Economic Analysis of Mobile Integrated Health Care Delivered by Emergency Medical Services Paramedic Teams

Key Points Question Is mobile integrated health care (MIH) delivered by emergency medical services more efficient than regular ambulance responses in addressing the needs of urgent care in the community? Findings This economic evaluation compared 1740 calls serviced by MIH in 2018 to 2019 with propensity score–matched ambulance calls for the same period and 2 years prior and found that MIH was associated with a decrease in the proportion of patients transported to the emergency department and saved health care costs compared with regular ambulance responses. Meaning These findings suggest that MIH is a promising and viable solution to meeting urgent health care needs while improving the efficiency in using emergency care resources.


Introduction
Health care systems are facing ever-increasing pressure to meet rising health care demand with limited resources. To address this issue, we must improve efficiency and reduce any wasteful spending in the system.
Mobile integrated health care (MIH) is a new model of community-based health care that uses community paramedics, who work directly with other health care professionals, to provide needs-based on-site urgent and nonurgent care. 1 The concept of MIH is to address specific health care needs of patients in their homes or within mobile environments and provide innovative approaches to health care that traditionally require emergency medical services (EMS), emergency department (ED) care, or hospital admission. This model was initially designed as a proactive means to improve health care access for underserved rural populations. It has since been adapted to real-time reactive response to 911 calls as a safe, timely, and mobile medical care delivery model in the community setting.
MIH programs can improve quality of life, reduce ED transport and admission, and reduce hospitalization for frequent users of the ED. 2 Existing MIH models have focused on the needs of specific community populations such as older adults with multiple chronic conditions, 3,4 individuals at high risk, 5 or individuals with acute mental health crises. 6 These types of MIH programs are delivered with physician oversight by interprofessional teams or community paramedics. MIH, which is integrated into and delivered by trained paramedics in partnership with other health care professionals, could also provide an efficient care service and reduce unnecessary use of ED and hospital resources. 7 Canadian literature 8 shows that 23% of EMS calls are low urgency, at Canadian Triage and Acuity Scale level 4 or 5. These patients may have been cared for in other ways, potentially by primary care.
Indeed, among those individuals calling EMS frequently, most experienced substantial loneliness and social isolation and had poor quality of life. Other responses to these EMS calls may be able to satisfy health care needs appropriately. In Canada, the MIH model has been incorporated into EMS in some jurisdictions. However, little is known about the performance of the MIH model delivered by EMS.
The primary objective of this study was to compare the time on task and costs between MIH and matched ambulance services from a public health care payer's perspective using real-world data routinely collected by the EMS system.

Methods
This economic evaluation was an analysis of administrative databases without contact with any patient or individual. Therefore, neither an ethics review board's approval nor informed consent was required, per 45 CFR §46. This study followed the relevant sections of the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) reporting guideline.

Niagara EMS MIH Program
Niagara EMS (NEMS) responds to unscheduled out-of-hospital emergent and nonemergent medical events for the Niagara Region of Ontario, Canada, a municipality of 1854 km 2 with a 2016 census population of 447 888. 9 NEMS operates an ambulance communication center with a land-based ambulance service that provides primary and advanced care paramedic services that approximate the scope of services provided by US emergency medical technicians and paramedics, respectively. 10 Since July 1, 2018, NEMS has implemented an MIH program that was devised to provide innovative and efficient responses to low-urgency and nondifferentiated emergency calls, which have accounted for one-third of call volume since its creation. The MIH teams primarily provide unscheduled services to those meeting the criteria, except for some scheduled follow-up visits between calls. Table 1 provides a brief description of the program.

Data Sources
The information for all emergency calls received and responded to from 2016 to 2019 was retrieved from the NEMS database, which records the location and priority description of the call, age and sex of the patient, the description of the response vehicle, whether transport to ED was required, and service times. Cost data were retrieved from the NEMS accounting database, including all equipment costs, operating costs, and salary and benefits of all staff employed.

