Is EMS Over or Under-Utilized in the South Bronx A Retrospective View

Tools



Abstract

In cities throughout the United States, access to health care is made possible largely through Emergency Medical Services. Unfortunately, the wide availability of services has led to claims of overuse and misuse of the transport system. Recently, the rising cost of healthcare has demanded a need to reduce waste.

Prior studies that investigated EMS usage patterns have used various criteria to assign “medical necessity”. We hypothesized that medically unnecessary non-urgent EMS transports were over utilized to St. Barnabas Hospital, Bronx, NY. Seven hundred and twenty patient encounters were examined through a retrospective chart review spanning one full calendar year from January 1 to December 31, 2002.

Our results reveal that 15% of the randomly selected patient encounters did not warrant emergent ambulance transport.  Although the outcome was less than our assumption and the figures reported in the review of the literature (The results have estimated that 31-92% of the ambulance transports to the hospitals were over utilized and determined “medically unnecessary” for emergent ambulance use), 1 2 3 4 5 6 7 adoption of alternate transport means could still result in significant savings.

Introduction

Health care costs in the United States have risen faster than the ability to pay for services.  Consequently, there has been an urgent need to reduce costs by streamlining services without sacrificing the quality of health care.  Emergency Medical Services (EMS) is one particular aspect of health care that has provided an invaluable service to the general public.  The health care safety net that the EMS system provides addresses and treats acute medical problems that would otherwise increase morbidity and mortality to the population.  However, the overuse of ambulance services and the inability to recoup costs has prompted cities to revamp their system by diverting funds to transport non-urgent medical problems by less expensive means.

Prior studies have analyzed the ambulance utilization pattern endemic to each area studied. 1 2 3 4  The results have estimated that 31-92% of the ambulance transports to the hospitals were over-utilized and determined “medically unnecessary” for emergent ambulance use. 1 2 3 4 5 6 7  The standards to determine “medical necessity” varied for each study, but despite the use of conservative guidelines the conclusions were similar.  The issue was not to determine whether there were “medically unnecessary” ambulance transports, but to calculate the magnitude of abuse.

The study looks to identify the ambulance utilization pattern to St. Barnabas Hospital, a Level 1 Regional Trauma Center in the Bronx, NY.  We analyzed the New York City 911 Triage Protocol to identify the lowest priority triage category deemed medical stable. By calculating the percentage of low urgency triage levels as “medically unnecessary” we are able to determine the percentage of 911 calls that can be diverted to lower cost transport services.  We designated a 911 call as low priority as long as there was no compromise in medical safety.  The percentage of overuse transports could be used to propose safer and more inexpensive alternative EMS transports.  These savings from alternative programs could have a significant financial impact to the borough of the Bronx in New York City while better utilizing ambulances.

Methods

Study Design

The study is a retrospective chart review spanning from January 1 through December 31, 2002.  The research sample was 720 patient encounters that were selected randomly from an electronic charting program for the 2002 calendar year. To reflect the natural variation in time of day, four patient transport visits per nursing triage shift were randomly selected and labeled 1, 2, 3 or 4 in order of their presentation to the Emergency department. There are three 8-hour shifts per day.  These shifts start from 12am to 8am, then 8:01 am to 4pm and lastly 4:01pm to 11:59pm. The sampling of data continued with the random selection of five days per month for the 2002 twelve calendar months.  Confidentiality was further ensured by keeping all information under lock and key in the Emergency Department Administrative Office with access only by the three investigators and Directors of the Emergency Department.

Setting and Population

The patient ambulance transports included in this population sample were from the St Barnabas EMS catchment’s area that phoned 911 in the 2002 calendar year.  We define our catchment’s area by using the EMS assignment of ambulance response to a catchment’s area, which are the Precincts (Battalions) the ambulances cover.  The St. Barnabas Hospital EMS units cover the 46 and 48 Precincts, an area approximately 4.5 square miles.  The catchment’s area is not a rigid boundary, but rather an area guideline for EMS response.  It expands and contracts due to the volume and acuity of calls each day.  This inherent and necessary variation in our measurement is a limitation of the study.

Measurement

The patient encounters were labeled anonymously using a four digit identification code representing month/day/shift (1,2,3)/patient order number (1,2,3,4) and the data recorded and compiled using Microsoft 2000® Excel spreadsheet software. The database included the identifier code (with four variables from ID code) and the NYC 911 triage code. The database also included: EMS diagnosis; SBH triage code; ED diagnosis, disposition from ED; and pre-hospital EM care.  The criteria for inclusion in this research study were patients who called 911 for transport, the determination of medically unnecessary was determined by the New York City 911 Triage system. Only those patient calls with the lowest priority (category 7-9) were being deemed medically unnecessary.  All triage diagnoses designated 1-6, as well as including those 7-9 with the diagnosis of “determination of DOA”, “Dead on Arrival”, “emotionally disturbed” and “jumper still up” will be deemed medically necessary.

