Emergency Medical Service Rescue Time in Fatal Road Accidents

Transcription

1 TRANSPORTATION RESEARCH RECORD Emergeny Medial Servie Resue Time in Fatal Road Aidents HAROLD BRODSKY Ambulane resue times in fatal road aidents in the United States have improved little over the past few years. In rural areas, it still frequently takes a half-hour, or more, for an ambulane to arrive after a rash. On the average, this happens in about one out of every five fatal aidents. Ambulane resue time onsists of two almost equal omponents: response (or travel) time and ommuniation time. A derease in response time is unlikely in the future, but more ould be done to redue the time it takes to all for an ambulane. In partiular, interageny delays in notifiation ould be eliminated. A mathed data set from Missouri shows that in 10 to 20 perent of fatal aidents the polie delayed 5 min or more in notifying an ambulane.dispather. Delays of this nature our beause a aller may fail to report injuries in the road aident. Perhaps an ambulane should be sent out anyway even if it is not ertain that injuries are involved. In 1988, the average ambulane response time in fatal road aidents in the United States was about 6 min in urban areas and about 11 min in rural areas. This level of aessibility is a produt of more than two deades of effort on the part of federal, state, and loal authorities. The goal of the Federal Emergeny Medial Servies At of 1973, to blanket the entire nation with ambulane servies, has largely been realized (J). The question that needs to be asked now is whether a plateau has been reahed or whether there are still opportunities for further progress. If ambulane speeds average about 1.5 min/mi in urban areas (40 mph) then a response time of 6 min an be equated to an average distane of about 4 mi from ambulane station to aident sene. If mile-a-minute ambulane speeds are usual in rural areas (60 mph) then the average rural ambulane station in the United States was within 11 mi of a rural road aident. Some areas of the United States are below aeptable standards of emergeny medial servie (EMS) aessibility, and no doubt with better training some servies may be able to redue their response times. Nevertheless, the prospet for a major improvement in ambulane response time in the years ahead seems unlikely. Indeed, the U.S. Department of Transportation 1988 annual Fatal Aident Reporting System (FARS) indiated that neither urban nor rural ambulane response times had shown a signifiant hange sine But is response time the only, or even the best, statisti to use to evaluate the aessibility of EMS in road aidents? Response time is defined as the number of minutes between EMS notifiation of an aident and EMS arrival at the sene Department of Geography, University of Maryland, College Park, Md of the rash. From the point of view of a publi health administrator, response times provide useful statistis for loating or reloating ambulane servies. Response times also provide performane standards that may be useful in judging the effiieny of a servie station relative to others. But from the point of view of the injured, response time is only one omponent of total resue time. What really matters most to the injured is the length of time it takes an ambulane to arrive after a rash. In 1988, for the entire United States, resue time in fatal road aidents in urban areas averaged about 12 min and in rural areas about 22 min. Resue time, whih inludes ommuniation delays, will always be longer than over-the-road travel time (or response time). Consequently, figures for resue time will always have a more sobering appearane. Given that a seriously injured person an go into an irremediable state of shok in 15 to 20 min, then the average rural resue time of 22 min is still not fast enough. Polie and emergeny medial tehniians do observe ases where the injured die during the resue proess. Time is a fator in survival, or in degree of reovery, otherwise there would be no sense of emergeny when responding to a road aident. The mathematial differene between resue and response time is ommuniation time. Communiation time is the duration from the time of an aident to the time when the EMS dispather was first alerted. Communiation time is often negleted beause it does not fall within the responsibility of any health professional. Medial professionals are trained to deal with the injured after they arrive. Emergeny 911 operators, polie ommuniation offiers, and EMS dispathers are onerned about delays in ommuniation, but little has been published about the problems they enounter in ommuniation. State funding has largely been devoted to purhasing state-of-the-art eletroni equipment, but the human element involved in sending unambiguous messages and in making diffiult deisions under unertainty has largely been left to the ommon sense of the bystanders who first all in the aident and to individual operators who must make sense out of what is frequently a garbled aident report. Every minute saved in dispathing an ambulane is equivalent to a minute saved in response time. But of the two, ommuniation time probably stands a better hane of improvement beause it is not only a matter of tehnology and eonomis, but also a matter of proedural effiieny. Existing proedures are so varied that it is hard to imagine that muh thought has been given to determining what works best. Complexity is taken for granted and therefore ommuniation time is ignored. Consequently, national or state-

