Cost-effectiveness analysis of workplace-based distributed cardiopulmonary resuscitation training versus conventional annual basic life support training

Abstract

Context

Although distributed cardiopulmonary resuscitation (CPR) practice has been shown to improve learning outcomes, little is known about the cost-effectiveness of this training strategy. This study assesses the cost-effectiveness of workplace-based distributed CPR practice with real-time feedback when compared with conventional annual CPR training.

Methods

We measured educational resource use, costs, and outcomes of both conventional training and distributed training groups in a prospective-randomised trial conducted with paediatric acute care providers over 12 months. Costs were calculated and reported from the perspective of the health institution. Incremental costs and effectiveness of distributed CPR training relative to conventional training were presented. Cost-effectiveness was expressed as an incremental cost-effectiveness ratio (ICER) if appropriate. One-way sensitivity analyses and probabilistic sensitivity analysis were conducted.

Results

A total of 87 of 101 enrolled participants completed the training (46/53 in intervention and 41/48 in the control). Compared with conventional training, the distributed CPR training group had a higher proportion of participants achieving CPR excellence, defined as over 90% guideline compliant for chest compression depth, rate and recoil (control: 0.146 (6/41) vs intervention 0.543 (25/46), incremental effectiveness: +0.397) with decreased costs (control: $C266.50 vs intervention $C224.88 per trainee, incremental costs: −$C41.62). The sensitivity analysis showed that when the institution does not pay for the training time, distributed CPR training results in an ICER of $C147.05 per extra excellent CPR provider.

Conclusion

Workplace-based distributed CPR training with real-time feedback resulted in improved CPR quality by paediatric healthcare providers and decreased training costs, when training time is paid by the institution. If the institution does not pay for training time, implementing distributed training resulted in better CPR quality and increased costs, compared with conventional training. These findings contribute further evidence to the decision-making processes as to whether institutions/programmes should financially adopt these training programmes.

INTRODUCTION

Cardiopulmonary resuscitation (CPR) training is typically required for all acute care providers. The conventional American Heart Association (AHA) Basic Life Support (BLS) course offered in a classroom-based format results in immediate acquisition of CPR skills but poor long-term skill retention. ^1–4 As a result, CPR quality in both simulated and real resuscitation events is suboptimal. ^5–7 Educational strategies such as distributed practice (ie, separating training into small sessions and dispersed over a period of time) ^8–11 and use of real-time feedback ^12–15 during training are effective at improving CPR skill acquisition and retention in healthcare providers.

More frequent training and use of advanced training equipment potentially increase the costs of equipment in a training programme. On the other hand, distributed practice with real-time feedback may also decrease the training time (by improving the efficiency) and costs for the instructors. The exact costs of conventional BLS course for healthcare systems are not clear. It is estimated that BLS training equates to more than 1 million training hours for practicing nurses in Canada, ¹⁶ as well as the extra costs for instructors and equipment each year. Although investing in medical training may benefit society by improving the quality of healthcare, careful evaluation of both effectiveness and resource allocation is necessary to inform the adoption of a training programme. To inform resource allocation decisions about implementing workplace-based distributed CPR training, decision makers need to know both costs and effectiveness of the distributed practice programme relative to the conventional training. To date, the literature reporting economic evaluations of simulation-based education training programmes is scarce. A systematic review described that only 1.6% of simulation-based studies have reported the cost of simulation training compared with other instructional approaches. ¹⁷ None of the studies identified were full economic evaluations involving a comparison of at least two alternative options in terms of both their costs and consequences. ¹⁸ The evidence supporting the efficacy of a distributed CPR training programme (ie, monthly practice) with real-time feedback compared with conventional CPR training method (ie, annual BLS course) has been well established. ¹⁹ In this study, we examine the cost-effectiveness of this educational strategy to inform whether programmes or institutions can feasibly adopt this educational method.

In our study, we aim to conduct an economic evaluation to exam the costs and learning outcomes of a workplace-based distributed CPR training programme relative to conventional BLS training. We hypothesise that distributed CPR training could result in decreased training costs and more effective learning outcomes relative to classroom-based BLS training.

