ACADEMIC EMERGENCY MEDICINE 2011; 18:972–976 © 2011 by the Society for Academic Emergency Medicine
Objectives: The objective was to determine if geospatial techniques can be used to inform targeted community consultation (CC) and public disclosure (PD) for a clinical trial requiring emergency exception from informed consent (EFIC).
Methods: Data from January 2007 to December 2009 were extracted from a Level I trauma center’s trauma database using the National Trauma Registry of the American College of Surgeon (NTRACS). Injury details, demographics, geographic codes, and clinical data necessary to match core elements of the clinical trial inclusion criteria (Glasgow Coma Scale [GCS] 3–12 and blunt head injury) were collected on all patients. Patients’ home zip codes were geocoded to compare with population density and clustering analysis.
Results: Over a 2-year period, 179 patients presented with moderate to severe traumatic brain injury (TBI). Mapping the rate and frequency of TBI patients presenting to the trauma center delineated at-risk populations for moderate to severe head injury. Four zip codes had higher incidences of TBI than the rest, with one zip code having a very high rate of 80 per 100,000 population.
Conclusions: Geospatial techniques and hospital data records can be used to characterize potential subjects and delineate a high-risk population to inform directed CC and public disclosure strategies.
Research in the fields of critical care medicine and resuscitation have been challenging due to the urgent nature of the disorders being studied and the need for rapid intervention. A major roadblock to conducting these trials has been the time required to obtain informed consent from patients prior to study enrollment. In emergency and resuscitation research, obtaining informed consent during time-critical conditions has significantly decreased the number of resuscitation studies in the United States and possibly hindered scientific advancement and public health improvements.1
In 1996, the Food and Drug Administration issued the Final Rule 21 CFR 50.24 (Exception from Informed Consent [EFIC] Requirements for Emergency Research). When specific criteria are met, this rule provides a mechanism for the enrollment of human subjects into clinical trials without their consent. The rule requires investigators to conduct community consultation (CC) and perform public disclosure (PD) before initiating the trial. However, many questions remain about what defines a relevant community for the study in question and how broadly to deploy PD activities.
In trials using EFIC, CC activities have included local community surveys, focus groups, local “town hall” meetings, and random-digit dialing. For PD, numerous methods have been employed, but most common are print media (newspaper articles or advertisements in local newspapers), radio broadcast media, flyers, and electronic media. There are limited data on the effectiveness of these different modalities for CC or PD.2,3
Despite broad efforts to engage the community of interest (those most likely to be affected by the study), as few as 5% to 8% of respondents in some studies were aware that EFIC studies were being performed in their communities, and less than half of those who were aware of the study could name one risk or benefit associated with study participation. Attendance at “town halls” or community forums has been dismal.4,5 The optimal methods for CC and PD are still open to debate.
Clinical trials in traumatic brain injury (TBI) present a major challenge, as patients do not have the capacity for decision-making, and the intervention must be instituted rapidly. The ProTECT III trial is a multicenter randomized double-blind placebo-controlled clinical trial of progesterone for the treatment of TBI that has FDA clearance to carry out the study under EFIC. The trial is being conducted through a large National Institutes of Health–funded network (the Neurological Emergencies Treatment Trials network). Although ProTECT III is a national study involving up to 31 hospitals in 15 different U.S. states, all CC and PD activities are conceived and implemented locally. The activities were not guided by the data in this report, as they had already been conducted. However, we used the ProTECT III trial and its inclusion and exclusion criteria as an example of how to determine if using geographic information systems (GIS) and hospital data could better define the community of interest for CC or PD activities.
This was a retrospective review of moderately to severely blunt head injured patients admitted to the trauma service with information on residence zip codes. This study received expedited approval from the university’s institutional review board due to the retrospective deidentified data required.
Study Setting and Population
The study was performed at a Level I trauma center with an annual volume of 105,000 emergency department (ED) visits and 2,800 trauma admissions lasting more than 24 hours. Admitted trauma patients were entered into the National Trauma Registry of the American College of Surgeon (NTRACS). Data were extracted for moderately to severely brain injured patients who presented from January 2007 to December 2009 with head injury by blunt mechanism and had an initial Glasgow Coma Scale (GCS) score between 3 and 12 (an inclusion criteria for the ProTECT III trial).
Data were extracted by the hospital NTRACS coordinator and included mechanism of injury, patient demographics (age, sex, race, town, and zip code of residence), injury severity data including Revised Trauma Score (RTS), which includes GCS, as well the Injury Severity Score (ISS), and the probability of survival (Ps) by Trauma and Injury Severity Score (TRISS) methodology.6 Initial GCS and vital signs were collected from the ED initial assessment nursing documentation and used to calculate the RTS. If the mental status was waxing and waning, the best GCS score documented during the patient's initial assessment was used. ISS is calculated by NTRACS nursing personnel who assigned an abbreviated injury score based on the final injury diagnoses.
Descriptive statistics of the individual injury cases were calculated for various demographic variables (e.g., race, age, and sex) and injury severity in SPSS 16.0 (SPSS, Inc., Chicago, IL). The injury data were then summarized according to the zip codes where patients lived when the injuries occurred. The resulting frequencies were joined with 2000 census data for Zip Code Tabulation Areas, and zip code boundaries that were downloaded from the U.S. Census Bureau (http://www.census.gov) in ArcGIS 9.3 (Environment System Research Institute, Redlands, CA). The injury rate per 100,000 people was calculated using the total population at each zip code.