2 90 wide guidane for dealing with ertain types of troubling but reurring situations are largely nonexistent. But how important is ommuniation time as a omponent of resue time? To answer this question, statistis are helpful, but reliane on overall averages may be misleading beause ommuniation varies with loation. The key to understanding why ommuniation delays our is knowing where they our. Statistis show that between 1983 and 1988, hanges in resue time have been negligible for the United States as a whole. Any trends within states are diffiult to verify beause of insuffiient data. An analysis of merged polie and EMS data from the state of Missouri shows that ommuniation time atually onsists of two separable omponents: all time and injury verifiation time. The distintion between all and injury verifiation times is more than perfuntory. Different strategies will be needed to improve eah of these separable aspets of ommuniation time. STATUS OF EMS DATA IN FARS The U.S. Department of Transportation has olleted data on EMS notifiation and arrival times in fatal aidents sine 1975 (FARS tapes). At first, only a handful of states were in a position to supply suh data, but by 1988 the majority of states were able to submit fairly omplete reords. Seventeen states had reasonably omplete reords going bak to However, ertain states apparently do not have legal requirements for ambulane distrits to supply trip information. California, New Jersey, Virginia, Massahusetts, and Washington are among the states that have a large proportion of missing data, at least as of Ambulane servies have traditionally been funded loally and a fiere independene often exists between loal and state offies. Even where a state an gain the voluntary ooperation of a loal ambulane servie, a math has to be made between the reords of two entirely different agenies: polie and ambulane servie. A orret mathing of reords an be diffiult and expensive. In some FARS data polie notifiation and arrival time have been found to be erroneously substituted for EMS notifiation and arrival time. For ertain analyses, these FARS reords are more than useless, they are absolutely misleading. Why then is there a relutane to publish figures on resue times along with other EMS road aident statistis? Data are available from the FARS tapes for the time of the aident and for the time of arrival, only a matter of subtrating the time of arrival from the time of the aident to obtain the resue time is Involved. However, the time of aident is estimated by the polie, usually by asking witnesses when the rash ourred, or by estimating the time of the rash on the basis of the first inoming all. Figures are often rounded, a sure sign that they are estimates (2). Beause the time of the aident is estimated it may seem less reliable than notifiation or arrival times, whih are based on atual observation. On the other hand, polie have no reason to bias their estimates of aident time beause their performane, like that of EMS, is based on response time, not on ommuniation time. Therefore, aident time averages are likely to be meaningful. But even notifiation and TRANSPORTATION RESEARCH RECORD 1270 arrival times are subjet to random errors beause of misreading loks, fast or slow loks, mismathes of ambulane and polie reords, and lerial errors. In 1988, there were 41,601 fatal aidents in the United States with about 58 perent in rural areas and 42 perent in urban areas. Of the 24,025 fatal aidents in rural areas, 27.9 perent were laking data that would enable a resue time to be alulated. In urban areas the situation was worse. Of the 17,576 urban fatal aidents, resue time ould not be alulated in 44.8 perent of the ases beause these urban figures are strongly influened by California, whih alone aounts for almost 40 perent of the missing urban data. Does this missing data introdue biases in the national averages? It all depends on how the data are used. If ambulane servie is no better, or worse, in California than in the rest of the United States, then this missing data may have little effet on overall national averages. A small proportion of the remaining data were not used in this analysis beause EMS notifiation time was given as earlier than aident time. Polie may simply have underestimated aident time. On the other hand, suh data may also be aused by lerial errors. Also, in a relatively small number of ases resue time took more than 2 hr after the aident. Without a doubt, ases of this nature are real beause a ar an rash into a dith late at night and not be disovered until the morning. But these data represent unusual irumstanes and perhaps should be analyzed separately. In any event, inluding suh data might strongly skew summary statistis. Aordingly, all negative resue times and times greater than 2 hr were left out of this analysis. COMMUNICATION TIME AS A PROPORTION OF RESCUE TIME In urban areas, an ambulane an be expeted to arrive quikly (in 5 min or less) at the sene of a rash in about one-third of the fatal aidents. In rural areas, this arrival time ours in only about 10 perent of the ases. From another point of view, in about 1 out of 20 fatal aidents (5 perent) in urban areas an ambulane may take an unonsionable half-hour, or more, to arrive at the rash site. In rural areas, even this dismal reord will be exeeded in one out of five fatal aidents (Figure 1). However, for rapid resues of 5 min or less, ommuniation may be almost instantaneous. In only 25 perent of the fatal aidents where resue was within 5 min did the ommuniations take more than 1 min. But ommuniation delays beome a progressively greater problem as resue time inreases. In resue times of greater than a half-hour, in half of the ases at least 21 min was required to ommuniate a need for an ambulane (Figure 2). Delays in ommuniating the need for an ambulane and delays in the length of time it takes to get out to the rural sene of an aident generally work together to exaerbate a problem that is virtually ertain to result in a large number of fatalities among injured people. Behind these dry statistis is a sense of frustration. An ambulane rew alled to the sene of a distant rural aident may rush to try to save a life. What the statistis above indiate, however, is that muh of the delay in the arrival of the ambulane may be due to initial diffiulties in ommuniation.

