Effectiveness of the community paramedicine at home (CP@home) program for frequent users of emergency medical services in Ontario: a randomized controlled trial

Abstract

Objective

To evaluate the impact of the Community Paramedicine at Home (CP@home) program, a community paramedicine home-visit intervention, on reducing emergency medical services (EMS) calls among frequent users.

Design

A 6-month, open-label, pragmatic, randomized controlled trial with parallel intervention and control arms. An online automated platform (randomizer.org) was used to randomly allocate participants using a 1:1 allocation sequence.

Setting

In homes of frequent EMS users in four paramedic services and regions across Ontario, Canada.

Participants

Eligible participants were frequent callers (≥ 3 EMS calls within six months and ≥ 1 EMS call within the previous month), or had ≥ 1 lift assist call within the previous month, or were referred by paramedics.

Intervention

Community paramedics conducted risk assessments, provided health education, referred appropriate resources, and reported to family physicians for up to three home visits. The control arm received usual care.

Primary outcome measure

EMS calls in 6 months during intervention.

Results

Two thousand two hundred eighty four eligible participants were randomly allocated to the intervention and control groups, with 265 participants lost to follow-up due to inability to retrieve participant records from EMS databases. There were 1025 intervention participants (52.7% female, mean age 69.65 years [standard deviation (SD) = 19.98]) and 994 control participants (52.0% female, mean age 69.78 years [SD = 19.09]). In the post-intervention intention-to-treat analysis (zero-inflated negative binomial regression), the EMS call rate was not significantly lower in the intervention group compared to the control group (incidence rate ratio [IRR] = 0.88, 95% confidence interval [CI]: 0.76, 1.01). In the subgroup analysis, the intervention had a significant effect in the lift assist caller subgroup (IRR = 0.73, 95% CI: 0.58, 0.92), but no significant effect among the frequent caller subgroup (IRR = 0.97, 95% CI: 0.82, 1.14). The sensitivity analyses found a similar association for the lift assist caller subgroup. There was a significant subgroup effect (p-value for interaction < 0.01).

Conclusions

CP@home had a significant impact on reducing EMS calls for those with a lift assist call but not for the overall sample. This program filled a healthcare gap by shifting primary care delivery, which could reduce the disproportionate number of EMS calls, thus reducing healthcare costs.

Trial registration

Registered with ClinicalTrials.gov NCT02835989 on July 14, 2016.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12913-024-11952-7.

Background

Emergency medical service (EMS) calls are rising worldwide alongside the growing and aging population [1, 2]. Frequent EMS calls worsen health system burdens by increasing demand, depleting limited resources, overcrowding emergency departments, and compounding healthcare expenditures [3–8]. Frequent EMS callers are generally older adults or people with multiple chronic diseases, substance use disorders, and/or mental health problems [4, 9]. These populations also have higher reported levels of poor health, complex health needs, and a lower quality of life [6, 10–12]. Frequent EMS callers are those who call often or can be potential frequent callers such as lift assist callers. Frequent callers have been defined as having three to ten calls in a year [12–14]. Although the definitions vary, many sources have defined frequent callers as having made four or more calls within one year [9, 15, 16]. Individuals who frequently call EMS are more likely to have unmet medical or social needs [4, 6, 12]. Medical-related frequent calls arise from individuals of any age who need help managing complex, chronic diseases [4]. Social factors that lead to frequent calls are loneliness, food insecurity and poverty, and having anxiety and depression [6]. Potential frequent callers can call EMS for a lift assist, which refers to help getting up after a fall [17]. These calls are most frequently for older adults but may also be in populations of all ages with specific conditions [17].

The burden of lift assist calls on healthcare systems is a growing concern, particularly among older adults who have been reported to account for over a third of all EMS calls [18, 19]. Lift assist callers may repeat their calls multiple times due to the recurrent need for help after falls [20]. In countries with aging populations such as Australia, Canada, and the United States, lift assist calls pose a particular challenge for paramedic services, contributing significantly to EMS demands and burdens [18, 19, 21]. The average age for lift assist callers is 75 years old, and the need for lift assistance is associated with short-term morbidity and mortality [20]. Lift assist calls divert resources, are physically demanding, and may require extra time to procure necessary resources resulting in a delay in response to other EMS calls [17, 18, 22]. These frequent calls can be prevented and more appropriately managed through community and primary care resources such as community paramedicine programs [23, 24]. Community paramedicine programs extend the traditional roles of paramedics to address primary and preventative health care needs in their communities [25].