METHODS

Overview of the cost-effectiveness analysis

This study was based on costs and outcomes from a prospective-randomised trial ¹⁹ conducted in the Emergency Department of Alberta Children’s Hospital in Calgary, Canada. Full details of the trial are published elsewhere, ¹⁹ but briefly, the aim of the trial was to compare the efficacy of a conventional annual BLS course to workplace-based distributed CPR training with real-time feedback among paediatric acute care nurses. ¹⁹ In the control group, participants received the standard 4-hour classroom-based BLS course led by a certified instructor according to AHA requirements, which includes didactic teaching (standardised video), CPR and automatic external defibrillator (AED) practice, and a written and skills test using a checklist. The control group received visual assessments by the instructor during practice and no real-time CPR feedback was provided during the course. The CPR quality was assessed right after the course for the control group. In the intervention group, participants were oriented to the equipment and expectations of the programme by a nursing staff member. Participants were required to practice CPR at least once a month, within the workplace, on a CPR training station equipped with objective real-time feedback (on compression depth, rate and recoil). The knowledge of resuscitation guidelines was also covered by online training material for the intervention group, which takes approximately 1 hour to complete (table 1).

Table 1.

A comparison between conventional BLS course and distributed CPR training

Content	BLS course (standardised 4 hours)	Workplace-based distributed CPR training with real-time feedback
Didactic teaching	50 min (standardised video)	60 min* (Online learning)
CPR practice	40 min	12×5=60 min
AED practice	10 min	0 min
Other logistics	85–90 min (waiting, observing, course break, questions and answers, etc)	5–10 min (orientation to the equipment)

*Didactic teaching for guidelines was covered by online training material for the intervention group. Instead of watching the videos in the classroom, participants watch this content using their spare time. Costs of didactic teaching is calculated as the costs of online material. In our local setting, time spent on online module was not compensated, therefore not calculated in the base case. However, the situation when online learning time is paid by the hospital is explored in the one-way sensitivity analyses.

AED, automatic external defibrillator; BLS, basic life support; CPR, cardiopulmonary resuscitation.

Research ethics board approval was obtained from the University of Calgary and written informed consent was obtained from all participants. The economic evaluation was conducted from the perspective of the institution (ie, the focus was on the costs incurred by the institutions), assessing the training costs and effectiveness (ie, CPR quality at 12 months) of distributed CPR training relative to conventional BLS recertification. The participants were asked to conduct chest compressions (CCs) on a manikin (with data collection capabilities) without any CPR feedback as measure of training effectiveness. Costs for both the distributed CPR training programme and the conventional BLS course were calculated (model structure presented as online supplemental efigure 1). This study is reported in compliance with the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) guidelines. ²⁰

Resource use and costs

The costs considered in this analysis were based on the educational cost-reporting framework by Levin et al, ²¹ which includes training programme costs, costs of training time paid by the hospital and costs for remediation in Canadian dollars in 2016. Total training programme costs were fixed, defined as the costs to implement the training programme for 1 year, including: (1) the cost of purchasing the equipment divided by the lifespan of each piece of equipment in different training groups, accounting for depreciation rate, (2) the costs of personnel (instructor, facilitator for equipment orientation and training and administrator) and (3) the cost for facility rental. Assumptions of lifespan and depreciation rate for all equipment were obtained from the manufacturers. It was assumed that the nurses receiving training would be compensated by the hospital for their time in training; salary costs for training time paid by the institution were calculated as the number of BLS course hours or training hours per participant multiplied by the hourly salary of nursing staff in 2016 based on the United Nurses of Alberta. ²² The standard BLS course time was 4 hours according to AHA BLS guidelines. ²³ The time participants received training in the intervention group was the actual time participants practiced CPR during the course of this research programme, which ranged from 0.75 to 2.08 hours with a median of 1.5 hours and IQR from 1.0 to 1.6 hours. Cost of remediation was calculated for those who did not achieve CPR excellence, which was estimated as the costs of one-to-one instruction by a nursing instructor for 30 min. A detailed description of cost estimation is provided in table 2.

