Cost–consequence analysis of daily continuous remote monitoring of implantable cardiac defibrillator and resynchronization devices in the UK

Abstract

Introduction

Cardiac implantable electronic devices (CIEDs) are frequently used for the treatment of bradycardia and heart failure as well as for preventing sudden cardiac death. Following device implantation, conventional care requires patients to regularly attend calendar-based in-clinic follow-up (FU) visits to monitor and optimize device function, to evaluate device diagnostics, and to assess patient health status.¹ However, the vast majority (>90%) of all calendar-based FU visits does not identify any issue and therefore does not result in further action.² Thus, they are an unnecessary burden for patients, their caregivers, and the health-care system as a whole.

The number of patients with CIEDs has been steadily increasing, and this trend is likely to continue.³ The burden on the health-care system generated by conventional follow-up (CFU) is thus expected to rise in the future, further straining the budgetary and technical capacity of health-care systems to provide ongoing care for patients with CIEDs. According to recently modelled estimates, the number of CIED patients in the UK was about 225 000 in 2010. At the current annual growth rate of about 8%, the patient pool is expected to reach ∼334 000 in 2015.⁴ All these patients will require FU services for as long as they have their device implanted, which is usually lifelong.

Professional societies have endorsed remote FU as a safe replacement for in-office services, and hospitals have developed workflow schemes that improve efficiency and suit their local needs.^1,5,6 Remote management reduces the travel burden for patients, thus potentially improving long-term adherence when compared with in-office FU.² However, adoption of remote CIED management in Europe has been hindered by the lack of reimbursement.^7,8 Uncertainty regarding the long-term economic consequences of remote CIED management has been cited as a hurdle against public funding.^7,9 The model presented here aims to address this gap.

The objective of this evaluation is to assess the long-term costs and consequences of using remote CIED management in patients implanted with implantable cardioverter defibrillators (ICDs) and cardiac resynchronization therapy defibrillators (CRT-Ds) for either primary or secondary prevention. This evaluation takes a UK National Health Service (NHS) perspective.

Methods

Remote cardiac implantable electronic device management

Although systems from different CIED manufacturers share some common features, there are inherent differences in technology and function.⁹ For the sake of simplicity and consistency, we chose to limit our analysis to the BIOTRONIK Home Monitoring® system (HM), as it currently has the most published data and since we had access to additionally required but unpublished information. Home Monitoring® provides remote monitoring with clinical and technical data transmitted automatically on a daily basis via the mobile phone network, plus instant automated alert transmission in case of a pre-specified parameter deviation (e.g. abnormal lead impedance, atrial fibrillation burden etc). This allows for an earlie detection, review and intervention in case of relevant clinical and technical events, compared with a scheduled in-office FU regimen.² Home Monitoring® is a fully automated system that neither requires patient interaction for data transmission nor a manual re-set after alert interrogation.

Modelling approach

The economic evaluation was performed as a cost–consequence analysis comparing long-term FU using either HM or CFU in ICD and CRT-D patients.

A cost–consequence analysis is a variation of the cost effectiveness approach that provides costs and outcomes (consequences) in disaggregated form, and thus is more transparent than, e.g. an analysis reporting cost per quality-adjusted life-year. It also leaves the decision regarding the relative importance of different outcomes to the reader.

Twelve consequences were examined in the model. These events were either clinical events or device-related technical events, namely scheduled and unscheduled FU visits, battery replacements, lead malfunctions, atrial fibrillation/flutter (AF), lead-related inappropriate shocks, non lead-related inappropriate shocks, stroke, hospital admission for acute decompensated heart failure (ADHF), sustained ventricular arrhythmia (SVA), appropriate shocks triggered by SVA, and death. Cost of the HM system and costs of managing the included consequences were captured in the model.