Data Analysis

The minimum necessary sample size required to analyze data for statistical inference was calculated as minimal sample size of 572 required. A sample size of 720 was collected with 638 being final sample size after incomplete records were excluded from study.  The randomly selected and anonymously coded data from the emergency department database were compiled and the EMS transport count, medically unnecessary transport count and rate, and medically necessary transport count and rate, were calculated in an Excel spreadsheet using Excel software.

The counts from the sample were then graphically analyzed using Excel bar graphs. The rates of medically unnecessary transports between our sample and the benchmark data from Billittier et al 1996 study were graphically analyzed using Excel bar graph (Graph I).  Billittier et al, 1996 is our benchmark study for one salient reason aside from the other similarities in populations sampled. This salient and important consideration is the conservative operational definitions of medically unnecessary transports that were employed by both research teams.  The chi square test statistic was calculated from a contingency table of counts of unnecessary transports between the samples compared.

The data samples were then statistically analyzed utilizing a chi square (x²) test statistic with Minitab® statistical software version 13.31. An alpha level of 0.05 between samples was determined by the research team as statistically significant value.  The data was then further analyzed by variables of shift, hospital triage score and season.  Shift night is from 00:00 to 08:00 hours, shift day is from 08:01 to 16:00 hours and shift evening is from 16:01 to 23:59 hours.  Hospital triage scores are A (highest priority), B (second priority), C (third priority).  Seasons were grouped by months as follows: spring included March, April and May; summer included June, July and August; fall included September, October and November, and winter included December, January and February.

The data samples were statistically analyzed utilizing a chi square (x²) test statistic with Minitab® statistical software version 13.31. An alpha level of 0.05 was determined by research team as statistically significant value between samples for each variable measured.

Results

The count for unnecessary ambulance transports for St Barnabas sample was 94 of 638 opportunities (percentage of unnecessary transports= 15%).  The count for unnecessary ambulance transports for benchmark data from Billittier et al was 71 of 626 opportunities (percentage of unnecessary transports= 11 %). The difference in the unnecessary ambulance transports between the St Barnabas sample and the benchmark sample (Billittier, et al 1996) was not shown statistically significant with a p value of 0.074 for an alpha set at .05, and a chi square test statistic of 3.203 DF= 1 (Table I in Appendix).  The chi square test statistic calculated was = 3.203, and the p value was = 0.074 for the data analyses.  The data was then further analyzed by variables of: shift, hospital triage score and season.  The chi square test statistic calculated was not statistically significant for shift comparison of unnecessary medical runs (chi square statistic= 0.063, p value of 0.969).

The chi square test statistic calculated was statistically significant for triage category in ED (chi square statistic= 18.977, p value of &ln; 0.001).  The chi square test statistic calculated was also statistically significant for comparison of data by season (chi square statistic= 14.477 p value of 0.002) (Table I in Appendix).

Dsicussion

The result of 15% unnecessary ambulance transport rate was not surprising due to the conservative operational definition we used to classify an ambulance transport as unnecessary.  Compared to the previously published studies with reports of 31% to 92% medically unnecessary transports, the operational definition of medically unnecessary that we applied (the New York City 911 Triage protocol) was conservative.  If we address the percentage of “unnecessary transports” using our more moderate guidelines we may be able to still have a significant impact on cost while not negatively impacting quality of emergency patient care through cavalier cost containment strategies, and arbitrary construction of operational definitions.

The close proximity of our percentage of unnecessary transports to that of our selected benchmark edifies our hypothesis that we in fact do have a percentage of medically unnecessary transports.  Even when measuring these transports conservatively we still have concern from a resource application perspective.  There is no universal metric to use to decide what constitutes a medically unnecessary ambulance transport.  The criteria vary and therefore conclusions must be drawn with some reservation.  We chose to compare our sample data to the sample data from the Billittier et al study for reasons of similarities in populations sampled and the conservative operational definitions of medically unnecessary transports that were utilized by both research teams. The consideration of these factors prior to statistical comparison of the data gives practical significance to the results. The populations sampled in both studies were patients in New York State that had called for an ambulance.  Both studies do not limit to specific age or complaint type population.  Both studies include urban hospital ambulance transports as the majority of the sample data.  The criteria for medically unnecessary transports were based on conservative criteria in both our study and the Billittier study. This factor may contribute to both studies having lower medically unnecessary transports than other studies cited by both research teams.

