not have to be serious, however. Even mild traumatic brain injuries can result in long-term cognitive problems, which impair an individual’s ability to perform activities of daily living.1,4 In this country, 5.3 million individuals are living with long-term or lifelong traumatic brain injury-associated disabilities; although, this estimate is suspected to under-represent the true number.1 Lifetime costs have been estimated at $60 billion annually.3
From 2004-2009, a total of 27,306 traumatic brain injuries occurred among Oklahoma residents that were fatal or serious enough to require hospitalization. The highest rate of traumatic brain injury was among persons 65 years and older, followed by persons 15-24 years of age. Males were over 1.5 times more likely to be injured than females. Falls and motor vehicle crashes were the most common causes of traumatic brain injury, accounting for 37% and 23% of injuries respectively, followed by gunshot wounds (8%) and assaults (8%). Of the 27,306 traumatic brain injuries in 2004-2009, 5,492 (20%) were fatal. Males had a higher case fatality rate than females (24% and 14%, respectively).
The 2009 Traumatic Brain Injury Data Report was submitted to the National Center for Injury Prevention and Control in October 2011.
The Injury Prevention Service has had the authority to collect and maintain traumatic brain injury surveillance data since traumatic brain injuries were mandated a reportable condition in April 1991 by the Oklahoma Board of Health and the Oklahoma legislature (HJR 1040).
History of traumatic brain injury Data Collection
Statewide surveillance for hospitalized and fatal traumatic brain injuries has been conducted in Oklahoma since 1992 using a standard morbidity and mortality code definition from the National Center for Injury Prevention and Control. Because a complete, consistent hospital discharge database was not available in Oklahoma until January 2005 (2002 data), traumatic brain injury surveillance data were collected directly from medical records for 1992-2003. A contact person was designated at each hospital’s medical records department to work with Injury Prevention Service staff to generate a list of traumatic brain injury patients based on the traumatic brain injury discharge codes and to make medical records available for review. Data elements were collected through medical record reviews by trained Injury Prevention Service staff at all 116 acute care hospitals (including federal facilities) in the state. From 1992-1998, approximately 100 variables, including most of the current basic and extended data elements recommended by the National Center for Injury Prevention and Control, were collected on all hospitalized cases. From 1999-2000, a 50% random sample of hospital medical records was selected and abstracted for both the basic and extended variables. For the remaining 50% of medical records, only the basic variables were abstracted. From 2001-2003, due to reduced funding, only basic variables were collected on all traumatic brain injury cases. A list of medical records that were not available during each hospital site visit was maintained and records were requested until they became available. Traumatic brain injury deaths were identified from the Office of the Chief Medical Examiner from 1992-1999 and from Vital Statistics beginning in 2000.
Traumatic Brain Injury Case Definition
Data were collected on men and women of all ages and racial and ethnic groups among Oklahoma’s 3.5 million residents. The traumatic brain injury mortality case definition codes in the Central Nervous System Injury Surveillance Data Submission Standards—2002 (referred to as Standards in the rest of the summary report) were used.5 Fatal traumatic brain injury cases were identified in the Vital Statistics database by searching all 20 multiple cause of death code fields for a code indicating a traumatic brain injury. Oklahomans who died out of state were included. For nonfatal injuries, the traumatic brain injury morbidity case definition codes in the Standards were used. Hospitalized traumatic brain injury cases were identified in the inpatient hospital discharge database by searching the principal diagnosis and all 15 other diagnosis codes for a code indicating a traumatic brain injury. Traumatic brain injury cases were limited to Oklahoma residents who died or were discharged from an acute care hospital during 2004-2009. Persons injured more than 12 months before the date of discharge or death were excluded.
The Oklahoma State Department of Health Vital Records Division maintains death certificates on all deaths that occur in the state. Death certificates are coded to multiple causes by the National Center for Health Statistics. A real-time electronic vital statistics file, which includes all deaths in Oklahoma and deaths of Oklahoma residents that occurred outside of the state, is made available to the Injury Prevention Service through the health department intranet and can be accessed daily. A final centralized statewide electronic database of deaths for the year, including multiple cause coding and personal identifiers such as name, date of birth, and date of death is obtained by the Injury Prevention Service annually.
Data for the centralized statewide hospital discharge database are collected and maintained by the Health Care Information Division of the health department. The hospital discharge database includes all state licensed general and specialized hospitals; federal government facilities are excluded. Discharge data records are submitted for persons discharged within a calendar year from all hospital beds. A separate record is submitted for each discharge, including information on the patient, provider, service, diagnosis and treatment, payer, and charges/bill type. A comprehensive data quality program is run on the database, including checks to ensure that all required fields are completed, ages are appropriate (0-115 years), date fields have the correct year of discharge and proper date sequences, there are not duplicate records, E codes are present for injury-related discharges, etc. Letters regarding missing and inappropriate data are sent to hospitals to obtain updated information and/or clarification. The Injury Prevention Service receives a finalized hospital discharge database each year. Personal identifiers, including name, date of birth, last four digits of the social security number, and medical record number are obtained by the Injury Prevention Service for all reportable injuries, including traumatic brain injuries.