Table 2.

Cost calculation for participants in the conventional BLS training and distributed CPR training programme per year

				Conventional BLS training (n=48)			Distributed CPR training (n=53)
	Unit cost (CAD)	Source	Unit of measure	Unit (n)	Depreciation (%)	Cost per trainee (CAD)	Units (n)	Depreciation (%)	Cost per trainee (CAD)
Equipment (fixed)
Adult manikin	$C2624.00	a	device	1	20	$C10.93	1	50	$C24.75
Infant manikin	$C1144.00	a	device	1	30	$C7.15	1	75	$C16.19
Feedback device	$C2013.00	a	device	−	−	$C0.00	1	40	$C15.19
Other training material	$C36.50	b	student	−	−	$C0.00	53	100	$C36.50
AED trainer	$C200	a	device	1	5	$C0.20*
Total						$C18.08			$C92.64
Personnel (fixed)
Facilitator (orientation)	$C40.00	c	hour	−	−	$C0.00	10	−	$C7.55
Administrator	$C30.00	c	hour	−	−	$C0.00	25	−	$C14.15
Instructor	$C50.00	c	hour	32	−	$C33.33	−	−	$C0.00
Facilitator (training)	$C20.00	c	hour	−	−	$C0.00	79.5	−	$C30.00
Total						$C33.33			$C51.70
Facility (fixed)
Rent for classroom with AV support	$C25.00	d	hour	32	−	$C16.67	−	−	$C0.00*
Training time paid by the institution
Cost for time spent in training	$C40.00	c	hour	192	−	$C160.00	79.5	−	$C60.00
Remediation (if not CPR competent)				Trainees (n)	Percentage (%)†	Cost per trainee (CAD)	Trainees (n)	Percentage (%)†	Cost per trainee (CAD)
Costs for instructor	$C50.00	c	hour	24	85.37	$C21.34	26.5	45.65	$C11.41
Clinical hour loss for trainee	$C40.00	c	hour	24	85.37	$C17.07	26.5	45.65	$C9.13
Total (adjusted)						$C38.41			$C20.54
Total costs						$C266.50			$C224.88
Source of unit costs in the analysis:	(a) Mean unit costs of Laerdal Resusci Anne QCPR, Laerdal Resusci baby QCPR, Laerdal SimPad SkillReporter in 2015 (b) AHA online learning module (c) Appendix of Nursing Salaries in Alberta 1948−2017 (Downloadable at URL: https://www.una.ab.ca/collectiveagreements/salaryappendix) (d) KidSIM facility rental

*Costs for storage of training equipment (a cart with manikin) and AED trainer do exist but are considered negligible in this analysis;

†Data obtained from Lin et al 2018. Resuscitation; CAD, Canadian dollars in 2016.

AED, automatic external defibrillator; BLS, basic life support; CPR, cardiopulmonary resuscitation.

Educational outcomes

Our educational outcome was ‘excellent CPR’, which is a composite dichotomous variable used in prior CPR studies. ^{8 11} We defined excellent CPR as achieving at least 90% compliance (in a 2-min skill session) for compression depth (greater than 50 mm), rate (100–120/min), and recoil (complete chest recoil) for each individual criterion. ^{24 25} CPR metrics were measured using Laerdal Resusci Anne and Laerdal SimPad SkillReporter. A light sensor is located in the chest of the manikin, which measures the timing in milliseconds (rate) and degree of the displacement in millimetres at initial (recoil) and peak depth (depth) of the chest during compressions. CPR quality was assessed for all participants at 12 months without real-time feedback, given that BLS training is required annually. Incremental effectiveness was calculated as the difference in the proportion of participants achieving excellent CPR for both complete cases (ie, lost-to-follow-up participants excluded from analysis) and with intention-to-treat approach (ie, the worst-case scenario, all the loss-to-follow-up participants were considered failing to achieve excellent CPR).