The Community Paramedicine at Clinic (CP@clinic), a chronic disease management and health promotion program delivered in social housing buildings, demonstrated a significant decrease in monthly EMS calls among older adults [26–28]. The Community Paramedicine at Home (CP@home) program, adapted from CP@clinic, is a novel and innovative community paramedic led primary care intervention delivered in the homes of frequent EMS callers. The research objective of this study is to evaluate the impact of CP@home among frequent callers and lift assist callers on EMS calls compared to baseline and a control group.

Methods

Trial design and setting

We conducted a multi-site, open-label, pragmatic randomized controlled trial (RCT) with parallel intervention and control groups between May 2018 and March 2020 in four separate and not linked paramedic regional services across Ontario, Canada. These semi-urban paramedic services provide usual emergency medical care (through 911 call and dispatch) to the surrounding population, and had an estimated annual call volume of 386,000 calls in 2018. The intervention was performed by community paramedics, who underwent structured training, and occurred in the homes of the participants. The control group received usual care, which was the free primary and specialty medical healthcare provided to all ages in Ontario as part of the universal healthcare system. The full RCT protocol with detailed study design has been published elsewhere [29].

The study was approved by the Hamilton Integrated Research Ethics Board (HiREB Project # 2153). A waiver of consent was obtained from HiREB to receive administrative data for all participants randomized (intention-to-treat) and written, informed consent was collected from all participants who agreed to receive the intervention. Participation in the study was voluntary with the option to withdraw at any point.

Participants

Eligible participants were identified from the four regions where the participating paramedic services operate in Ontario. Participants were eligible if they (1) had in the previous 6 months called EMS three or more times and within the last month called EMS at least once, or (2) had a lift assist call to EMS within the last month, or (3) were referred by paramedics. Paramedics could provide a direct referral to the program for any patient they felt would benefit from inclusion in a home visit program, understanding that the patient would need to be randomized. Participants could only be included in the trial and randomized one time (i.e. they could not re-enter the trial if they called EMS again later).

Participants were excluded if they lived in a long-term care facility or were participating in a similar paramedic-led program that targeted frequent callers and/or a community paramedicine program that was delivered in the home.

Procedures

The data were collected during the home visits and stored in the electronic CP@home database, using the Research Electronic Data Capture (REDCap) [VIP Research Lab, Department of Family Medicine, McMaster University, ON, Canada] platform. Paramedic services administrative data were also obtained from the local paramedic services’ databases to assess change in EMS call volumes.

Intervention: CP@home

The CP@home intervention consisted of up to three home visits performed by community paramedics. The first home visit consisted of introducing the program, obtaining informed consent, and conducting various health assessments and risk analyses. Community paramedics completed validated assessments pertaining to chronic disease risks and management, social determinants of health, and health-related quality of life. Participant information and assessment results were recorded in the CP@home database that calculated and summarized risk factors from validated tools into a summary risk factor profile. Based on the results of each validated tool, the CP@home algorithm directed participants to appropriate healthcare and community services (e.g. healthcare and social services, legal aid resources, or wellness programs) and/or health education was provided (e.g. chronic disease management, health promotion strategies, and appropriate EMS use). The health and risk assessments were also sent to the participants’ family physicians with participant consent.

An appointment for a following home visit was made automatically and varied depending on participant availability and the time needed to implement plans and resources. The second home visit happened 2 to 4 weeks after the first visit and consisted of follow-ups and reassessments. The third and final home visit occurred 6 to 8 weeks after the first visit for final assessments. Participants were generally discharged from the intervention at this point. If the participant called EMS after the third visit and they still met the eligibility criteria, they could be re-enrolled in the intervention. If medical attention was needed, appropriate actions were taken such as to treat or facilitate transport to appropriate healthcare facilities.

Outcomes

The primary outcome was the difference in the number of EMS calls in frequent users compared to baseline and a control group. The data were collected from paramedic service administrative databases 6 months prior to the start of the intervention and 6 months during a participants’ enrollment within the intervention.

Sample size

The sample size was calculated based on the primary outcome compared between the intervention and control groups and a mean number of 5.3 EMS calls (5.7 SD) per frequent user over a 6 month period. A total of 261 participants per group was needed to detect a 25% difference between groups using 0.05 alpha and 0.80 statistical power. See details in published protocol [29]. As it was discovered that contact information was missing for nearly 40% of individuals in the paramedic service administrative data [30], oversampling was used to support a sensitivity analysis of those who received the intervention (per protocol) in addition to the intention-to-treat analysis.

