Utility of the Injury Severity Score: A Confirmation by John L. Semmlow and Randall Cone
Independent results obtained from Illinois Trauma Registry data confirm an earlier finding of a fixed, monotonic relationship between expected mortality and the Injury Severity Score (ISS) value of injuries sustained in both vehicular and nonvehicular incidents. Further results show the ISS measure to be a good indicator of hospitalization demand from accident victims and suggest that this variable can be used to control for variations in injury severity in comparative evaluations of EMS systems. The unique and immediate needs of the critically injured patient have proven a strong stimulus toward the development of comprehensive systems for the delivery of emergency medical services (EMS) [1-3]. Effective management of the necessary specialized facilities and personnel is essential if such care is to be available to a broad spectrum of the population; in fact, it is only through better utilization of presently available resources that any immediate improvements can be effected in a poorly-organized emergency health system [4]. Recognition of this situation has stimulated widespread interest in EMS planning, and with it the demand to develop effective methods for evaluating developing EMS systems [5, 6]. Evaluation of an EMS system, as of any large organizational structure, requires descriptors for important system variables and processes. When possible, such descriptors should be quantitative so that advanced technological tools such as statistical methods, systems analysis, and computer modeling and simulation may be used. Unfortunately, many of the components in an EMS system are highly complex, requiring a large number of variables for complete description. It has been suggested that this fact may have discouraged investigators from applying quantitative methods to EMS planning and evaluation [7]. Thus there is a need to isolate the most critical processes and variables of an EMS system and develop descriptors for those that require a minimum of variables. Parsimony in the describing variables is desirable This work was supported by a grant for computer time by the University of Illinois Circle Campus Computer Center and by a contract from the Department of Public Health, State of Illinois. Address communications and requests for reprints to John L. Semmlow, Ph.D., Bioengineering Program, University of Illinois at Chicago Circle, Box 4348, Chicago, IL 60680.
Spring 1976
45
Semmlow
&
Cone
to reduce the amount of data that must be collected to evaluate system performance. Probably the most important descriptors in any system are those that define the inputs and outputs of the system. In an EMS system the input variables should completely define the patient's health status upon entering the system and should include such factors as mechanism of injury, injury profiles, severity measures, and epidemiologic data, along with other indexes of general function. Similarly, the output variables should completely define the patient's health status when he finally leaves the system (not necessarily by hospital discharge) and should include factors describing mortality, morbidity, degree of disability, and the like [5]. Outcome measures might also be extended to include variables relating to discomfort or inconvenience such as days of hospitalization and days of disability. A complete description of EMS system inputs and outputs is clearly not feasible because, even if all t-he required variables could be uniquely defined and measured, the cost of collecting the necessary data would be prohibitive. The number of describing variables must be reduced to only the most informative ones. With regard to outcome measures there is little agreement on what constitutes a minimally acceptable set of defining variables [5], and a number of quantitative evaluation studies have used mortality alone [8-10] in lieu of a more comprehensive outcome descriptor. Of the many input variables, the most important is likely to be a measure of injury severity that correlates with the life-threatening potential of the injury. The focus of this study is one such injury descriptor, the Injury Severity Score (ISS) developed by Baker et al. [11]. This measure, derived from the Abbreviated Injury Scale (AIS) for injuries sustained in motor vehicle accidents [12], provides a numerical description of the overall severity of injury in persons who have injury to more than one area of the body. In their study, Baker et al. [11] showed a nonlinear correlation between the ISS value and the percentage of fatalities for the patient group they studied. In the present study, data from an independent source, the Illinois Trauma Registry [13, 14], are used to confirm the relation between the ISS input descriptor and the outcome variable, mortality. The area of relevance of the ISS is shown to include trauma sustained through nonvehicular incidents. Finally, the relationship between this injury descriptor variable and two other important intermediate variables is examined: days of hospitalization (which is both an outcome measure and an indicator of the amount of service required by EMS users) and the percentage of major surgery performed, another indicator of demand load. An adequate injury severity descriptor such as the ISS coupled with only a few additional input variables (such as patient age and treatment delay) should allow comparison of outcomes and other system performance measures across a broad spectrum of patient injury states by permitting adjustment of these variations in input case mix. This would greatly increase the amount of data applicable to any given study. Since data relevant to EMS are always