Cost-effectiveness analysis

This analysis was conducted with a 1-year time horizon, given that the BLS courses were required every year for acute care providers. Incremental cost was calculated as the cost difference between the distributed training group and conventional training group. Incremental effectiveness was calculated as the difference in the proportion of participants achieving excellent CPR between groups. There are several different combinations of incremental costs and effects possible. For instance, the intervention may be both more effective and less costly—in which case it dominates the control group and we can recommend that it should be adopted. Or it may be both more effective and more costly, in which case cost-effectiveness is expressed as the cost per increased providers achieving CPR excellence according to the incremental cost-effectiveness ratio (ICER), enabling a judgement about whether the extra benefit (ie, extra number of excellent CPR provider) is worth the extra cost.

ICER = (Cost_Intervention − Cost_Control)/(Effectiveness_Intervention − Effectiveness_Control)

The ICER is the difference in mean costs over the difference in the proportion of participants achieving excellent CPR, and indicates the additional cost required per additional excellent CPR provider to implement the distributed practice rather than conventional training.

We conducted one-way sensitivity analyses to determine whether the change of costs of uncertain parameter estimates, such as the cost of equipment or the policy (ie, training time not paid, no remediation for failing students, a self-directed approach without facilitation) would affect the ICER and the conclusions regarding cost-effectiveness. In our local setting, the time participants spent on the online module was not compensated. However, some institutions might pay for this amount of time. The control group received a conventional BLS course including AED training that was not covered in the intervention group. Although the cost of a complete BLS course should be evaluated, we were curious to see if a shortened version of BLS (excluding time for AED training, but assuming the same effectiveness) could influence the cost-effectiveness. These assumptions were assessed with a one-way sensitivity analysis as well as the impact of variability in effectiveness on cost-effectiveness of the intervention. One-way sensitivity analyses only indicate how individual parameter change affects the cost-effectiveness of the intervention, therefore, to determine the impact of the uncertainty of both cost and effectiveness parameters jointly, we conducted probabilistic sensitivity analyses in scenarios when CPR training is paid and not paid by the institution.

Probabilistic sensitivity analyses were conducted using Monte Carlo analyses, which allows every parameter in cost and effectiveness estimation to vary simultaneously over a specific probability distribution. This way, the uncertainty of a model input variables (ie, cost components, effectiveness) can be used to generate CIs for outputs (ie, incremental costs, incremental effectiveness, ICER). Parameters related to costs were assigned a log-normal distribution ²⁶ where the ranges of costs were used to calculate the parameters of the distribution [ie, mean = ln(minimum) + ln(range)/2, SD = ln(range)/4]. The probabilities of excellent CPR were obtained by bootstraps of the study data using technique of resampling with replacement. Five thousand replicates of data simulation were performed to approximate the CIs of outputs and presented as a scatterplot in a cost-effectiveness plane. Neither the costs nor the educational outcome was discounted given that the study period does not exceed 12 months.

RESULTS

Trial participants

A total of 101 healthcare providers were enrolled in the study; 53 participants were randomly assigned to the intervention group and 48 were assigned to the control group. During the trial, one participant (1.9%) in the intervention group did not receive training due to medical reasons. At 12 months, 7 participants (14.6%) in the control group and 6 in the intervention group (11.3%) were lost to follow-up because we were unable to contact them, or they transferred to a different institution. In the intervention group, 85% (39/46) participants completed the monthly practice (total yearly practiced session range: 12–25) and 15% participants missed some monthly practice session (total yearly practiced session range: 9–12). The demographic characteristics between two groups were similar.

Costs

Table 2 provides details of the cost components for both groups in the study. Compared with the control group, the intervention group costs were higher in equipment (control $C18.08 vs intervention $C92.64) and in personnel (control $C33.33 vs intervention $C51.70) per learner, but were less in facility rental costs (control $C16.67 vs intervention $C0) and costs of training time (control $C160.00 vs intervention $C60.00) per participant. After adjusting for costs of remediation, the incremental cost of the intervention group relative to control group was −$C41.62 per participant (control $C266.50 vs intervention $C224.88) for base case. A detailed calculation of costs was provided in online supplemental efigure 1.

