Outcomes and costs of remote patient monitoring among patients with implanted cardiac defibrillators: An economic model based on the PREDICT RM database

Abstract Background Remote monitoring of implantable cardioverter‐defibrillators has been associated with reduced rates of all‐cause rehospitalizations and mortality among device recipients, but long‐term economic benefits have not been studied. Methods and Results An economic model was developed using the PREDICT RM database comparing outcomes with and without remote monitoring. The database included patients ages 65 to 89 who received a Boston Scientific device from 2006 to 2010. Parametric survival equations were derived for rehospitalization and mortality to predict outcomes over a maximum time horizon of 25 years. The analysis assessed rehospitalization, mortality, and the cost‐effectiveness (expressed as the incremental cost per quality‐adjusted life year) of remote monitoring versus no remote monitoring. Remote monitoring was associated with reduced mortality; average life expectancy and average quality‐adjusted life years increased by 0.77 years and 0.64, respectively (6.85 life years and 5.65 quality‐adjusted life years). When expressed per patient‐year, remote monitoring patients had fewer subsequent rehospitalizations (by 0.08 per patient‐year) and lower hospitalization costs (by $554 per patient year). With longer life expectancies, remote monitoring patients experienced an average of 0.64 additional subsequent rehospitalizations with increased average lifetime hospitalization costs of $2784. Total costs of outpatient and physician claims were higher with remote monitoring ($47 515 vs $42 792), but average per patient‐year costs were lower ($6232 vs $6244). The base‐case incremental cost‐effectiveness ratio was $10 752 per quality‐adjusted life year, making remote monitoring high‐value care. Conclusion Remote monitoring is a cost‐effective approach for the lifetime management of patients with implantable cardioverter‐defibrillators.


| INTRODUCTION
The survival of patients at high risk of sudden cardiac arrest can be improved with the use of implantable cardioverter-defibrillators (ICDs). 1 The long-term mortality and morbidity of patients who receive ICDs remain substantial, however. In addition to the physician visits needed to manage disease-related morbidity, current guidelines recommend that patients with ICDs should be evaluated every 3 to 6 months to assess device function. 2 This regimen can impose a considerable burden on both patients and physicians if patients must be evaluated in the office. As a consequence, device follow-up is not reliable in routine clinical practice, with nearly one-quarter of patients not seen in-person within a year of device implantation. 3 Remote patient monitoring (RPM) has been promoted as a strategy to reduce this burden. It can improve the efficiency of care delivery by replacing at least some in-office visits with remote monitoring transmissions 4-7 without compromising safety. [7][8][9] Remote monitoring may also improve patient satisfaction and quality of life as it entails less travel time, time off work, and interruption of patient activities. Data suggests that clinically actionable events are detected sooner with remote monitoring than with standard in-office follow-up, 10 potentially allowing clinicians to act on these issues before they cause increased morbidity, hospitalizations, and costs. RPM also provides a convenient means for regular assessment of device-related parameters, such as lead impedance and battery status, which may allow early detection of a device and lead malfunction. [11][12][13][14][15] RPM can, therefore, enhance device safety and potentially improve clinical outcomes. 10,[16][17][18][19][20] RPM was associated with lower hospitalization rates and reduced mortality in the large, real-world PREDICT RM study, 21 and its routine use has been endorsed by professional societies. 22 However, while RPM is widely available, it is still not universally utilized by clinicians. In a recent U.S. study, fewer than half of ICD recipients enrolled in and activated RPM, 21 and utilization is significantly lower in Europe. To determine whether RPM has economic benefits in addition to the associated clinical benefits and to determine the magnitude of the health and economic incentives for increased use of RPM, we developed an economic model to conduct an analysis of the clinical outcomes and costs of RPM versus no RPM from a Medicare perspective. Previous studies done over limited time horizons have shown RPM to be relatively cost-effective, 23

| Patient population
The patient population for this study was composed of Medicare patients with RPM-capable devices (N = 15 254; control = 9906; RPM = 5348) taken from the population studied in the PREDICT RM database. Simulated individual-patient profiles were created based on the categorical distributions of patient characteristics in the PREDICT RM database (Table 1). To reflect the heterogeneity of the real patient population and to better preserve correlations among patient characteristics, the patient population was stratified into subgroups based on the predicted times to rehospitalization and death-the key outcomes of interest. Details of the risk stratification (eTable 1) and patient characteristics for the risk-stratified subgroups (eTable 2) can be found in the Appendix ("Patient Characteristics by Risk Strata").

| Model inputs: Clinical
Parametric survival equations

Regression analysis of predictors
The survival fitting yielded distributions for each time-to-event curve for the overall population; the fits were then adjusted to account for individual patient characteristics. Times to first rehospitalization and death were modified based on patient characteristics and, for mortality, history of the first rehospitalization. Model building with predictors is described in detail in the Appendix ("Model Building with Predictors").

