Long-term impact of intrathoracic impedance findings on survival and heart failure hospitalizations after cardiac resynchronization therapy in ICD Registry patients

Abstract

Aims

To determine adjusted associations among OptiVol^® threshold crossings, long-term survival, and hospitalizations among heart failure (HF) patients with Medicare coverage in the United States.

Methods and results

A cohort with OptiVol^®-enabled cardiac resynchronization therapy defibrillators (CRT-D) devices from the Implantable Cardioverter Defibrillator Registry was linked to both Medicare claims/summary data and Medtronic’s CareLink^® Network data. An extended multivariable Cox model was used to analyse associations among OptiVol^® threshold crossings (treated as time-dependent covariates), mortality, and HF-related hospitalizations (HFH). We analysed N = 1565 patients with OptiVol^®-enabled CRT-D devices (mean age 72.8, 28% women). The median follow-up was 6.3 years. Patients with >15.1% of days above OptiVol^® threshold (highest quartile) had more than a 4-fold increase in mortality [hazard ratio (HR) 4.2, 95% confidence interval (CI): 3.3–5.3] and more than a 3-fold increase in HFH (HR 3.2, 95% CI: 2.4–4.2) compared with patients having <4.1% of days above threshold (lowest quartile) after adjustment for key covariates. In addition, a single OptiVol^® crossing was associated with significantly increased rates of both mortality (HR 1.87, 95% CI: 1.27–2.75) and HFH (HR 1.70, 95% CI: 1.28–2.27).

Conclusion

In a CRT-D cohort with over 6 years of follow-up, both single OptiVol^® crossings and time above OptiVol^® threshold were associated with increased rates of mortality and hospitalization, which has important implications for clinical care. This is the first study integrating device data with Medicare outcomes to validate the long-term significance of OptiVol^® findings.

What’s new?

• Our study integrates CRT device-measured data, registry data, and administrative claims data to examine the potential utility of using a well-known device algorithm to monitor heart failure patients.

• A single OptiVol^® threshold crossing as well as time above threshold are strong predictors of overall patient survival and HF-related hospitalization after adjustment for key covariates in a large Medicare cohort of CRT-D patients. These findings represent the first time that such OptiVol^® threshold crossings have been associated with long-term clinical outcomes after adjustment for a complete range of covariates.

Introduction

Treatment and management of heart failure (HF) in the United States remains a major public health burden, with 5.7 million patients suffering from the disease, many of them facing a poor prognosis. These patients are hospitalized over 1 million times a year, and are often hospitalized on multiple occasions throughout their treatment.¹ Therefore, predicting death and future hospitalizations in these patients is a highly desirable goal. One strategy to do so, intrathoracic impedance monitoring, has been shown to be safe and effective for patients with cardiac resynchronization therapy defibrillators (CRT-D) in clinical trials.^2–8 While these trials have shown some predictive power of future adverse events when a pre-programmed threshold value is reached, limitations have included a lack of adjustment for patient characteristics and comorbid conditions, shorter follow-up times, the need for external validation in a real-world cohort, and inclusion of threshold crossings occurring only during the first 6 months of follow-up.⁹

In the current era, the use of intrathoracic impedance monitoring in the clinical setting remains controversial. There remains debate regarding not only which specific clinical actions should be taken in response to device readings in order to avoid adverse outcomes but also whether device-measured information should be used alone or in combination with an in-person clinical evaluation.¹⁰ Another method for monitoring HF patients, pulmonary artery pressure monitoring, has shown some promise in preventing decompensation-related hospitalization when guided by a specific management strategy,¹¹ but requires implantation of additional hardware. Additional long-term data demonstrating the independent impact of thoracic impedance monitoring on clinical outcomes in a real-world cohort would be of great interest with respect to the value of this strategy. In order to address this unmet need, we examined the impact of OptiVol^® threshold crossings and the time above threshold on long-term survival and HF-related hospitalizations (HFH) after adjusting for key covariates in a large population of patients with Medicare coverage with over 6 years of follow-up.