Effectiveness of distributed training on CPR quality

The performance of the distributed training group was superior to the conventional training group in CPR competency (incremental effectiveness: 0.397, 95% CI 0.217 to 0.517, p<0.001) at 12 months. ¹⁹ When all lost-to-follow-up participants were considered as failing to achieve competency, the proportion of participants achieving excellent CPR is still significantly higher in the distributed training group than the conventional training group (Incremental effectiveness: 0.346, 95% CI 0.183% to 0.511%, p<0.001) (table 3).

Table 3.

Cost-effectiveness of distributed training versus conventional training and one-way sensitivity analysis

	Conventional BLS training		Distributed CPR training		Incremental costs	Incremental effectiveness	ICER/interpretation
	Mean cost (CAD)*	Effectiveness	Mean cost (CAD)*	Effectiveness
Base case	$C266.50	0.146	$C224.88	0.543	−$C41.62	+0.397	Distributed training dominates
Sensitive to change of costs
No remediation	$C228.08	0.146	$C204.33	0.543	−$C23.75	+0.397	Distributed training dominates
Training not paid by the hospital	$C106.50	0.146	$C164.88	0.543	+$C58.38	+0.397	$C147.05 per extra excellent CPR provider
No remediation & training not paid by the hospital	$C68.08	0.146	$C144.33	0.543	+$C76.25	+0.397	$C192.07 per extra excellent CPR provider
Least expensive equipment	$C263.24	0.146	$C210.78	0.543	−$C52.46	+0.397	Distributed training dominates
No facilitation for training (assuming self-directed distributed training in achieving the same effectiveness)†	$C266.50	0.146	$C194.88	0.543	−$C71.62	+0.397	Distributed training dominates
Conventional BLS course shortened to 3 hours (removing costs for AED and didactic video time)†	$C226.50	0.146	$C224.88	0.543	−$C1.62	+0.397	Distributed training dominates
Online module time (1 hour) in the distributed CPR group got compensated†	$C266.50	0.146	$C264.88	0.543	−$C1.62	+0.397	Distributed training dominates
Sensitive to effectiveness
All lost-to-follow-ups assumed failure (intention-to-treat)	$C266.50	0.125	$C224.88	0.472	−$C41.62	+0.347	Distributed training dominates
Least effective‡	$C266.50	n/a	$C224.88	n/a	−$C41.62	+0.217	Distributed training dominates

*Mean costs per trainee per year (Canadian dollars in 2016).

†Assuming change of condition will influence cost estimation only, but not the performance of both groups.

‡Incremental effectiveness set as lower bound of 95% CI of base case.

AED, automatic external defibrillator; BLS, basic life support; CPR, cardiopulmonary resuscitation; ICER, incremental cost-effectiveness ratio.

Cost-effectiveness

As presented in table 3, we found an incremental effectiveness of +39.5% for excellent CPR and incremental costs of −$C71.62 per trainee (distributed training relative to conventional training) in the base case, which indicates that distributed CPR training is both more effective and less costly and thus dominates the conventional BLS course. With the intention-to-treat analysis, when all participants lost to follow-up were assumed incompetent, distributed CPR training with real-time feedback still dominates the BLS recertification course, with an incremental effectiveness of +34.5% with the same incremental costs.

This finding is robust to a range of one-way sensitivity analyses. One-way sensitivity analyses revealed that distributed CPR training consistently dominates the annual BLS training when the costs of equipment and remediation change, or the intervention is less effective. When time spent during online module training is paid or AED training in the control group is not included in the cost calculation, the incremental costs between groups are close to 0 (incremental cost $C1.62). In these scenarios, distributed CPR training costs almost the same as, but more effective than the conventional training. The results of the cost-effectiveness model are sensitive to whether the training time for staff is paid by the institution, in which case the distributed training is no longer dominant. In the most extreme case, where no remediation was provided, and training time is not paid by the institution, distributed CPR training is more effective but also more expensive with an ICER of $C192.07 per extra participant achieving CPR excellence (table 3).