46
Research Health Services Health Services Research
INJURY SEVERITY SCORE
likely to be scarce, the development of concise descriptors and data-efficient computational techniques is essential for the successful application of powerful quantitative methods to EMS planning and evaluation. THE DATA
The ongoing Illinois Trauma Program [1] includes continuous monitoring by a specially designed information system, the Illinois Trauma Registry [13, 14]. This unique data resource is invaluable for quantitative research in EMS. The central registry employs a computerized data base and a generalized information retrieval system to collect and maintain data on all aspects of the trauma incident; it allows for identification of problem areas within the developing EMS system and provides a guideline for their solution [2]. The Illinois Trauma Registry classifies injury profiles into categories given in the 1968 edition of the International Classification of Diseases, Adapted (ICDA) [15]. It is possible to convert many of these values to equivalent modified Abbreviated Injury Scale (AIS) values [12] through the use of conversion tables generously supplied to us by Baker [16]. (These conversion tables are available from the authors.) The modified AIS values are then categorized by body area, and the ISS is computed by taking the sum of the squares of the three highest AIS values occurring in different body areas. The data used in this study were obtained directly from the Illinois Trauma Registry. The required variables were transferred via magnetic tape, and computations were implemented on an IBM 370-158 computer with plotting facilities at the University of Illinois at Chicago Circle Computer Center. The data cover two years of the initial operation of the Illinois EMS system (July 1971-June 1973). Two different geographical regions of the state are included, Regions 3A and 3B, covering central Illinois and including both rural and moderately urban areas. (Data from these areas were selected since previous experience [10, 17] has shown them to be the most reliable and Injury Severity Score Distribution of Patients Injured in Vehicular and Nonvehicular Incidents
ISS 0-79 .......... 0-9 .......... 10-19 .......... 20-29 .......... 30-39 .......... 40-49 .......... 50-59 60-69 .......... 70-79 .......... ..........
Spring 1976
Incident type Vehicular No. of
Both No. of
patients
patients 8 852 3 911 4 076 643 160 48 9 3 2
99.99 44.18
46.05 7.26 1.81 0.54 0.10 0.03 0.02
3 350 1 238 1 476 460 124 40 7 3 2
Nonvehicular No. of
patients 100.00 36.96 44.06 13.73 3.70 1.19 0.21 0.09 0.06
5 502 2 673 2 600 183 36 8 2 0 0
100.01 48.58 47.26 3.33 0.65 0.15 0.04 0 0
47
Semmlow & Cone 100
3.0
*
El
80
0~~~~~~~~~~~~8 El
2.5
z 2.0,
0
a
~2.0
*
Independent results obtained from Illinois Trauma Registry data confirm an earlier finding of a fixed, monotonic relationship between expected mortality and the Injury Severity Score (ISS) value of injuries sustained in both vehicular and nonvehicular incidents. Further results show the ISS measure to be a good indicator of hospitalization demand from accident victims and suggest that this variable can be used to control for variations in injury severity in comparative evaluations of EMS systems. The unique and immediate needs of the critically injured patient have proven a strong stimulus toward the development of comprehensive systems for the delivery of emergency medical services (EMS) [1-3]. Effective management of the necessary specialized facilities and personnel is essential if such care is to be available to a broad spectrum of the population; in fact, it is only through better utilization of presently available resources that any immediate improvements can be effected in a poorly-organized emergency health system [4]. Recognition of this situation has stimulated widespread interest in EMS planning, and with it the demand to develop effective methods for evaluating developing EMS systems [5, 6]. Evaluation of an EMS system, as of any large organizational structure, requires descriptors for important system variables and processes. When possible, such descriptors should be quantitative so that advanced technological tools such as statistical methods, systems analysis, and computer modeling and simulation may be used. Unfortunately, many of the components in an EMS system are highly complex, requiring a large number of variables for complete description. It has been suggested that this fact may have discouraged investigators from applying quantitative methods to EMS planning and evaluation [7]. Thus there is a need to isolate the most critical processes and variables of an EMS system and develop descriptors for those that require a minimum of variables. Parsimony in the describing variables is desirable This work was supported by a grant for computer time by the University of Illinois Circle Campus Computer Center and by a contract from the Department of Public Health, State of Illinois. Address communications and requests for reprints to John L. Semmlow, Ph.D., Bioengineering Program, University of Illinois at Chicago Circle, Box 4348, Chicago, IL 60680.