A deterministic Markov cohort model was developed, using TreeAge Pro 2009 (TreeAge Software Inc.). The Markov cycle length was 1 year, with a base case modelling period of 10 years, ensuring that device replacement would be captured by the model. Half-cycle correction was applied to all costs except to the initial one-off investment for the HM transmitter (CardioMessenger II, BIOTRONIK) in the first cycle (HM arm only). Costs and outcomes were discounted at 3.5% per annum, in line with the UK guideline.¹⁰

Model structure

The patient population represented in the model consists of patients who have undergone an ICD or CRT-D implantation and are managed in an outpatient setting. One treatment arm represents patients managed by HM, while the other arm consists of patients managed with CFU. Four main health states were included: Well, Post-stroke, Post-ADHF, and Dead as an absorbing state (see Figure 1). Stroke and ADHF were deemed to have long-term clinical and/or economic implications, warranting separate ‘post’ health states. If patients in the ‘Post-ADHF’ state experienced a stroke, they were re-assigned to ‘Post-stroke’. Patients who entered the ‘Post-stroke’ health state stayed in this health state until death (or model termination).

Figure 1

Transition diagram for the presented model.

Open in new tabDownload slide

In every Markov cycle and health state (except ‘Dead’), the probabilities of experiencing the clinical and technical events (see above) were applied. The model consists of individual modules for exposing every patient to all possible events (e.g. battery life module, lead issue module, non-lead issue-related shock module, and cardiovascular (CV) events module; refer to Supplementary Data for the exact structure). The order of these modules does not indicate a certain order of events in a patient. In fact, the modules may be re-arranged with no effect on the results, as the actual sequence and timing of multiple events within one cycle is irrelevant for estimating health outcomes and costs.

The model structure was identical in the two compared FU options (see Supplementary Data for model structure).

Model data inputs

Clinical and cost data were identified via structured searches using MEDLINE and systematic review of the identified sources. While most of the event data were taken from randomized controlled trials (RCTs) for HM; all costs were specific for the UK. Data from multiple sources were pooled or meta-analysed. Where required, data specific for ICD and CRT-D patients or for gender were weighted, based on the number of procedures performed in the UK or the gender split in the UK population. Further data on e.g. patient age at implantation were gathered from relevant sources. While the complete data appraisal cannot be reported here in detail, key model input data are provided in Table 1, with the complete model input data reported in Supplementary Data.

Event rates were converted into annual probabilities.¹¹ Some of these probabilities remain constant, while others change over time depending on age or prior clinical events.

While there are some data suggesting that HM can reduce the risk of CV events such as stroke^2,12 these data are still at early stages of development. To this end, no differential risks of CV events were applied to the two compared strategies. The model can, however, easily host new clinical data as they become available over time.

As the RCTs comparing HM with conventional FU had a relatively short FU, they were not suited to provide data on mortality. Saxon et al.¹³ report overall survival in ICD and CRT-D patients based on the ALTITUDE registry. A matched pair analysis of 10 272 patients showed a 50% reduction in mortality in patients on remote monitoring over the first 4 years. However, due to the inherent limitations of a registry we did not attribute a survival benefit with HM, but considered the mortality as observed for conventionally followed patients for both FU groups.

By providing earlier detection of clinical and technical complications that drain the battery reserve, HM prolongs the life of a CIED and thus reduces the necessity for device replacement when compared with CFU. In an analysis of the ECOST trial, HM reduced appropriate and inappropriate shocks by 71% and the number of capacitor charges by 76%.¹⁴ Battery life of CIEDs is primarily determined by two factors; frequency of capacitor charging and percentage of pacing. No extra battery consumption attributable to the operation of HM was considered, given that it is negligible (equivalent to about a single 30 J shock over the entire device lifetime).¹⁵ Capacitor charges occur for routine capacitor formatting, for aborted shocks, and for delivered shocks (appropriate, inappropriate, and ineffective). The relationship between capacitor charging, percentage of pacing, and battery drainage was incorporated in the model, and as described for the BIOTRONIK device models most widely used in the UK: the LUMAX 540 DR-T at 0% pacing was considered as ICD, the LUMAX 540 HF-T at 100% pacing as CRT-D.¹⁶

Predictive validity of the model was tested by assessing how well event numbers from the pivotal RCTs were predicted by the model. Debugging of the model was undertaken by entering zero or extreme values for mortality, number of FU visits, and shock probabilities. Multiple sensitivity analyses were performed to address data uncertainty.