During 2004 traumatic brain injury data collection, the Injury Prevention Service collected data for all traumatic brain injury cases from the hospital discharge database on all available basic variables as described in the Standards. In addition, all available extended variables were obtained during medical record reviews of a sample of 1,200 cases; these variables were updated in the traumatic brain injury database. In 2006, numerous conference calls were held with traumatic brain injury funded states and the National Center for Injury Prevention and Control to enhance and update the Standards. In a document used by funded states, but yet to be formally published and released, basic and extended data elements were clarified and updated. In addition, 44 new variables were added. These revised and new elements were used in the collection of 2005-2009 sampled traumatic brain injury data. The creation of new variables was part of the original grant proposal guidelines, which charged grantees with identifying a topic of emerging public health importance and collecting additional information on that topic during regular traumatic brain injury surveillance. Funded states selected falls among individuals aged 65 years and older as the emerging public health topic and 37 of the 44 new variables related to this module. The remaining seven new variables were to be collected on all traumatic brain injury sampled cases.
Data Linkage and Sampling Methodology
Deaths in the 2004-2009 centralized electronic hospital discharge database and vital statistics database were linked using the probabilistic linking software SAS LinkPro. For persons with multiple hospitalizations for the same event, back-to-back stays were combined and the definitive care hospital was documented. Patients transferred from one hospital to another were identified using the source of admission and personal identifiers. Protocols were established to determine if non-consecutive stays were for the same injury (deleted second stay from the database) or for a second injury (included second stay in the database). Discharges for the same person that occurred 2-10 days after the initial stay were removed from the database unless the external cause of injury code (E code) indicated a different type of injury (e.g., fall and motor vehicle crash). If subsequent discharges occurred 11 or more days later, the stays were considered to be related to separate injuries and were included in the database. Without a comprehensive review of all medical records, it was unknown exactly how many of the discharges were for follow-up care of a previous injury. Persons in the vital statistics database were also compared to patients in the hospital discharge database using personal identifiers to identify and combine duplicates.
As specified in the Standards, a representative sample of at least 1,000 traumatic brain injury cases (preadmission deaths not included) needed to be successfully abstracted using a stratified sampling approach. An initial sample of 1,200 cases was selected using 2004 data to allow for false positive cases and missing records. Hospitalized cases were divided into strata based on hospital size (<100 beds and ≥100 beds). The proportion of cases in each of the hospital stratum was calculated and the sample followed the same proportions. The sample was formed by selecting the predetermined number of cases in each stratum from a randomized, stratified database of traumatic brain injury cases. After 2004 data produced a very low number of false positives, the sample size was reduced to 1,050 for subsequent years.
Extended Medical Records Surveillance
The Injury Prevention Service abstracted all basic and extended variables for all sampled cases. On average, one abstraction took 15-20 minutes for an experienced abstractor. Cases requiring the completion of the additional fall variables involved at least five more minutes per record. Data were recorded on site at the selected hospitals on a hard copy abstraction form. Data for 2004-2005 were obtained by telephone from a medical records employee for hospitals that had only one or two randomly selected cases. Beginning with 2006 data, hospitals with less than 10 cases (depending on the distance to the hospital) submitted copies of the records by mail. The Injury Prevention Service determined and documented if the case met the National Center for Injury Prevention and Control’s clinical case definition for a traumatic brain injury as defined in the Standards. Quality assurance measures included double-checking forms for missing or inconsistent information and periodic double reviews to assess inter-rater reliability. In addition, a document was produced with definitions for most variables and notes about unusual/confusing issues, which was readily available during all abstractions. A list of sampled records was maintained and information on the basic and extended variables from the abstraction forms was entered under security into an Microsoft Access file. The file was combined with a database of non-sampled cases. Abstraction forms were kept in locked cabinets in a locked room. As outlined in the Standards, traumatic brain injury cases found to be false positives during abstraction were included in the data set submitted to the National Center for Injury Prevention and Control; false positive cases were flagged by the “abstract” variable.
Usefulness of the Data
The need for standardized traumatic brain injury data is well documented.6 In Oklahoma, traumatic brain injury data are needed for describing the problem and demand for services and for funding treatment, prevention, and research. Standardized data on traumatic brain injury allow the identification of high-risk populations and risk factors and the development of targeted prevention programs and evaluations. Data also enable policymakers and the public to put various health conditions in perspective. It is also important to frame the costs of treating injuries versus expenditures for prevention.