Propensity Score Matching
Propensity score matching is a widely used matching method that mimics a randomized controlled design from an observational study on the basis of particular characteristics. 11 This method balances propensity scores between treated and untreated participants, achieving a similar distribution of baseline characteristics between the matched groups.
The MIH model was started in July 2018 with 4 teams deployed to supplement ambulance services. All calls serviced by the MIH team between July 1, 2018, and June 31, 2019, were used as an intervention cohort. Propensity score matching was used to identify a matched cohort of calls managed by EMS ambulance for the same period. Because EMS practices and budgets change over time, we also identified a matched cohort from each of the 2 years prior to the MIH implementation We applied and compared different propensity score matching methods. Nearest neighbor matching first selects a treated participant and then a matched participant whose propensity score is closest to that of the treated participant, while the optimal matching method matches participants so as to minimize the mean within-pair difference in propensity scores. 11 Caliper nearest neighbor matching is a modified nearest neighbor matching that matches 2 participants only if the absolute difference in their propensity score is within a prespecified caliper distance. 11 In our analyses, nearest neighbor matching within a caliper distance of 0.2 performed better than optimal matching. Therefore, for the base-case analysis, we chose nearest neighbor matching, with optimal matching included in the sensitivity analysis. The propensity score matching was conducted using R statistical software version 3.5.3 (R Project for Statistical Computing) and its MatchIt Package. Tests of statistical significance were run on all the variables used to drive the propensity scores, and there was no significant difference between the intervention and matched control cohorts.

Time on Task
Two time intervals recorded in the EMS system are time on scene (ie, from response to either final disposition on scene if no patient transported or to arrival at hospital if there was transport) and time at hospital (ie, from arrival at hospital until transfer of care to ED, only applicable for those with an ED transport). The mean time on task was calculated as the sum of time on scene and time at hospital One advanced care paramedic with community paramedic training and one mental health nurse Calls with a chief concern of mental health and/or addictions issues, including overdose; providing patients with community resources, referrals for ongoing medical or social needs, harm reduction training and supplies b a This involves scheduled follow-up visits with patients who have had multiple falls to provide an assessment to seek ways to increase independence and therefore decrease reliance on emergency medical services.
b This involves scheduled follow-up visits to known locations where drug use occurs and providing harmreduction kits.
and was compared between the MIH and the matched ambulance cohorts. We also compared the mean time on task for those with and without ED transport separately.

Costs Associated With MIH and Ambulance Services
Canada has a single-payer, publicly funded health care system. Therefore, our economic analysis was conducted from the public payer's perspective. Costs incurred for the provision of health care, whether by EMS or hospital, are all relevant and included in the analysis.

Base-Case Cost Analysis
We converted the time on task to cost and compared the total cost per 1000 calls associated with the MIH model vs the ambulance service. First, we calculated per minute cost by dividing cost by time in minutes. It is necessary that we differentiated between operating cost and capital or personnel cost in calculating the mean cost per minute. The latter is the cost incurred regardless of whether a vehicle is in service (ie, fixed costs), whereas the former is directly determined by the time in service (ie, variable costs). Therefore, mean operating cost per minute was calculated by dividing the cost by Second, we calculated the mean cost per call for on-scene task by multiplying the mean cost per minute with the mean time on scene, and the mean cost per call for hospital transport by multiplying the cost with the mean time at hospital. Each of the mean cost per call was multiplied by 1000 to derive the total cost. In addition, the cost of an ED visit was calculated by multiplying $184 (estimated cost per ED visit in Ontario adjusted to 2019 Canadian dollars) 12 by the number of ED transports. The total cost associated with each service model was, therefore, calculated using Equation 3: where C denotes the mean cost per minute, T the mean time on task, and N ED the number of ED transports per 1000 calls. It is important to note that for ambulance transport, the same mean cost per call was used for both on-scene task and hospital transport. For MIH, the mean cost per call for on-scene task was the cost derived for MIH, but the mean cost per call for hospital transport was the cost derived for ambulance transport of the same year because an ambulance was always dispatched and transferred the patient to ED even for those calls initially served by the MIH team.

Statistical Analysis
In the sensitivity analyses, we used propensity optimal matching to identify the matched cohorts. For the costing, we used the ratio of the number of MIH vehicles to the number of ambulances to allocate the administration cost between the MIH and ambulance services for 2018 to 2019. Furthermore, we excluded the administration personnel cost when calculating the cost per minute in the sensitivity analysis in a scenario assuming that the budget for administration was fixed and would not be changed even with the operation of the MIH program.
Statistical tests were conducted using t test for continuous variables and χ 2 test for categorical variables. A 2-tailed P < .05 was considered significant. Statistical analyses were performed using R statistical software version 3.5.3 (R Project for Statistical Computing) from January to April 2020.