3 Brodsky 91 RURAL URBAN RURAL Communiation Time 50% 31 75% 18 23% 52 Response Time 100'Yo of fatal aidents 120., T J. Q) "' Q) E I= 30.2 r::'!i Categories.?. 1. :g.: : 'E 15.. E '. Upper Range 3rd quantile Median First quantile Lower range URBAN RURAL 9o/ 35 57% 37 URBAN I 33% 46 Elapsed time: - Resue Time - o to 5 6 to to or more mnutes minutes minutes minutes FIGURE 1 EMS ommuniation, response, and resue time for fatal aidents, (Soure: FARS tape, 1988, based on 16,561 rural and 9,661 urban fatal aidents.) COMMUNICATION TIME AND ENVIRONMENT Communiation delays vary with the environment. Low travel densities and low land use densities next to the road will result in fewer passersby when an aident ours. For those observing the aident, a low density may make it diffiult to find a telephone to notify the authorities. The density effet an be studied by seleting relevant variables: first, the United States an be divided roughly into two disrete density regions: (a) lower density for the mountain and plains states, and (b) higher density for the remaining states. Seond, a ommuniation delay an be expeted to be more likely during the late hours of the day (between 11 p.m. and 6 a.m.), than during other hours. Finally, the type of road might be examined beause the more housing adjaent to a road the more likely ommuniation will be quiker. This situation suggests that limited aess highways (Interstates) would be more likely to have longer ommuniation delays than other roads. When data from the 1988 FARS tape are examined with eah of these fators (region, hour, type of road) in mind, it appears that rapid ommuniation (of 5 min or less) is less likely in mountain and plains states than in the rest of the United States, during late hours than during normal hours, 0-5 min 6-14 min min Resue Time >30 min FIGURE 2 Communiation as omponent of resue lime for rural fatal aidents, (Soure: FARS tape, 1988, based on 15,334 rural fatal aidents.) and on Interstates than on other roads. Overall, rapid ommuniation is also less likely in rural than in urban areas (Figure 3). No surprises are apparent in these statistis. Indeed, if these relationships did not hold one would be inlined to suspet the auray of the data. Density relates to aessibility to a telephone and being lose to a telephone seems to be the major fator in EMS ommuniation delay. EMS RESCUE TIME AND MULTIFATALITIES In aident analysis one would like to know how many lives might be saved if ertain ations were undertaken. But unless one has a perfetly mathed sample it will be diffiult to obtain ompletely onvining results. In any nonexperimental analysis a possibility will always exist that one fator or another may have been left out of onsideration. All persons in a vehile involved in a rash are at risk of beoming a fatality. But in some aidents only one fatality will our, whereas in others there will be multifatalities. In part, the age and health of the individuals involved in a rash will affet the probability of a multifatality aident ourring. In part, the probability will also depend on the nature of the aident. Certainly, the probability of more than one person dying in a rash will also depend on the number of persons involved in the rash. If there are four people in a vehile when it rashes, the probability of more than one person dying, all other things being equal, will be greater than if only two persons were involved. However, the probability of a multifatality will also depend on how fast EMS arrives at the sene to render aid. In order to verify these assumptions, 1988 FARS data for the United States were used onsisting of an initial sample size of 9,381 fatal aidents in whih there were exatly two persons involved in the aident. Overall, in only 5.5 perent of these two-person aidents did multifatalities (two fatalities

