Obstacles to emergency medical services system design and operational features

Study design

This was a qualitative embedded multi-case study (Scholz and Tietje, 2002; Yin, 2003). Five mini cases that represent different urban (>100 000 population) provider models (i.e. fire, third service, public utility model, private, hospital-based) in the US are presented. The cases relied on multiple sources of data including: i) data metrics, ii) document review, iii) interviews, and iv) archival records. Data was self-reported by executive leadership from each participating EMS system.

Population and setting

All participating provider organisations served urban centres with populations greater than 100 000 citizens. Design considerations and opportunities for the provision of ambulance service vary dramatically when discussing urban versus rural settings. The scale of geographic coverage, volume of requests for service and transport, available resources, and options for increasing efficiency and productivity are significantly different in an urban system (Stout, 1994). EMS systems were selected to represent examples of the five most common provider types found in urban centres in the US (Williams and Ragone, 2010). Executive leaders interviewed included the physician medical director and senior operational leaders (i.e. chief executive officer, deputy chief) from each location. In the public utility model system, the chief executive of the authority and the chief executive of the contracted, private provider were interviewed.

Human subjects review

The research protocol was reviewed and approved by the Saybrook University Institutional Review Board. Objectives of the research study and description of how the data would be used were presented to the participants verbally and in writing in advance and at the start of the interviews. Written permission from each participant was obtained in advance using a consent form.

Participants were advised of all data collection methods (e.g. recorded telephone line). Complete transcripts and any written interpretations were made available to the participant at his or her request. Best interests of the participant's rights and confidentiality took priority when considering the reporting of data. Participants had the power to dictate their level of anonymity (Miller, 1992).

Experimental protocol

Fifteen potential design and operational features of an EMS system were identified by the author through a literature search (see Table 1). The search included government reports, peer-reviewed journal research, dissertation research, and non-peer reviewed EMS publications.

A survey was distributed by email to an executive leader at each EMS system (http://bit.ly/1PJyK4A). The survey was administered using a cloud-based survey tool (SurveyMonkey.com, Palo Alto, CA, USA). Completed survey results were downloaded to an Excel file (Excel 12.0 for Macintosh, Microsoft, Redmond, WA, USA). Submitted data was reviewed for omissions and potential data entry errors. Data in question was returned to the submitter for completion or confirmation of accuracy.

Interviews were conducted by telephone using voice-over-IP (Skype, Luxembourg City, Luxembourg) and recorded to mp3 files with an add-on programme: Call Recorder (Ecamm. North Andover, MA, USA). Participants were informed in advance and at the start of the interview of the audio recording and all granted their consent. Completed interviews were transcribed. Participants were provided an electronic copy of their transcribed interview to confirm the accuracy of the transcription and to consider adding or modifying comments. No requests were made to alter the transcript after review.

Each transcript (n=11) was reviewed and obstacles described by participants were highlighted in the text using the Microsoft Word (Microsoft Word 12.0 for Macintosh, Microsoft, Redmond, WA, USA) comments feature. An outside reviewer independently coded a stratified sample of the clinical and operational leader (n=6; 54.5%) interview transcripts. The independent coder received an example of a coded transcript and a list of codes to use. Additional codes could be added at the reviewer's discretion. Inter-rater reliability was high with only one missed obstacle identified by the external reviewer.

Each obstacle was recorded in a Bento Database version 3.0.3 (Filemaker, Inc, Santa Clara, CA, USA). Each record entry cited the i) name of participant, ii) system model identifier (e.g. hospital-based), iii) participant position/role (e.g. clinical or operational), iv) transcript location by page number and line number, v) type of obstacle, vi) description of the obstacle, and vii) number identifying the obstacle (Dean, 2004).

The participants identified 582 obstacles to the 15 features. Obstacles were initially summarised into a few descriptive words (e.g. not sure alternative transport destinations are less expensive). Then each description was assigned a broad theme (e.g. unclear cost savings). Finally, an affinity process was used to group obstacles into similar coding, resulting in 39 coded categories (e.g. cost/funding) (see Figure Legend) (Langley et al, 2009). The codes were rank ordered by frequency overall, which allowed for an analysis of a Pareto distribution showing the most frequently identified obstacles (Kenett, 1991). Analysis of the obstacles across features was conducted with the same approach.