Mean Times on Scene and on Hospital Transport
As shown in Table 2, of the 1740 calls, 498 (28.6%) serviced by MIH had ED transport (ie, after MIH team assessment transport to ED was deemed to be necessary or demanded by the patient),   As shown in Table 3 (Table 3).  Table 4.

Costs and Mean Cost Per Minute
The Figure shows the difference in total cost per 1000 calls between the matched ambulance service and MIH for each of the 3 years. The cost was substantially higher for regular ambulance response  than for MIH, with the difference ranging from $163 000 to $180 000 in the base-case and sensitivity analyses.

Discussion
This study estimated and compared the time on task associated with MIH compared with regular ambulances for matched emergency calls identified from the NEMS system. On average, the MIH team spent approximately 10 to 15 minutes longer on scene but reduced ED transport by 45% to 50%, each of which took approximately 40 minutes of ambulance service time. We further converted the time measures to dollar amount using cost per unit time. MIH was associated with savings of approximately 60% of the total cost compared with matched ambulances. reported that up to half of patients in the ED at any one time experience stress and its effects. 14 Avoiding an unnecessary ED visit but receiving the appropriate health care solution at home or at the place of the emergency call has the added potential to improve quality of life and satisfaction.

Strengths and Limitations
There are several strengths in our analysis worth highlighting. First, we used the real-word information routinely collected by EMS. Because of the nature of the EMS services, it is difficult and potentially unethical to design a randomized study to conduct a head-to-head comparison between MIH and ambulance transport. We used propensity score matching to identify a cohort of calls matched on important factors associated with demographic characteristics and the urgency of calls that were received from the same year. We further compared with the matched calls from 2 previous years to adjust for time and other unknown factors that might be associated with the characteristics of calls over time. We also assessed the impact of an alternative propensity score-matching algorithm and costing methods in the sensitivity analyses. We believed that the robustness of our estimates and conclusion have been enhanced by this analytical approach.
This study had some limitations. Although routine quality assurance was a part of this initiative, including qualitative feedback from patients, the EMS database does not record whether those patients called back and required ED transport later. Another limitation is that the cost of ED was an estimated mean cost for an ED visit in Canada and likely underestimated the true cost incurred by patients who were assessed and transported to hospital by EMS. Had all subsequent costs after ED admission been included, the cost savings would likely be even larger in favor of the MIH model. To overcome these limitations, we plan to link the EMS data with a public health administrative database in future studies to allow us to assess outcomes and subsequent use of health care resources, including any increased access to primary care, by those who receive care from MIH. MIH teams provided scheduled follow-up visits between calls. These service times were not recorded in the database. Because the fixed and variable costs associated with MIH have already been included in our analysis, if these between-call services had also been included, the cost per unit time for MIH would decrease, which would further favor the MIH model. It was a challenge when converting time to dollar value because both MIH and ambulances use the same infrastructure and staff support.
While recognizing the methodological challenges, we believe that presenting the time in monetary terms can deliver the message in a way that is relevant and understandable by health care budget planners, a target audience of this type of research.
NEMS only started the MIH program in 2018 with 5 response vehicles in operation. It is likely that the MIH team was not able to respond to all calls that met its responding criteria and not all calls serviced by the MIH team represented the best use of this service (eg, those calls that were initially dealt with by the MIH team but had ED transport). There is room for further improvement of this new service model so the resource use can be optimized. To do this, future research should include the collection of detailed information on the health conditions of patients and link EMS data with hospital admission databases. This will help accurately identify and map out the health care needs in the community that can be served by MIH. The MIH team can then be deployed with enhanced precision.

Conclusions
This economic evaluation's findings suggest that MIH delivered by NEMS was associated with reduced ED transport and saved substantial savings of EMS staff time and resources compared with ambulance for the matched emergency calls. This service model could be a promising and viable solution to meeting urgent health care needs in the community, while substantially improving the use of scarce health care resources.