4 92 TRANSPORTATION RESEARCH RECORD 1270 BJLB.d.L l:ll No<mal Hou" Lalo Hou"s URBAN Mountain and Plains Region ! - 'lo'lf. 90. xi -s 80. :i 1. :i 70. j - eo"' eo. u. :. so. so. : ; 40. Rest of the USA 30. '30.!:!: >C. a. a. 10% 10.,. o.. Interstate US & State Roads Interstate US & State Roads I] Normal Hou"s Late Hou'I Mountain and Plains Region , 90% 90%.!! 80.C.., ::> 80..!! 70.C. ::> % so. " u. 5 " > 50. :. u: j 40.C. 40. :. 30% l 30!: 20% a. 10%,. a. o o. Rest ol the USA Interstate US & State Roads Streets Interstate US & State Roads Streets FIGURE 3 Perentage of rural and urban fatal aidents with a ommuniation time within S min, by region, hour, and type of road, in this ase) our. When broken down into groups of resue time ategories, the perent of multifatalities varied. Multifatalities ourred in 3.9 perent of ases where EMS arrived within 5 min. But the rate inreased to 8.3 perent in ases where resue time for EMS was 30 min or more. The same EMS resue time relationship was also examined for involvement ategories onsisting of three-, four-, and five-person aidents with similar results. In all ases, as one would expet, multifatalities inreased with involvement ategory. But more important, multifatalities also inreased with the length of time it took EMS to make the resue. However, these results were based on aidents at all speeds. Speed also affets the probability of multifatalities. Further, slower speeds may be assoiated with loations where resue time is faster. Therefore, it seemed reasonable to sample only those aidents where posted speed was ;:::: 55 mph. When this proedure was followed, the sample size for the two-person involvement ategory dropped to 4,676 aidents. The overall perent of multifatalities inreased from 5.5 to 6.5 perent (not a large inrease, but ertainly in the expeted diretion for inreased speeds). In fat, all involvement ategories (three-, four-, and five-person) showed an inrease in multifatalities. Clearly, probability of a multifatality in a fatal aident inreased with posted speed. However, ontrolling for speed did not affet the general relationship of an inrease in multifatalities with an inrease in EMS resue time. Posted speed was not a onfounding variable, although it ould have been. Next, pedestrian aidents were removed from the sample. For two-person involvements the sample size dropped to 3,893. Here, the underlying assumption was that although two people may be involved in a pedestrian aident, only the pedestrian will atually be at risk. Removing pedestrian aidents from the sample atually improved the relationship between speed of EMS resue and perent multifatalities. (Removing pedestrian aidents would have had an opposite effet if they ourred more frequently in remote loations.) The variation now beame 6.1 perent for resue times of 5 min or less, and 9.9 perent for resue times of 30 min or more. But the presene of pedestrian aidents in the original sample did not alter the general relationship between resue time and multifatalities. The type of aident might possibly affet the assoiation beause the risk of having more than one person die may vary depending on whether the rash was single or multiple vehile. Consequently, the sample was redued to multiple-vehile rashes. Sample size now dropped to 1, 913 aidents for twoperson involvements. Nevertheless, the relationship between speed of resue, number of persons involved, and multifatalities still remained (Figure 4). The validity of this assoiation was heked for statistial signifiane with logit analysis. Both of the independent variables, involvement and EMS resue time, were found to be signifiant at the 1 perent level. Additional ontrol variables suh as age of persons involved and preise nature of the aident ould be used to further refine the proess. However, eah refinement redues the sample size and introdues the possibility that a valid relationship may be obsured by random variation.