Case studies

The five cases where compared using five bundles of metrics, including i) socioeconomic indicators, ii) system design, iii) system activity, iv) performance measures, and v) financial measures.

Socioeconomic indicators across cases

The city populations ranged from 101 365–709 893; the median population was 537 734. The metropolitan statistical area populations ranged from 1 206 142–5 376 285, with a median of 1 652 602. Land area of the cities ranged from only 6 square miles to as much as 606 square miles; the median was 251 square miles. Population density, or the number of people per square mile, ranged from 833.8 to 15 762.8 persons per square mile; the median was 2 610.4 persons per square mile.

The percentage of the population identifying as White ranged from 33.2%–68.4%; the median was 65.4%. The percentage of persons aged 25 years or older reported as high school graduates ranged from 76.4%–89.5%; the median was 81.3%. The percentage of persons aged 65 years old or older ranged from 6.7%–11.5%; the median was 9.7%.

Per capita income ranged from $17 838–$31 156; the median was $24 163. The median household income ranged from $32 285–$47,979; the median for the five cities was $34 947.

Comparison of system design across cases

Three of the EMS systems were full service—providing both emergency 911 and non-emergency, interfacility ambulance service. One (the public utility model) was the exclusive provider of emergency and non-emergency ambulance service in its jurisdiction. The two governmental EMS systems (the fire service and the third service) were the only systems that exclusively served emergency 911 patients. In these two cases, non-emergency, interfacility ambulance service was provided by other organisations.

All five case cities were advanced life support (ALS) systems with a paramedic-level ambulance going on every emergency call. All but one EMS system staffed the paramedic ambulance with one paramedic and one EMT; the governmental third service staffed with two paramedics. Three EMS systems had medical first response at the ALS level: one was at the basic life support level and one was at the intermediate life support level. All five systems used a commercial protocol-based dispatch process for 911 caller interrogation and triage.

Response time goals for life-threatening emergencies ranged from 7 minutes 59 seconds to 9 minutes 59 seconds. All five systems measured reliability in achieving this goal at 90%. All but one of the EMS systems measured response times with the clock starting at some trigger for call receipt (e.g. phone pick up, first keyboard key stroke, etc.) and ending when the ALS ambulance arrived at the curb of the call location. The fire service system started the clock later in the call process (at the time the ambulance was dispatched) and did not include the call processing time.

Three of the five EMS systems have ambulances fully deployed at temporary locations throughout their service area such as key intersections able to respond quickly or relocate as coverage requirements dictate. One EMS system used a hybrid approach to deployment, mixing ambulances stationed in fixed locations (e.g. fire stations) with temporary locations. One EMS system used fixed posting locations only (e.g. fire stations).

Comparison of system activity across cases

Annual responses ranged from 21 000–97 000; the median was 73 845. Annual transports ranged from 16 000–68 000, with a median of 59 000. The percentage of patients transported ranged from 67%–76%, with a median of 73%. Non-transport rates ranged from 24%–33%, with a median of 27%.

The response unit hour utilisations range from 0.27–0.67, with a median of 0.52. Transport unit hour utilisations ranged from 0.20–0.47, with a median of 0.39.

Comparison of performance measures across cases

All cities reported a response time goal of 90% compliance. Four of the five EMS systems measured the time from call receipt at the medical dispatch centre to ambulance arrival at the call location. Compliance ranged from 85.9%–93.0%, with a median of 89.0%. Two EMS systems (governmental third service and the hospital-based service) were not in compliance with their goal at the time of the survey.

Cities reported data on one clinical outcome measure: the percentage of patients in sudden cardiac arrest (SCA) transported to the emergency department resuscitated with return of spontaneous circulation (ROSC). Four case cities reported data and the percentage ranged from 15%–50%, with a median of 20.23%. Cities reported data on three process measures related to the cycle time in STEMI, strokes, and life-threatening trauma. Two case cities reported the mean cycle time from symptom onset to percutaneous coronary intervention in STEMI patients, where the goal is 90 minutes. The mean times were 34 minutes 5 seconds and 42 minutes 2 seconds. Two case cities reported data on the mean cycle time from onset of stroke symptoms to emergency department arrival to enable the hospital to achieve the goal of intervention within 3 hours of onset of symptoms. The mean times were 38 minutes 47 seconds and 30 minutes 44 seconds. Two case cities reported data on the mean on-scene time in trauma patients with life-threatening injuries, where the goal is 10 minutes or less. The mean times were 17 minutes 5 seconds and 9 minutes 57 seconds.