Randomization

The regional paramedic services generated a weekly list of eligible participants that was reviewed and verified by the McMaster Community Paramedicine Research Team at McMaster University, Hamilton, Ontario. A random number generator was used to create a 1:1 allocation sequence via online automated platform (randomizer.org, seed #314,159, blocks of two). The allocation sequence was uploaded to the REDCap project database and eligible participants were allocated to either the intervention group, which received the CP@home intervention, or the control group, which received usual care.

Statistical methods

Descriptive statistics were used to describe baseline characteristics of participants, the number of EMS calls pre-intervention and post-intervention, and change in EMS calls after the intervention. Intention-to-treat analysis was conducted using General Linear Model (GLM) regression with zero-inflated negative binomial distribution to evaluate the incidence rate of EMS calls in the intervention group compared to the control group, as per allocation, adjusting for baseline EMS call rate. Sensitivity (per protocol) analysis was conducted to compare those who received the CP@home intervention to control. Subgroup analyses were conducted for both the intention-to-treat and sensitivity analyses to compare the differential effect of the intervention between frequent callers and lift assist callers on the incidence rate of EMS calls by the test of interaction between the intervention and subgroup variables. The results are reported as an incidence rate ratio (IRR) with corresponding 95% confidence interval (CI) and associated p-value. All statistical analyses were conducted on SPSS version 28. See detailed analysis description in the published protocol [29].

Results

Participant recruitment and characteristics

Participants were recruited and assessed for eligibility from regional paramedics service databases between October 2018 to March 2020. All eligible participants were based on the ‘lift assist’ or ‘frequent caller’ criteria; no participants were identified for the trial through the ‘paramedic direct referral’ option. There were 2284 participants identified, of which 1149 participants were randomly allocated to the intervention group and 1149 participants were randomly allocated to the control group (Fig. 1). At follow-up, there were 117 records and 148 records that could not be retrieved from the intervention and control groups, respectively [30].

Fig. 1.

CONSORT flow diagram

In the 1025 intervention participants, the mean age was 69.65 years (SD 19.98), 540 (52.7%) participants identified as female, 737 (71.9%) participants were identified as frequent callers, and 288 (28.1%) participants were identified as lift assist callers (Tables 1 and 2). Comparatively, in the 994 control participants, the mean age was 69.78 years (SD 19.09), 515 (52%) participants identified as female, 701 (70.5%) participants were identified as frequent callers, and 293 (29.5%) participants were identified as lift assist callers (Tables 1 and 2).

Table 1.

Characteristics of the total study population

Variables		Overall
		Intervention n = 1025	Control n = 994
Gender: Female	n (%)	540 (52.7)	515 (52.0)
Gender: Male	n (%)	484 (47.3)	476 (48.0)
Age (in years)	Mean, SD Median, IQR	69.65 (19.98) 76 (27)	69.78 (19.09) 75 (24)
EMS Calls Pre-Intervention	Mean, SD Median, IQR	4.41 (3.42) 4 (2)	4.35 (3.46) 3 (2)
EMS Calls Post-Intervention	Mean, SD Median, IQR	1.64 (3.15) 1 (2)	1.86 (3.72) 1 (2)
Difference in EMS Calls	Mean, SD Median, IQR	−2.76 (3.43) −3 (3)	−2.49 (3.31) −3 (3)

Notes: EMS Emergency Medical Services, pre-intervention and post-intervention were both 6-month periods

Table 2.

Characteristics of the frequent caller and lift assist caller groups

Variables		Frequent Caller Group		Lift Assist Caller Group
		Intervention n = 737	Control n = 701	Intervention n = 288	Control n = 293
Gender: Female	n (%)	404 (54.8)	360 (51.6)	136 (47.4)	155 (52.9)
Gender: Male	n (%)	333 (45.2)	338 (48.4)	151 (52.6)	138 (47.1)
Age (in years)	Mean, SD Median, IQR	66.56 (21.45) 74 (31)	66.31 (20.66) 71 (30)	77.56 (12.53) 80 (15)	78.07 (10.86) 80 (15)
EMS Calls Pre-Intervention	Mean, SD Median, IQR	4.48 (3.11) 4 (2)	4.44 (3.14) 4 (2)	4.22 (4.12) 3 (3)	4.12 (4.11) 3 (3)
EMS Calls Post-Intervention	Mean, SD Median, IQR	1.65 (3.44) 1 (2)	1.69 (3.14) 1 (2)	1.63 (2.25) 1 (2)	2.25 (4.82) 1 (3)
Difference in EMS Calls	Mean, SD Median, IQR	−2.83 (3.17) −3 (2)	−2.74 (3.03) −3 (2)	−2.58 (4.00) −2 (2)	−1.87 (3.85) −2 (3)

Notes: EMS Emergency Medical Services, pre-intervention and post-intervention were both 6-month periods

For the sensitivity analysis, there were 994 participants in the control group and 89 participants in the intervention group, of which 43 participants were identified as frequent callers and 46 participants were identified as lift assist callers. The demographic profile of the intervention participants is available in Supplemental File 1 (see eTable 1).