Results

The results of the 10-year base case analyses are presented in Table 2 (event numbers) and Table 3 (costs).

Our model reflects what is reported from the clinical trials: HM reduces the number of patients affected by inappropriate shocks due to lead issues and AF (as well as the actual number of shocks in case of shock events but that is not reported as an outcome here). Home Monitoring® also reduces the need for battery replacements (as a result of less battery charges), and lowers the number of in-office FU visits significantly. The increase in the number of unscheduled visits is caused by the need to FU automatic alerts issued by the HM system. Overall, HM is cost neutral compared with CFU, with only a minor cost advantage predicted for HM over CFU (GBP 33 over 10 years). This includes all costs for the initial investment in the HM technology and all ongoing remote CIED management services.

The findings of the sensitivity analyses are shown in Figures 2 and 3. The data relate to the incremental changes in the discounted cost with HM vs. CFU. A number of scenarios were identified where a change to the model input data would result in notable changes to the incremental cost with HM: Considerable cost savings could be expected over a 15-year modelling period, if patients would routinely attend four or more CFU visits per annum, if HM reduces stroke risk (RR = 0.28) as reported in the TRUST trial,² if device replacement costs were higher than considered for the base case, and if cardiologists were to participate more frequently in in-office FU visits (82% instead of 20%). Home Monitoring® would cause additional costs if replacement costs were lower than considered for the base case, if a 5-year modelling period would be considered, and if patients would routinely attend two CFU per annum. The base case results were insensitive to changes in the rate of SVA, costs for treating lead issues, and patient age at implantation (not shown).

Figure 2

Univariate sensitivity analysis: discounted incremental costs per patient on HM compared with conventional in-clinic FU, negative data indicate cost savings with HM.

Open in new tabDownload slide

Figure 3

Univariate sensitivity analysis for modelling period: Discounted total costs per patient on HM vs. conventional in-clinic FU.

Open in new tabDownload slide

A separate analysis of different modelling periods was performed (Figure 3). If shorter modelling periods are considered, the initial investment in HM results in a higher cost per patient. The effect of extended battery longevity with HM is clearly visible at year 7, when the majority of CFU patients require box changes for battery depletion, which occurs at year 8 for HM patients (resulting in cost savings of GBP 5433). For use of HM for 8–9 years, additional costs are negligible (GBP 56 and GBP 9, respectively). If used for 10 years or longer, HM becomes increasingly cost saving.

Discussion

Main findings

This is the first comprehensive long-term economic model for any CIED remote management system. A Markov cohort model assuming no mortality benefit predicts cost neutrality when applying HM in the FU of ICD and CRT-D patients. Apart from being cost neutral in the UK NHS environment, HM reduces the demand for in-office services, thereby allowing to redirect health-care services. This study fills the gap of a thorough long-term economic assessment of costs and consequences which often has been cited as a hurdle towards reimbursement.

Previous studies

The impact of HM on health-care resource utilization and clinical outcomes has been studied and reported before as part of observational studies and RCTs such as Home ICD,¹⁷ OEDIPE,¹⁸ TRUST,^2,19 COMPAS,¹² and ECOST.^14,20 Abbreviated cost assessments have been undertaken in the past, for example, by Fauchier et al.²¹ who estimated the costs of using HM in France, including costs for in-office FU visits, associated medical services, and transportation. However, the report did not consider costs for clinical and technical issues requiring unscheduled visits, or reimbursement for remote FU.