5 Brodsky The results as they now stand are onsistent with the general understanding that speedy resues save lives. The results may atually underestimate this effet quantitatively beause some of the individuals who died in single-fatality aidents may also have been affeted by delays in EMS arrival. As in many studies of risk, no one demonstration will be suffiient. A ombination of studies using different methods of standardization of risk will ertainly make a stronger ase for a ausal relationship. In a study done in Texas, for example, the effet of EMS aessibility on fatalities was measured using a severity ratio, rather than multifatalities, as a means of standardization (3). But a logial relationship between EMS resue time and survival is not the issue beause there an be little doubt about the emergeny of injury road aidents. The major purpose of a statistial analysis is to provide a better quantitative assessment of the numbers of fatalities affeted by variations in EMS resue time. As the quality of the data improves, one an expet to see progressively more aurate assessments. TRENDS IN RESCUE TIME, 1983 to 1988 The quantity and ompleteness of EMS data on fatal aidents in FARS has improved sine For example, in 1983 only 53.3 perent of rural resue times in fatal aidents ould be alulated. By 1988, 72.1 perent of rural fatal aidents had EMS data assoiated with it; only 27.9 perent were missing. But preisely beause the proportion of missing data has hanged, espeially within ertain states, it is not always possible to assume a trend. Improvement, or lak of it, may simply be aused by the effet of having more omplete data. Mapping state-by-state hanges between 1983 and 1988 indiates that 20 states improved their resue time, at least in rural areas, with respet to the perent of fatal aidents with a resue time of 30 min or more. But 11 states showed an inrease in perent of EMS resues in rural areas that were 30 min or more (Figure 5). No pattern is apparent in the state-by-state omparisons, and overall national statistis do not indiate signifiant improvements. Delays in ommuniation time atually inreased from 5.7 to 6.3 perent in fatal aidents with 30 min or more needed for ommuniation. Response time delined slightly from 3.6 to 3.2 perent in ases of 30 min or more needed for response. Resue time remained about the same, or slightly delined, from 7.4 to 7.2 perent when 30 min or more are needed for resue. The stati nature of ommuniation time over the past 5 years, like the stati nature of response time, might indiate that little an be done to improve resue time, or it might indiate that the ommuniation aspet of EMS has simply been negleted. CLARIFYING COMMUNICATION TIME WITH MATCHED MISSOURI DATA What is not lear from statistis based on FARS data alone is the relationship between polie and EMS notifiation of a % minutes minutes minutes EMS Resue Time Note: Fatal aidents on roads wtth 55 mph. or greater posted speeds and multiple vehile aidents only. FIGURE 4 Multiple fatality aidents as perentage of total fatal aidents by involvement and EMS resue time. (Soure: FARS tape, 1988.)

6 94 TRANSPORTATION RESEARCH RECORD 1270,o. _ Legend ::J 20 States dereased - 11 States inreased * 19 States had insuffiient reords for romparison FIGURE 5 Change in perentage of resues of 30 min or more in fatal aidents, (Soure: FARS tapes, 1983 and 1988.) road aident. Who gets notified first, the polie or EMS? Or do they both get notified at the same time? Are there any delays in notifiation between agenies? How muh of ommuniation time is from delay by a passerby in getting to a telephone, and how muh is from interageny delays in transmission? In order to answer these questions it is neessary to examine both EMS and polie notifiation times. FARS data, unfortunately, omit polie notifiation time, although it ould have been obtained easily from the state polie aident reord. The need for mathing different data soures to produe a riher file of data is beoming more apparent in aident researh. Daniel Fife (4) desribes mathing of FARS data with a file from the National Center for Health Statistis (NCHS). Using suh riteria as age, sex, and date of death, he was able to uniquely m;ith 85 perent of the FARS data with NCHS data to produe a data set that an examine the nature of the injury with aspets of the motor vehile rash. Similar researh in mathing data files urrently underway by Sandra W. Johnson in Maine is sponsored by the National Highway Traffi Safety Administration. In the Maine study, ambulane run reports are being linked with polie, hospital, and other data files to produe a sensitivity index for statewide systems that will evaluate the sequene of events from time of rash to hospital release of the injured. In this study, EMS data were taken from the FARS tapes for 1985 to 1988 and mathed with data from Missouri polie injury