EMS call outcomes

Overall, the descriptive analysis found that the mean call difference was greater in the intervention group (−2.76 calls/person; SD = 3.43) compared to the control group (−2.49 calls/person; SD = 3.31) (Table 1). In both the frequent caller and lift assist caller subgroups, the intervention group had a greater decrease in the number of mean calls compared to the control group.

Intention-to-treat analysis

The intention-to-treat regression analysis determined EMS calls were not significantly lower in the intervention group compared to the control group (IRR = 0.88; 95%CI: 0.76, 1.01; p = 0.06) (Table 3). There was also a significant subgroup effect (p-value for interaction < 0.01), and the intention-to-treat regression analysis found that the intervention had a significant effect in reducing EMS calls in the lift assist caller subgroup (IRR = 0.73, 95% CI: 0.58, 0.92; p < 0.01), but no significant effect among the frequent caller subgroup (IRR = 0.97; 95% CI: 0.82, 1.14) (see Table 3 and Fig. 2).

Table 3.

Intention-to-treat zero-inflated negative binomial regression analysis of the incidence rate of EMS calls per person over 6 months between intervention and control groups, adjusting for baseline EMS call rate

	Incidence Rate Ratio (IRR)		Subgroup Interaction Effect
Type of Analysis	IRR (95%CI)	p-value	p-value
Overall ● Intervention (n = 1025) ● Control (n = 994)	0.88 (0.76, 1.01)	0.06	––-
Subgroup: Lift Assist ● Intervention (n = 288) ● Control (n = 293)	0.73 (0.58, 0.92)	< 0.01	< 0.01
Subgroup: Frequent Callers ● Intervention (n = 737) ● Control (n = 701)	0.97 (0.82, 1.14)	0.73

Fig. 2.

Sensitivity analysis using zero-inflated negative binomial regression of the incidence EMS calls per person in the frequent and lift assist caller groups, adjusting for baseline EMS call rate

Sensitivity analysis

The sensitivity analyses found no association between the intervention and control groups (IRR = 1.00; 95% CI: 0.85, 1.17; p = 0.996). However, while not statistically significant, it does appear there may be a trend in the differential impact between subgroups (subgroup interaction, p = 0.05), with a trend towards decreased calls for the lift assist subgroup and increased calls for the frequent caller subgroup (see Fig. 2).

Discussion

The RCT evaluating the impact of the CP@home program demonstrated a significant reduction in the rate of EMS calls in the intervention arm compared to the control arm for the lift assist group but not for the frequent caller group. This could be attributed to the fact that a lift assist is a more homogenous and well-defined reason for an EMS call with specific management options that have immediate results. The CP@home program may have been particularly effective in this group because it provided them with the necessary resources to address their specific needs such as fall prevention education (i.e., identifying and addressing risk factors for falls) and referring them to specific community resources. As well, individuals in the lift assist group may have been more receptive to the program and more likely to make behavioral changes that led to a reduction in their EMS calls [30].

The frequent caller group, however, is a heterogeneous population with a wide range of needs and various different reasons for requiring EMS urgently that can vary from call to call [10]. For example, frequent callers may have mental health illnesses, substance abuse issues, chronic health conditions, and/or unmet social care needs that require assistance to navigate solutions for [4, 6, 9, 12]. Frequent callers also have different socio-economic backgrounds, access to healthcare, and levels of social support influencing their health outcomes and EMS use [6, 16, 31]. Due to these differences and varying needs, it is more challenging to tailor a program to effectively meet the needs of frequent callers compared to lift assist callers. In addition, frequent callers may be lack the appropriate resources and support to implement health behavior change recommendations, and their call patterns may be more challenging and time-consuming to modify [12]. It is also hypothesized that contact with paramedics may reinforce the behavior of frequent EMS callers, where types of calls are largely undifferentiated, especially if the participant feels that they receive immediate and effective care during each encounter that they are not receiving elsewhere in the health system, leading to a cycle of dependence on EMS as their primary source of healthcare [32–35]. This may explain the observed increase in EMS calls among the frequent callers in the sensitivity analysis.