When training time is paid, 4993 of 5000 (99.9%) bootstrap replication estimates were in Quadrant IV (lower right quadrant) of cost-effectiveness plane, where distributed CPR training is more effective and less costly than the conventional BLS course (figure 1A). If the training time is not paid by the institution, distributed CPR training no longer dominates the conventional BLS training, and instead is more effective and more costly with an ICER of $C147.05 (95% CI: $C100.89–294.10) per extra participant performing excellent CPR (figure 1B).

Figure 1.

Monte Carlo simulation for cost-effectiveness. (A) Training time is paid by healthcare institution. In the graph, 4993 out of 5000 points (99.9%) were in quadrant IV (lower right), where the distributed training is more effective and less costly, compared with conventional training (origin). (B) Training time is not paid by the institution. In the distributed training group yields an ICER of $C147.05 (95% CI 100.89 to 294.10) per extra excellent CPR, compared with conventional training (origin).

DISCUSSION

To our knowledge, this is the first study examining the cost-effectiveness of workplace-based distributed CPR training relative to the conventional CPR training method. We found distributed CPR training is more effective than the conventional method. And it could be delivered with decreased costs compared with conventional training under certain conditions (ie, the hospital compensates staff members for their time spent in CPR training). In this scenario, the workplace-based distributed training can definitely be recommended for adoption. However, in a setting where CPR training time is not paid, the distributed practice group is still more effective, but now costs $C58.38 more than conventional training group per participant (incremental cost). It would cost an additional $C147.05 per additional excellent CPR provider trained, if adopting distributed training instead of conventional BLS training. In this case, the decision-maker must decide if the improved training outcomes are worth the extra costs.

Cardiac arrest is relatively rare event that healthcare providers will face in clinical practice. As a result, healthcare providers often lack opportunities to apply CPR skills other than during annual training. This explains the decay in CPR skills after annual BLS training. ^1–4 Our intervention provided refresher training within the workplace to boost and maintain CPR skills on a monthly basis, thus preventing skill decay. These findings are consistent with previously published literature. Niles et al have shown in cohort studies that workplace-based bedside ‘rolling refresher’ training significantly improves the retention of CPR skills, ⁹ with providers taking less time to achieve CPR competency. ⁸ Other studies examining the effect of ‘low-dose, high frequency’ training for either 6 months ¹¹ or ‘6 minutes of monthly practice’ on nursing students ¹⁰ have supported our results. However, none of these studies reported the costs of the training, leaving educators and programmes administrators wondering if improvements in skills were at the expense of incremental financial costs.

The cost of classroom based, annual BLS recertification training sessions is huge. Based on statistics from the Canadian Nursing Association (CNA) in 2017, there are 301 010 nurses in Canada. ¹⁶ A 4-hour standard BLS recertification course ²³ leads to more than 1.2 million hours for nurses each year. Assuming all nurses require BLS certification, and an estimated hourly nursing salary of $C40/hour, ²² this equates to nearly 50 million dollars spent on BLS recertification training for nurses only each year in Canada. Optimising the cost-effectiveness of CPR training could help release more financial resources and allocate them to other more pressing healthcare priorities.

Very few studies have examined the cost-effectiveness of resuscitation training or medical education programmes. ¹⁷ Studies that claim to examine the cost-effectiveness of interventions often suffer from inadequate estimation of costs (ie, only including equipment costs) and false presentation of cost-effectiveness (ie, not presenting ICER). ^{27 28} Key components of an economic evaluation, such as timeline, perspectives of analysis, sensitivity analyses are typically not included. ¹⁸ These have illustrated challenges encountered by many investigators aiming to conduct cost-effectiveness analyses of medical education interventions.