Between 2007 and 2009, a total of 179 patients presented with moderate to severe TBI as defined by a GCS score of between 3 and 12 and were enrolled in NTRACS. Patients with TBI were more likely to be male (67%), Caucasian (50.3%) or African American (39.7%), and middle-aged (42.7 years; SD ± 18 years). Mechanism of injury data are listed in Table 1. The mean (±SD) RTS was 4.37 (±1.93), mean (±SD) ISS was 20.2 (±13.6), and mean (±SD) number of intensive care unit days was 9.98 (±10.94). The mortality rate was high (31.3%), with most deaths occurring at a mean (±SD) age of 44 (±18) years, in men (62%), and in road traffic injuries (79.6%). Deaths were more likely among patients with worse ISS and RTS.
|Cause of Injury||n||%|
|Road traffic injuries||114 of 179||63.7|
|Motor vehicle||74 of 114||64.9|
|Motorcycle||21 of 114||18.4|
|Pedestrian||17 of 114||14.9|
|Bicycle||2 of 114||1.8|
The spatial patterns of where TBI patients live reveal several areas of concern in terms of the frequency and rate. Figure 1 shows the frequency of TBI during the study time period. The figure shows a higher number of TBI in urban areas compared to rural areas. Four zip codes within the study region (labeled 1–4 in the figures) have the most TBI patients. The zip code region 1 had the highest number of TBI patients, especially when comparing to neighboring zip codes. Figure 2 shows the rate of head injury per 100,000 people. When controlling for population, TBI rates are higher in the rural areas than urban regions. Within the city, zip code region 5 notably has the highest rate of TBI. Table 2 shows demographics of TBI patients in the zip codes with the highest rates and frequencies and the demographics of these zip codes.
|Region, by Zip Code||Study Findings||Census Data by Regions|
|N||Rate per 100K||% Patients||Mean age, yr (±SD)||% Male||% Black or African American||Population||% Male||Median age,* yr||% Black or African American|
Given the paucity of literature describing the best mechanisms of CC and PD, this study attempted to use GIS and patient zip code to determine a more specific target population. In determining GIS “hot spots” for focusing CC initiatives, there are three main factors that must be evaluated: the frequency of the event, the time period, and the geography or spatial footprint. To ensure that GIS and retrospective analysis can direct CC and PD, the highest frequency over the shortest time period in the smallest spatial footprint would most accurately detect hot spots. Therefore, in preparation for using GIS for CC and PD, each of these factors must be evaluated.
In our study, we analyzed both the TBI frequency and the rate per 100,000 population. The benefit of looking at the overall frequency is the ability to reach the greatest number of potential patients with the smallest intervention, using population density as an asset. Frequency of TBI as defined as rate per 100,000 can determine populations that are at greater risk, essentially removing population density to allow for other risk factors to be evaluated. While evaluating frequencies, keep in mind relationships between the event and its causes; for instance, blunt head injuries are commonly caused by road traffic crashes occurring outside city limits on open highways or deserted roads. Figure 2 shows several regions with high frequencies that a simple rate-driven hot spot would fail to identify. Complicating the sampling size is population growth, movement, and changes over time. The shortest time period for a retrospective review would ensure the least amount of population “drift.”
Geographic location or region size is important to be able to delineate hot spots and link community demographic information. Routinely, NTRACS databases include patient zip code or city, but not address. Utilizing patient zip code as the geographic region is not optimal; zip codes differ in geographic size and population size, they are developed by the US Postal Service (USPS) for mailing purposes without regard for neighborhoods, and are subject to change as determined by the USPS.7,8 Census data are collected on census block groups that do not match zip code boundaries. Obviously, the most accurate method of determining the appropriate communities for intervention would be to use patient addresses, but these data are not always collected. A second difficulty with using zip codes as the spatial descriptor is the number of zip codes per region or the population within that zip code. For instance, a city could use one zip code that would not allow for specific localization of at-risk communities. It is for this reason that census blocks are often used for spatial analysis, as they have a fixed population size. Some authors have suggested using Federal Information Processing Standards (FIPS) codes and that other geocoded data be included in trauma registries for not only spatial analysis and improved care, but also prevention initiatives.9
This was a retrospective review of data confined to a small sampling of patients at a single trauma center. However, the purpose of the study was to show conceptually how GIS could be used to focus CC efforts, so the retrospective nature of the data, the risk of random sampling error, and the local focus of the study do not affect the purported significance. In addition, there were no geographic data on the location of accidents; therefore, this information could not be used to target efforts. While the study raises the possibility of using GIS as a tool for targeting CC and PD analysis, it did not evaluate the effect of this method. The next study would be to use GIS data to target CC and PD in a portion of the identified regions of interest and determine its effectiveness compared to other methods.
Geographic information systems can be used to determine the frequency and rate of events in a spatial footprint. These hot spots can be used to guide community consultation and public disclosure to fulfill exemption from informed consent Final Rule requirements.