A rapid review identified seven community paramedicine studies that conducted a RCT analysis, none of which were community paramedicine programs delivered in homes of frequent callers [36]. Therefore, this study presented novel findings of the first RCT of a community paramedicine program for frequent callers delivered in their homes. Of the literature published on community paramedicine programs that were delivered in the home and evaluated using non-RCT designs, there was limited evaluation on the impact on EMS calls with a focus on ED visits instead. We found that most interventions had not been effective in frequent callers, however individualized case management programs were recommended and found to significantly reduce EMS calls among frequent callers [12]. At this time, in Ontario, it is beyond the scope of practice for paramedics to perform case management in individuals with severe mental illnesses.

Strengths and limitations

Strengths of this trial were the robust, pragmatic study design that increased generalizability and applicability. The results are applicable to frequent EMS users and lift assist callers and are generalizable to suburban and urban communities across Ontario. The results may also be generalizable to other regions and provinces outside of Ontario, but may be limited by differences in population and healthcare system organization and delivery. The intervention was also successful in reaching a vulnerable and often overlooked population. The study reported both intention-to-treat and sensitivity analyses, which aligns with best practices and provides a more comprehensive evaluation of the intervention’s impact. The study also utilized administrative data which has a low likelihood of response and recall bias. Another strength was the planned subgroup analysis between lift assist and frequent callers that allowed us to highlight differences in effectiveness between different populations [29].

A limitation of the study design was the lack of blinding, which was not feasible due to the nature of the program and could have led to performance bias. Not all participants who were randomized to the intervention group received the intervention since vulnerable populations are difficult to recruit, retain, and survey; alternatively they may have been in the hospital. Detailed recruitment challenges can be found elsewhere [30]. For example, their high rate of health and mental health issues, and a lack of trust in research and medical institutions all potentially could have impacted study recruitment and participation. Difficulties in contacting this population were evident as 48% were missing contact information [30]; in addition, some phone numbers provided were unavailable, participants did not answer calls or participant home addresses could not be located. There were also challenges in not being able to retrieve data records, which could be due to several reasons such as duplicate records from multiple ambulance vehicles responding to a single call, routine data cleaning that may have removed or deleted the records, and multiple calls for only one incident. Despite oversampling to support a sensitivity analysis (per protocol), the sample size of those who received the program was still limited and the sensitivity analysis was less than 10% of the intention-to-treat sample. Due to the COVID-19 pandemic, home visits had to cease; another site had been due to start CP@home visits in March of 2020 but were unable to due to the restrictions. Had the pandemic not taken place it is likely that the sample size that received the intervention would have been greater.

Conclusion

The CP@home intervention demonstrated mixed results across a diverse groups of patients in a semi-urban region of Ontario. The intervention is particularly effective in those patients with a lift assist call; however, the benefit for those with frequent calls was weaker and may have the unintended consequence of increasing EMS calls. To date, this is the only RCT to provide robust evidence of a reduction in EMS calls using a CP program delivered in the homes of frequent callers. Further analyses are needed to demonstrate CP@home impact on health outcomes and an economic analysis to evaluate its cost-effectiveness. CP@home could be a valuable health promotion program to implement and integrate in similar healthcare systems. This community paramedicine program could also have policy implications for primary care and paramedic services to inform the development of community paramedicine policies and roles. This program can contribute to shifting primary care delivery to address community needs and reduce healthcare burdens.

Abbreviations

CI

Confidence Interval

CP@clinic

Community Paramedicine at Clinic

CP@home

Community Paramedicine at Home

EMS

Emergency medical services

GLM

General Linear Model

HiREB

Hamilton Integrated Research Ethics Board

IRR

Incidence Rate Ratio

REDCap

Research Electronic Data Capture

RCT

Randomized Controlled Trial

SD

Standard Deviation

Authors’ contributions

G.A., R.A., M.P., F.M., and B.M. were involved in the study design. G.A., R.A., M.P., J.B. and L.T. analyzed and interpreted the study data. All authors were involved in the writing and preparation of the manuscript. All authors read and approved the final manuscript.

Funding

Canadian Institutes of Health Research, FRN 152951.

Data availability

De-identified data can be shared upon reasonable request by the corresponding author.

Declarations

Ethics approval and consent to participate

The study obtained ethics approval from the Hamilton Integrated Research Ethics Board (HiREB), Project # 2153 and was conducted in accordance with the Declaration of Helsinki and the Canadian Tri-Council Policy Statement: Ethical Conduct for Research Involving Humans. A waiver of consent was obtained from HiREB to receive administrative data for all participants randomized. Written informed consent was collected from all intervention participants.

Consent for publication

Not Applicable.

Competing interests

The authors declare no competing interests.