Spring 1976
45
Semmlow
&
Cone
to reduce the amount of data that must be collected to evaluate system performance. Probably the most important descriptors in any system are those that define the inputs and outputs of the system. In an EMS system the input variables should completely define the patient's health status upon entering the system and should include such factors as mechanism of injury, injury profiles, severity measures, and epidemiologic data, along with other indexes of general function. Similarly, the output variables should completely define the patient's health status when he finally leaves the system (not necessarily by hospital discharge) and should include factors describing mortality, morbidity, degree of disability, and the like [5]. Outcome measures might also be extended to include variables relating to discomfort or inconvenience such as days of hospitalization and days of disability. A complete description of EMS system inputs and outputs is clearly not feasible because, even if all t-he required variables could be uniquely defined and measured, the cost of collecting the necessary data would be prohibitive. The number of describing variables must be reduced to only the most informative ones. With regard to outcome measures there is little agreement on what constitutes a minimally acceptable set of defining variables [5], and a number of quantitative evaluation studies have used mortality alone [8-10] in lieu of a more comprehensive outcome descriptor. Of the many input variables, the most important is likely to be a measure of injury severity that correlates with the life-threatening potential of the injury. The focus of this study is one such injury descriptor, the Injury Severity Score (ISS) developed by Baker et al. [11]. This measure, derived from the Abbreviated Injury Scale (AIS) for injuries sustained in motor vehicle accidents [12], provides a numerical description of the overall severity of injury in persons who have injury to more than one area of the body. In their study, Baker et al. [11] showed a nonlinear correlation between the ISS value and the percentage of fatalities for the patient group they studied. In the present study, data from an independent source, the Illinois Trauma Registry [13, 14], are used to confirm the relation between the ISS input descriptor and the outcome variable, mortality. The area of relevance of the ISS is shown to include trauma sustained through nonvehicular incidents. Finally, the relationship between this injury descriptor variable and two other important intermediate variables is examined: days of hospitalization (which is both an outcome measure and an indicator of the amount of service required by EMS users) and the percentage of major surgery performed, another indicator of demand load. An adequate injury severity descriptor such as the ISS coupled with only a few additional input variables (such as patient age and treatment delay) should allow comparison of outcomes and other system performance measures across a broad spectrum of patient injury states by permitting adjustment of these variations in input case mix. This would greatly increase the amount of data applicable to any given study. Since data relevant to EMS are always
46
Research Health Services Health Services Research
INJURY SEVERITY SCORE
likely to be scarce, the development of concise descriptors and data-efficient computational techniques is essential for the successful application of powerful quantitative methods to EMS planning and evaluation. THE DATA
The ongoing Illinois Trauma Program [1] includes continuous monitoring by a specially designed information system, the Illinois Trauma Registry [13, 14]. This unique data resource is invaluable for quantitative research in EMS. The central registry employs a computerized data base and a generalized information retrieval system to collect and maintain data on all aspects of the trauma incident; it allows for identification of problem areas within the developing EMS system and provides a guideline for their solution [2]. The Illinois Trauma Registry classifies injury profiles into categories given in the 1968 edition of the International Classification of Diseases, Adapted (ICDA) [15]. It is possible to convert many of these values to equivalent modified Abbreviated Injury Scale (AIS) values [12] through the use of conversion tables generously supplied to us by Baker [16]. (These conversion tables are available from the authors.) The modified AIS values are then categorized by body area, and the ISS is computed by taking the sum of the squares of the three highest AIS values occurring in different body areas. The data used in this study were obtained directly from the Illinois Trauma Registry. The required variables were transferred via magnetic tape, and computations were implemented on an IBM 370-158 computer with plotting facilities at the University of Illinois at Chicago Circle Computer Center. The data cover two years of the initial operation of the Illinois EMS system (July 1971-June 1973). Two different geographical regions of the state are included, Regions 3A and 3B, covering central Illinois and including both rural and moderately urban areas. (Data from these areas were selected since previous experience [10, 17] has shown them to be the most reliable and Injury Severity Score Distribution of Patients Injured in Vehicular and Nonvehicular Incidents
ISS 0-79 .......... 0-9 .......... 10-19 .......... 20-29 .......... 30-39 .......... 40-49 .......... 50-59 60-69 .......... 70-79 .......... ..........
Spring 1976
Incident type Vehicular No. of
Both No. of
patients
patients 8 852 3 911 4 076 643 160 48 9 3 2
99.99 44.18
46.05 7.26 1.81 0.54 0.10 0.03 0.02
3 350 1 238 1 476 460 124 40 7 3 2
Nonvehicular No. of
patients 100.00 36.96 44.06 13.73 3.70 1.19 0.21 0.09 0.06
5 502 2 673 2 600 183 36 8 2 0 0
100.01 48.58 47.26 3.33 0.65 0.15 0.04 0 0
47
Semmlow & Cone 100
3.0
*
El
80
0~~~~~~~~~~~~8 El
2.5
z 2.0,
0
a
~2.0
*