The statistical significance by season of unnecessary medical EMS transports shows unnecessary transports in spring and winter seasons higher than expected, and unnecessary EMS transports in summer and fall lower than expected.  Further analysis of this phenomenon could reflect patterns related to weather, pollen counts, and seasonal disease factors etc.  The statistical significance in triage category by hospital personnel reflected an agreement between hospital personnel triage categories and field EMS triage category assignment.  This agreement of patient clinical assessment by pre hospital and hospital practitioners supports other studies and proposals that give EMS personnel more decision making options regarding transport of persons who call the 911 system.

The data did not prove statistically significant for any variability in EMS call among the three-shift designation.  This result is surprising, especially due to the large p value (p= 0.969) associated with the chi square statistic.  Medical cases, which are the majority of EMS transports, will occur more without time variation compared to trauma cases.  If you hypothesize that patients use EMS transport when it is inconvenient then there should be a spike in EMS use during those time periods (i.e. Night shifts).

Therefore, if the EMS transports are occurring when medical emergencies occur, then that should explain a more consistent nature of the transports.

There are limitations to the conclusions that can be drawn from our study. The research was conducted retrospectively.  This limits the interrogation of the data and design of data collection tools (no real time survey possible, no recording of possible confounding factors by researchers).  Our study researched ambulance transports over a one-year period, whereas our comparison study (although of same sample size and geographic area and patient population) was collected over one week.  The question as to the seasonal variation that may be a confounding variable in this type of study, or other possible cyclic phenomena that is captured over a twelve-month study is not measured in a one-week study. In addition, we define our catchment’s area by using the EMS assignment of ambulance response to a catchment’s area.  The catchment’s area is not a rigid boundary, but rather an area guideline for EMS response. The comparison study included data from four hospital’s catchments areas (one suburban and three urban).

The study looked to determine the EMS utilization pattern to St. Barnabas Hospital, Bronx, NY.  We presumed that the EMS utilization pattern to St. Barnabas Hospital was representative of the ambulance pattern in the Bronx borough of New York.  Our hypothesis was that the EMS utilization to a regional trauma center with a high indigent population would have a pattern exceeding the national average.  However, we calculated that the EMS utilization to St. Barnabas Hospital was 15%, which was significantly lower than the national average of 31%.  However, the NYC 911 Triage Protocol extensively delineates urgent vs. non-urgent medical cases compared to the criteria used to calculate the national average.  Although utilization of EMS transportation varies from region to region, the efficiency of the New York City EMS Triage protocol underestimates our claim that over utilization is a significant problem in this metropolitan region. A major flaw in this conclusion can be contributed to the fact that the medical necessity of ambulance transports does not directly correlate with the Emergency Physicians discharge diagnosis.  Further delineation of the lower triage categories (7-9) is needed to differentiate urgent vs. non-urgent cases.

In the study, we chose to use a conservative category level for deeming an EMS transport an appropriate medical necessary 911 emergency call.  Although an argument can be made to downgrade the designation level we felt that this action would only increase health risks to the general population.  Furthermore, only appropriate direct medical evaluation can truly determine a patient stable for non-medical transportation.  Alternative means of EMS transportation may still be advocated to address the remaining 15% overuse. The savings from a more efficient EMS system could be used for better equipment, more staff, training programs, and health education programs.

Appendix

Table 1

Summation values of data inferential statistical analysis
VARIABLE X² TEST STATISTIC DF PROBABILITY
by study (St Barnabas v benchmark) 3.203 1 0.074
by shift 0.063 2 0.969
by hospital triage code 18.977 2 <0.001*
by season 14.477 3 0.002*

* statistically significant at alpha (α) of 0.05

Graph 1

Image of graph

 

References

  1. Billittier AJ I, Moscati R, Janicke D, Lerner EB, Seymour J, Olsson D. A multisite survey of factors contributing to medically unnecessary ambulance transports. Academic Emergency Medicine. 1996;3:1046-52

  2. Morris DL, Cross AB. Is the emergency ambulance services abused? British Medical Journal. 1980;281:121-123

  3. Garner GJ. The use and abuse of the emergency ambulance service: some of the factors affecting the decision whether to call an emergency ambulance. Archives Emergency Medicine 1990;7:81-89

  4. Rademaker AW, Powell DG, Read JH. Inappropriate use and unmet need in paramedic and nonparamedic ambulance systems. Annals of Emergency Medicine 1987;16:553-556

  5. Brady WJ, Jr, Hennes H, Wolf A, Hall KN, Davis M. Pattern of basic life support ambulance use in an urban pediatric population. American Journal Emergency Medicine 1996;14:250-253

  6. Camasso-Richardson K, Wilde JA, Petrack EM. Medically unnecessary pediatric ambulance transports: a medical taxi service? Academic Emergency Medicine 1997;4:1137-1141

  7. Lammers RL, Roth BA, Utecht T. Comparison of ambulance dispatch protocols for nontraumatic abdominal pain. Annals Emergency Medicine 1995;26(5):579-589