Rates of subsequent rehospitalizations
The model accounts for rehospitalization events subsequent to a patient's first rehospitalization using a constant rate. Counts of second and additional rehospitalizations were divided by the duration of follow-up (counting from the time of the first rehospitalization) to give subsequent rehospitalization rates by treatment arm (RPM vs no RPM).

Rates of outpatient claims
The model accounts for three classes of outpatient claims: hospital outpatient claims, ambulatory surgical center (ASC) claims, and physician claims. ASC claims constituted less than 0.6% of the total, however, and so were combined with hospital outpatient claims.
For each type of outpatient claim, rates were calculated separately before and after the first rehospitalization as the number of unique claims per patient divided by the appropriate average time-time to first rehospitalization or time from the first rehospitalization until death.

Utilities
Utility values for the patients are dependent on a patient's baseline characteristics, comorbidities, and rehospitalization. 30 Patients were assumed to have a utility of 0 for their assigned length of stay during rehospitalization events. The effects of both patient characteristics F I G U R E 1 Overview of the model development process (outpatient claims not shown). The figure illustrates the data sources and inputs used in the economic model, and how the final inputs were derived. Boxes next to or over arrows describe the process completed to map PREDICT RM data to model inputs

| Model validation
To verify that the model would reproduce the observed results upon which the model inputs were based, we simulated two cohorts of patients, one with RPM and one without, and compared the model outputs to the observed data from the PREDICT RM database.   The rates of hospital outpatient/ASC claims and physician claims were consistently higher in the RPM arm than they were in the no-RPM arm. This was true before and after the first rehospitalization, and it was true for the overall population as well as for most of the risk-stratified bins of patients (eTable 7).

| Event rates
We also investigated whether the number of outpatient claims varied by patient risk. We assigned a risk rank to each risk stratum based on the combination of baseline risks of rehospitalization and death (see Appendix (eTable 1)). Plotting the mean outpatient claim rates by risk ranking showed a clear trend toward lower outpatient claim rates in the lower-risk groups (Figure 2).

| Cost-effectiveness analysis: Base case
Base-case results are presented in

| Cost-effectiveness analysis: Scenario and sensitivity analyses
Key scenarios in the analysis were those that removed specific RPM effects from the RPM-arm projections (  2) Although total costs and the number of rehospitalizations are increased due to improved survival, the number of rehospitalizations and overall costs PPY decrease with RPM.
3) With only a direct effect of RPM on the hospitalization rate, RPM becomes a cost-saving strategy that still provides health benefits above that of no-RPM.

5)
These results were robust to various sensitivity and scenario analyses.
Previous cost-effectiveness analyses from small randomized studies have shown RPM to be cost saving or neutral compared with conventional in-office follow-up. 23,27,35,36 Nonhospital costs were generally lower with RPM due to fewer scheduled office visits in the RPM arm (as defined by the protocol). In these studies, the number of unscheduled visits was higher with RPM possibly related to increased detection of arrhythmias and device malfunctions. However, the total number of scheduled plus unscheduled visits was still reduced. In our study, RPM was cost-effective despite an increase in the rate of outpatient claims. In addition to lower visit rates in these previous studies, inpatient costs were also reduced due to fewer hospitalizations and shorter lengths of stays. 35 In addition, device cost savings were seen in the ECOST trial 37 due to improved battery longevity. Had such cost savings been included in our analysis, RPM would have been even more cost-effective.
Clinical equipoise remains regarding the effect of remote monitoring on mortality. Retrospective analysis of two large independent databases 19,38 both showed an association between RPM utilization and improved survival, with a graded benefit related to the level of adherence to RPM. 38

ACKNOWLEDGMENTS
The views expressed in this presentation represent those of the author(s), and do not necessarily represent the official views of the NCDR or its associated professional societies identified at CVQuality.ACC.org/NCDR.
For more information, go to CVQuality.ACC.org/NCDR or email ncdrresearch@acc.org.

AUTHOR CONTRIBUTIONS
JPH, JPC, and JGA contributed to design of the study; analysis and interpretation of the data; and writing of the manuscript. HB and LSO contributed to design and conduct of study; management, analysis, and interpretation, of the data, and review of the manuscript. RJL, KAD, ARK, and SS contributed to design and conduct of study; the development of the economic model; management, analysis, and interpretation of the data; and writing and review of the manuscript. SLA, PWJ and KS contributed to the design and conduct of the study; the collection, management, analysis, and interpretation of the data; and the review and approval of the article.

FUNDING INFORMATION
This study was sponsored by Boston Scientific Corporation.