Comparison of financial measures across cases

Annual EMS operating budgets in the four case cities reporting ranged from US$7 200 000–US$42 000 000. The median was US$23 180 033. The fire service case was excluded because it was an integrated department serving both a fire suppression mission and EMS service. It was not possible within the scope of this research to reliably determine cost of EMS delivery alone. The percentage of the annual EMS operating budget coming from tax subsidy ranged from 0%–36%, with a median of 10%. Two EMS systems received no community tax support and operated exclusively on user fee revenue.

The cost per capita ranged from US$43.97–US$71.03, with a median of US$45.21. The cost per unit hour ranged from US$147.26–US$178.43, with a median of US$149.34. The cost per response ranged from US$220.49–US$568.76, with a median of US$325.58. The cost per transport ranged from US$220.49–US$568.76, with a median of US$436.63.

Comparison of cases to the EMS system features

Table 2 summarises the case cities' adoption of the 15 EMS System features. All five case systems had adopted two design features: no call screening and all ALS. None of the systems had adopted alternative transport to destinations other than the emergency department. The public, exclusively 911 case systems had adopted the fewest system design features with just three: no call screening, all ALS, and quality improvement. The public utility model system and private provider systems exhibited adoption of the most system design features, with 10 and 8 features respectively. Five design features had ‘limited’ or no adoption: public intervention, reduced call cycle time, outcome-based performance measures, preparedness, and EMS health screening. The public utility system was the only case system to adopt a balanced scorecard.

Qualitative interviews

Interviews were conducted with senior clinical and operational leadership in each case EMS system, and a total of 582 obstacles were identified. Obstacles were organised into 39 categories and ranked from most commonly noted to the least common and displayed in a Pareto chart (Figure 2). Figure 3 displays the number of obstacles identified per EMS system feature and Figure 1 presents the obstacles to each EMS system feature in individual Pareto charts.

The top 10 categories represented 50.4% (293) of the total obstacles identified and were as follows: 1) cost and funding (70, 12.0%), the perception that adding a feature meant additional cost and required additional funding; 2) data measurement (46, 7.9%), the need to establish data collection methods and measure data effectively; 3) process and outcome focus (31, 5.3%), focusing on what the EMS system is aiming to achieve and then creating reliable processes to deliver the results; 4) systems view or design (26, 4.5%), appreciating an EMS system as interdependent parts working in concert and designed to produce the desired outcomes; 5) public information and education (22, 3.8%), increasing public awareness and knowledge about what the EMS system does, what citizens can do to help during life-threatening emergencies, and understanding the value of an effective EMS system; 6) knowledge or understanding of productivity (21, 3.6%), internal appreciation for the theories and methods of production strategies; 7) training or education (20, 3.4%), additional staff development is required to implement a feature; 8) will (20, 3.4%), individuals require the will to implement; 9) data definitions (19, 3.3%), a need for universal, operational definitions for measurement indicators; and 10) culture or tradition (18, 3.1%), requires overcoming the status quo or embedded professional culture.

Limitations

This study does have several limitations:

The multi-case study represents a small sample of urban EMS systems only and the case sites are all within the US. To understand whether these findings are relatively universal would require a larger sample of EMS systems representing broader demographics.

Uniform and universally accepted outcome and process measure are not defined and operationalised across EMS systems. The lack of an accepted data dictionary makes lateral comparison difficult. Data was self-reported and not independently validated.

The data analysis of identified obstacles relied on self-reported experiences or perceptions. The interview participants included clinical and operational leaders only. Adding in care providers, middle managers, support staff, and policy maker perspectives may produce additional insights.

The 15 features were identified through a literature review and the experience of a single researcher. A cohort of researchers may identify additional or different features worthy of consideration and further exploration.