In this cost-effectiveness analysis, we demonstrated that although the distributed training group incurs more equipment and personnel costs compared with conventional methods, the overall cost was in favour of the distributed training group in a setting where training time is paid by the institution. The cost estimation represents the situation when we designed and implemented this research. At the present time, there might be more inexpensive products on the market that provide feedback and collect CPR quality data. This could further decrease the total costs of training programme as well as the incremental costs, making the distributed practice group even more cost-effective relative to conventional training. Participants in the intervention group practiced within the workplace in a longitudinal fashion, spending less time for course logistics and more time on CPR practice with objective feedback, therefore leading to decreased training time. Conventional training, however, involves training participants in groups, where the participants use training time less efficiently (spending less time practicing and more time on didactic teaching and course logistics). Furthermore, the intervention group incurred less remediation costs due to increased effectiveness of training. In a situation where the training time is not paid, the distributed CPR training programme resulted in increased costs and more effective learning outcome with an ICER of $C147.05 per extra excellent CPR provider. In this case, adoption of the distributed training programme relies on the budget impact on training and opportunity costs (ie, the forgone benefits that would have been achieved if not implementing distributed CPR training). To provide a hypothetical example, consider that there are 100 providers that need to receive CPR training annually in an institution. Implementing distributed CPR training could result in an additional 39 excellent CPR providers compared with conventional training (15% vs 54%). This will require additional training costs of $C5734.95 ($C147.05×39) annually. Policy makers would need to decide whether there is a better way to achieve the same learning outcome with lower costs in their local setting (in which case, opportunity cost of distributed practice is too high to implement it). Because some of the costs are for capital equipment, and the cost-effectiveness analysis spreads these costs over their lifetime, a practical consideration may also be cash-flow issues to purchase equipment up front.

The AHA is implementing the Resuscitation Quality Improvement (RQI) programme, which involves distributed CPR training with real-time feedback (self-directed, a minimum of quarterly). It is worth mentioning that the results of our study reflect a monthly CPR practice programme with facilitation of training. In implementation without facilitation, additional effort and/or costs may be required to ensure compliance of distributed training. Thus, our results are not generalisable to programmes where CPR training is self-directed or less frequently practiced, given that the effectiveness and costs might be different depending upon training frequency.

Limitations

Our study has several limitations. We did not examine direct patient outcomes in the cost-effectiveness analysis. Survival and neurological outcome of the patient are associated with other factors, such as time to administer epinephrine, time to defibrillate (for shockable rhythms) and underlying conditions of the victims. ^29–31 In addition, CPR quality is also influenced by factors other than training, such as the type of mattress on which the patient was put, ³² the use of backboard ³³ and height of the CPR providers. ³⁴ At present, it is not practical to account for all of these confounding factors in a trial-based economic evaluation. The cost-effectiveness conclusion of distributed practice relative to conventional training might also change, if taking the costs of post-arrest care into consideration. Future research should consider linking the cost-effectiveness of CPR training with patient outcomes, such as quality-adjusted life year gain. Second, we conducted this study in the emergency department of a paediatric tertiary care centre that is heavily involved in CPR research. The cost-effectiveness analysis is conducted from an institutional perspective. Generalising the conclusion to a different setting or a broader perspective (ie, healthcare system) should be with caution. Third, we only examined the quality of CC as the educational outcome. There are other important components in the BLS course (ie, ventilation, use of automated external defibrillator) whose effectiveness is not evaluated within the current study. However, bag-mask ventilation and AED are less practical to most acute care providers, since cardiac arrest patients are usually managed with advanced airway and manual defibrillators. The difference in training content could potentially introduce bias to this study, however, we attempted to adjust the costs of these different components in the sensitivity analyses. Further research would be needed to address these benefits of AED and ventilation in BLS training.

CONCLUSIONS

The use of distributed practice and real-time feedback resulted in decreased costs with increased CPR training effectiveness when training time is paid by the institution. Adoption of distributed training is recommended in this situation. In contrast, in a setting where the institution does not pay the wages for the healthcare providers to receive training, distributed training resulted in better CPR quality, but increased costs. Decision-makers in such a setting must weigh whether the improved CPR training outcomes are worth the additional training costs. This research provides evidence for administrators to make decisions regarding the adoption of educational interventions.

What is already known on this subject.

• Current CPR training method does not result in high performance in healthcare providers.

• Yet the healthcare system spent millions of dollars each year for this type of training.

• Distributed CPR training with real-time feedback improved the CPR quality in healthcare providers.

What this study adds.

• Workplace-based distributed CPR training results in decreased training costs and better learning outcomes in certain condition.

• This study provides evidence to the cost-effectiveness of distributed CPR training and promotes the implementation of this.