Limitations of the Data
The Oklahoma traumatic brain injury surveillance system excluded individuals with less severe traumatic brain injuries who were treated in an emergency department and released home and individuals treated in a physician’s office. Traumatic brain injury data for hospitalized cases were obtained from the hospital discharge database. The hospital discharge database did not obtain data from federal hospitals (military and Native American facilities) nor did it include Oklahoma residents who were hospitalized out of state.
Persons with two consecutive hospital stays were likely transferred from one facility to another for additional care for the same injury and the two stays were combined. If there was a gap between stays, it was difficult to determine if the person was readmitted for the same injury or suffered a second injury. Without a date of injury variable in the hospital discharge database, standardized methods were used to classify cases as one injury or two separate injuries; however, all cases may not have been correctly classified.
The traumatic brain injury database was evaluated prior to being sent to the National Center for Injury Prevention and Control to ensure complete and quality data. All cases were reviewed to verify that they had a traumatic brain injury code, were discharged or died in the surveillance year, were residents of Oklahoma, and were not duplicate records. Date fields were checked for improper sequences and conflicts. Frequencies were run on all variables to ensure completeness of the data and to verify that they were in the proper format required by the National Center for Injury Prevention and Control. Checks were made to ensure that false positive cases were included and identified in the database and that all available ICD-9-CM, ICD-10, and E codes were submitted in the proper position. The crude death and hospitalization incidence rates were calculated and compared to previous years. Predictive value positive was calculated based on medical records reviews to determine the probability that persons with a traumatic brain injury code actually experienced a traumatic brain injury. A marker for sensitivity was also calculated based on cases of hospitalized traumatic brain injury deaths.
Preparing for Site Visits
Collecting Data from Hospitals
After the Visit/Telephone Review/Submission of Copies
(not a comprehensive listing of variables)
County of Injury:
Race (7 variables):
Abbreviated Injury Scale (AIS) Score—Head:
Blood Alcohol Level:
Glasgow Outcome Scale (GOS) Score:
Length of Time Unconscious:
Personal Protective Equipment:
Module Variables—All Patients
Clinical Case Definition:
Other Motor Vehicle Type:
Module Variables—65 Years and Older Fall Patients
The difficulty with the majority of the fall-related module variables was twofold. First, there was a significant lack of documentation in the medical records that hindered the ascertainment of specific circumstances of the fall (e.g., factors and objects involved), which left many variables with a high percentage of unknown values. Secondly, without a comparison group of older adults with similarly abstracted information, it remains challenging to interpret and draw conclusions on the traumatic brain injury sample beyond what is already well documented in the literature. Of course, surveillance is not research, but it is important in the planning of prevention initiatives to understand if/how the fall-related variables are unique to fall patients rather than being a part of aging in general. Surveillance should initiate control and prevention activities; this purpose becomes a struggle when variables are missing so many details (up to 66% unknown).
Factors Involved with the Fall:
Location at the Time of the Fall:
Objects Involved with the Fall:
Mail merge was again used to import known information onto abstraction forms prior to visiting each hospital. The information came from the database used to select the sample and included the following variables: hospital name, medical record number, patient’s name, patient’s address (including street, city, county, and zip code), patient’s last four digits of his/her social security number, patient’s date of birth and age, patient’s race, Hispanic ethnicity, all diagnosis and E codes, and discharge disposition. It was much quicker to verify these data in the medical record, than to abstract them on site.
Out of the 76 hospitals that had at least one sampled traumatic brain injury record, 61 facilities had 10 or fewer records. All of these facilities were willing to mail copies of the selected records. IPS staff then had several “in-office” surveillance days where a conference room was reserved and abstractors worked as if they were reviewing records at a hospital. Not only did this option save time and travel for Injury Prevention Service staff, but hospitals seemed to prefer it over having to dedicate personnel to conduct telephone reviews or cater to a site visit.
The abstracting team remained the same as the year prior. Aside from a brief refresher, little time was needed for training.
Beginning with 2005 data collection, 44 new variables were added to the Standards. The Injury Prevention Service tested all new variables for that year’s data, except for one (the optional variable on antihypertensive use among older falls). For 2006 data collection, the National Center for Injury Prevention and Control made several more of these variables optional. Using what was learned during 2005 data collection and afterwards assessing the value of each optional variable, the Injury Prevention Service decided to drop several more optional variables that had exceptionally large rates of missing information or were infrequently positive (i.e., AVPU, frailty, incontinence, multiple sclerosis, peripheral neuropathy, seizures/epilepsy, syncope, hip fracture, and wrist fracture). This decision improved efficiency without affecting data quality or scope.
Note: The following links open in a new window.
Injury Prevention Service
Publication Date: October 2011
Page last updated: October 2012
Copyright © State of Oklahoma