aidents. The FARS data inlude the time at whih EMS was alled as well as loation date and time of aident. Loation and time of aident make it possible to math the FARS file with the polie file relatively easily. For example, in the 1987 fatal aident data it was possible to math 918 out of 927 fatal aidents. The unmathed nine either had missing or inorretly oded reords. Among the 918 sueessful mathes, some required a small hange in the oding. For example, a set of six observations was inorretly oded in FARS by the ounty, whih was obvious beause no ounties in Missouri have that digit. Usually, omputer onsisteny heks made with the FARS data are omprehensive, but this partiular data set was not heked for valid ounty odes. All times were onverted to minutes from the beginning of the day to enable simple subtrations to be made. Beause there is only one date on the reords, a day was added to the subtration when the resue times went past midnight. Adjustments were also made when it was lear that military time was not used. Data from the fatal aidents demonstrate that most frequently both polie and EMS are notified at about the same time. In urban areas, polie and EMS are notified within 1 min of eah other in 41.7 perent of the fotol oidnts. In rural areas, the figure is less, only 27.0 perent of fatal aidents. Most rural areas in Missouri do not have 911 emergeny numbers, so the aller has to all either the polie or EMS. However, what is most disturbing is that in urban areas in 14.7 perent of the fatal aidents EMS was notified 5 min or more after the polie had been notified. In rural areas, the perentage inreased to perent of the fatal aidents (Figure 6). Polie are apparently not always notified first. In urban areas in 15.1 perent of the fatal aidents, EMS was apparently notified more than 5 min before the polie. In rural areas, the figure was apparently 19.3 perent. Apparent is used, but the figures for EMS notifiation before the polie are ambiguous. Polie notifiation indiates the time that the offier who filled out the aident report was notified, not neessarily the time that polie were first informed of the aident. If a polie offier is too oupied with the aident to be able to fill out an aident report another offier may

7 Brodsky 95 be alled on to do this job. The offier who arrives later will reord the time he or she was notified of the aident on the aident report. Consequently, some, many, or possibly all instanes where the polie were apparently notified after EMS beome louded. In some ases, EMS ould really have been notified before the polie but there is no way of knowing this for sure without aess to the original polie logbooks. But why the delay in polie notifying EMS? When the polie reeive a all they always ask about injuries. If the aller indiates that there are injuries, then the polie immediately radio the EMS dispather for an ambulane. However, if the aller is vague then the polie may hesitate until onfirmation that an injury is involved before they notify EMS. Consequently, in 15 perent or more of the fatal aidents there may be a delay of at least 5 min beause the polie are unsure about injuries. This omponent of ommuniation delay is well known to both polie and EMS ommuniations operators and there!!l 40% Cl> "C 8 < 50% % 0 20% Rural Sample size= 2,366 EMS "apparently". noified before. a; ;f;ed. polie 10%-t----t 50% 40%.!!l Cl> "C g 30 /o < LL l'i 20% 0 Cl> Cl> 10% < to -5 4 lo -2-1 to 1 2 to 4 5 to 1 O > 10 Urban Differene in Notifiation lime in Minutes Sample size= 927 EMS "apparently" notified before polie EMS notified i----- aher'iioiie. <-10-10to-5 4to-2-1to1 2to4 5to10 >10 Differene in Notifiation lime in Minutes FIGURE 6 Polie minus EMS notifiation time versus perent of fatal aidents, Missouri, are onsiderable differenes in opinion as to what, if anything, an be done about it. Beause the vast majority of reported aidents do not require an ambulane and beause fatal aidents are omparatively rare ourrenes even among aidents that do require an ambulane, there is a tendeny to pereive this problem as minor. Only when a statistial analysis is done for an entire state over a period of 1 year or more do the serious dimensions of this problem begin to emerge. Some EMS people are of the opinion that the EMS dispather should be notified of a road aident, regardless of whether the polie think that an injury is involved or not. This pratie would plae more of the responsibility on EMS for deiding whether or how to respond. EMS dispathers would then have to deide whether the desription of the rash warranted an ambulane resue. Perhaps EMS should be sent out more frequently to road aidents even when it is unertain that an injury is involved. The extra burden that this might plae on an ambulane servie ould best be understood by people within the servie and might be a fator in deision making. If a hange