An evidence review undertaken by the NHS Purchasing and Supply Agency in 2005 provided a modelled economic analysis from a UK NHS perspective, and for patients with prior AF and fitted with a CIED.²² Based on the evidence available at the time, and not considering technical issues requiring medical intervention, changes in battery longevity, or a survival gain from remote monitoring (RM), it was concluded that the initial investment into RM would not be offset by cost savings within 10 years—a finding that is contrary to the comprehensive economic assessment presented here. However, the Centre for Evidence-based Purchasing assessment also concluded that a relative reduction in stroke incidence by RM of as little as 5% would be sufficient for it to be cost-effective by National Institute for Health and Clinical Excellence standards. Ongoing studies (IN-TIME and IMPACT) assessing the impact of HM on major CV events including stroke will provide clarification. Previous RCTs have already pointed towards a favourable effect on stroke risk, although they were not powered for this endpoint,^2,12 A Monte Carlo simulation by Ricci et al.²³ also showed a reduction in risk of stroke by HM. This early evidence warranted considering a stroke risk reduction in our model as part of the sensitivity analysis, with cost savings by HM of 580 GBPs per patient (see Figure 2).

The main progress of the economic model presented here is that it captures the clinical, technical, and economic consequences of using HM to monitor CIEDs, and details the balance of costs and consequences. For the first time, the impact of HM on the need for device replacement has been considered. By taking a long-term perspective, our model extends and translates the available clinical evidence and applies it to the UK NHS-insured population.

Study limitations

In the model, every CV event was considered to be treated individually, whereas in reality multiple (successive) events might have occurred and been treated during one hospitalization. Such possibility could not be explored further due to lack of information on the relative timing of the individual events. Another limitation of the input data might be the constant probability of lead issues, whereas in reality this probability might increase with time.²⁴

Our model focused on economic aspects, and did not take into account patient-perceived preferences for different FU strategies and outcomes.

Considering that almost all clinical evidence in this model comes from randomised controlled trials for the BIOTRONIK HM system, this evaluation could be considered applicable to HM only. For example, a recent report with different remote monitoring systems showed that 49.2% of all scheduled remote transmissions were not received due to patient non-compliance, requiring considerable resources to contact these patients by phone.²⁵ By contrast, 93% of the daily HM transmissions in the TRUST trial were successfully received.¹⁹ Also, data reported from RCTs may not reflect outcome in a routine clinical setting.

Possible additional fields of cost savings by Home Monitoring®

Reimbursement decision makers might disregard inappropriate shocks as clinically relevant outcome (and focus predominantly on ‘hard’ outcomes such as CV events including mortality). However, apart from affecting device longevity, the psychological consequences resulting from inappropriate shocks may result in increased use of medical resources.

Our economic analysis may be conservative due to the fact that we did not consider any reduction in duration of hospital stay with HM (an 18% reduction was observed in the remote monitoring arm of the CONNECT trial).²⁶ The NHS savings presented here may also be underestimated as we did not include transportation costs (which are often subsidized by the NHS), and also did not consider improvements in efficiency of the device clinic resulting from HM. Likewise, we did not take into account the possible increase in device longevity resulting from delaying box change in patients on HM (in whom automatic alerts are available to warn of low battery charge), as well as an increase in FU frequency in patients with CFU in whom the battery is reaching the elective replacement indicator.

In conclusion, the presented model establishes HM as an economically viable technology when applied within the UK NHS system.

Supplementary material

Supplementary material is available at Europace online.

Funding

This work was supported by Biotronik. H.B was supported in part by a grant from la Tour Foundation for Cardiovascular Research.

Conflict of interest: H.B. has received consulting fees and speaker honoraria from Biotronik, Boston Scientific, Medtronic, Sorin and St Jude Medical. C.S. has received consulting fees and speaker honoraria from Biotronik, Boston Scientific, Medtronic, Sorin, and St Jude Medical. J.W. has given educational lectures on behalf of Biotronik, Boston Scientific, St Jude Medical, and Medtronic, and received travel and accommodation from all four companies at international conferences. A.S. is an employee of BIOTRONIK SE & Co KG, Berlin, Germany. K.M. and D.T. have received consultancy honoraria from BIOTRONIK SE & Co KG, Berlin, Germany.

References

Author notes