in poliy is initiated, it should be monitored over time to weigh inreases in suessful resues against probable inreases in dry runs (no transportation of injured persons). At present, no state has a program for evaluating the auses and onsequenes of delays in emergeny ommuniation in road aidents. CONCLUSION Although ambulane response time is the most widely used statisti in evaluating EMS, it is meaningful only when making performane omparisons between otherwise similar servies. A muh more important statisti is resue time. From the time of the rash until the time EMS arrives, the injured may suffer irreversible physiologial hanges affeting survival or omplete reovery. Therefore, the faster the ambulane arrives the more likely the individual will reover. The results presented are onsistent with an understanding that delays in EMS arrival do affet the number of fatalities. Resue time in road aidents has hanged little over the past several years. Response time will likely remain frozen at its present level unless additions are made in the number of ambulane stations, whih ould be ostly. Communiation time may be more promising to pursue for improvements in the aessibility of EMS in road aidents. Communiation time is a major omponent of resue time, about equal in importane to response time. Communiation delay ours in situations where it is diffiult to find a telephone. The effet of travel and population densities on ommuniation time is quite lear. However, new tehnologies involving use of satellites and ellular phones may in the future redue this problem (5). Enouragement should be given to the development of this kind of tehnology for its potential value in medial emergenies. But another aspet of ommuniation delay is quite independent of telephone availability. Apparently, not all inoming alls to the polie learly speify the need for an ambulane, and most ar rashes do not result in injuries. Consequently, the polie generally do not notify EMS about an aident unless they are sure that injuries are involved.

8 96 TRANSPORTATION RESEARCH RECORD 1270 Polie ommuniations offiers assume the primary responsibility for sorting out road aidents likely to need EMS from those that probably do not. But, in about 15 perent or more of fatal aidents, the ommuniations offier makes the wrong deision by failing to notify EMS immediately. Should this responsibility of sorting things out be shifted to the EMS dispather? And if EMS is allowed to make these deisions, what poliy should EMS use? Should EMS automatially respond to all reported aidents, or should EMS wait for onfirmation of atual need as the polie usually do? Or is there a middle ground that has yet to be explored? Polie and EMS are separate agenies that are often relutant to examine ontroversial boundary issues unless drawn to it by external pressures. The general publi is probably unaware that this problem even exists. But ommuniations offiers know that their judgment ould be questioned in ourt. Negligene has been brought up in other situations, but never in onnetion with EMS delays in road aidents. However, for good reasons both the polie and EMS operators reord all inoming alls. Statistial results from the Missouri data on differenes in ageny notifiation are dealt with in greater detail by Brodsky (6), but the kind of analysis presented here ould be repeated in most states. Missouri has a ommuniation system similar to many others in the United States. Therefore, it is likely that delays in EMS resue in road aidents beause of ommuniation problems will be found to be widespread in this ountry and perhaps in other nations as well. ACKNOWLEDGMENT This researh was supported by the AAA Foundation for Traffi Safety, Washington, D. C., Sam Yaksih, Diretor. The Missouri State Highway Patrol and the Missouri Department of Health provided some of the data used in this analysis. Lisa Wolfish, graduate student at the University of Maryland, helped in the preparation of the graphis. REFERENCES 1. Emergeny Medial Servies at Midpassage. A Report to the Committee on Emergeny Medial Servies, National Aademy of Sienes, Washington, D.C., S. P. Baker. Digit Preferene in Reported Time of Collision. Aident Analysis and Prevention, Vol. 3, 1973, pp H. Brodsky and S. Hakkert. Highway Fatal Aidents and Aessibility of Emergeny Medial Servies. Soial Siene and Mediine, Vol. 17, 1983, pp D. Fife. Mathing Fatal Aidents Reporting System Cases with National Center For Health Statistis Motor Vehile Deaths. Aident Analysis and Prevention, Vol. 21, No. 1, 1989, pp D. K. Willis. IVHS Tehnologies: Promising Palliatives or Popular Poppyok? Transportation Quarterly, Jan. 1990, pp H. Brodsky. Geographi Perspetives on Improving Emergeny Notifiation in Road Aidents. AAA Foundation for Traffi Safety, Washington, D.C., Publiation of this paper sponsored by Committee on Traffi Reords and Aident Analysis.