Methods

Study population

For this analysis, we analysed patients from the Medicare Implantable Cardioverter-Defibrillator (ICD) Registry with an OptiVol^®-enabled CRT-D device. The Center for Medicare and Medicaid Services (CMS) requires providers who implant ICDs, including CRT-Ds, to record various patient demographic, clinical, and device information. For the period January 2005 to April 2006, the registry was maintained by the Iowa Foundation for Medical Care and is available for download from CMS (see https://www.cms.gov/Medicare/Medicare-General-Information/MedicareApprovedFacilitie/Downloads/icdregistry1.pdf for more information). These registry data were then linked with both CMS claims data to ascertain time from device implant to hospitalization and death¹² and Medicare summary data to determine Medicare enrolment and type of coverage.

We then linked these Medicare data with CareLink^® data regarding patient monitoring. The CareLink^® network was established in 2002 as a de-identified database containing longitudinal device-programmed and device-measured information on implantable cardiac monitors, implantable pulse generators, ICDs, and CRT-D devices manufactured by Medtronic in the United States, and is still ongoing. To align with the timeframe of available Medicare data, the CareLink^® database was queried to find all CRT-D devices equipped with the OptiVol^® feature implanted in the United States between 1 January 2005 and 30 April 2006. Some devices equipped with the OptiVol^® feature, however, did not actually have the feature ‘turned-on’ in the device. Information from devices with actual CareLink^® transmissions was then linked with the Medicare data on the basis of device implant date (±2 days), patient age at implant (±1 year), patient sex, device model, and de-identified (3-digit) ZIP code. Patient mortality and hospitalization follow-up information was available through December 2011. A pictorial representation of how these data were merged is shown in Figure 1. Patients who had a CRT-D implanted, but were outside the established indication guideline criteria were excluded from the Medicare dataset.¹³ In addition to the variables used for linking the two databases, the time until a device experienced an OptiVol^® threshold crossing and the time above threshold were also taken from the CareLink^® registry. Patients were followed up until end of Medicare enrolment or time of the last CareLink^® record if they did not experience any threshold crossing. A total of 1565 patients were matched between the Medicare data and the CareLink^® registry.

Figure 1.

Cohort selection and data merging protocol—the process of cohort selection is shown in the flow chart. CRT-D, cardiac resynchronization therapy with defibrillator; ICD, implantable cardioverter-defibrillator; MDT, medtronic.

Endpoint definition

Heart failure-related hospitalization in the Medicare dataset was determined by an inpatient admission with a primary HF diagnosis code. For mortality endpoints, the available data included all-cause mortality rather than the cause of death.

Exposure definition

The CareLink^® database was queried to determine when an OptiVol^® threshold was exceeded. The CareLink^® database contains information on the OptiVol^® threshold setting and the accumulated fluid index value. These values were examined on a daily basis, and the first instance of the accumulated fluid index value being greater than the programmed OptiVol^® threshold value was considered to be the time of threshold crossing. Although the threshold value can be programmed to various values, the time until the first crossing was defined when the OptiVol^® fluid index value exceeded the programmed value, independently of the threshold value setting.

Covariates

Other variables available in the Medicare registry that were considered in this analysis included patient age, duration of HF, LVEF, QRS duration, systolic blood pressure, diastolic blood pressure, heart rate, sex, New York Heart Association class, presence of ischaemic cardiomyopathy, prior coronary artery bypass graft, atrial fibrillation (AF), anticoagulation therapy for AF, presence of ventricular tachycardia, prior sudden cardiac arrest event, diabetes mellitus, prior myocardial infarction, chronic kidney disease, end-stage renal disease, smoking status, and the prescription of beta-blockers, ACE-inhibitors or angiotensin receptor blockers, digoxin, diuretics, amiodarone, and/or Coumadin. The diagnosis of chronic kidney disease (CKD) and end-stage renal disease was established based on administrative diagnosis codes from inpatient and outpatient encounters, available in CMS utilization files from the year the device was implanted.

While the main exposure of interest was time until first threshold crossing, a patient may experience many such events over their follow-up time, or may have extended periods of time above the threshold, indicating a chronic congestive condition. To capture this phenomenon, we counted the number of days an individual patient spent with a fluid index value above the threshold in CareLink^® prior to the first clinical endpoint. Days above threshold could be caused by either multiple crossings or by extended periods of time for a single crossing, or both. We then calculated the percentage of their entire follow-up time spent in this state, and patients were categorized into quartiles based on this parameter.

Statistical analysis

Extended Cox models were run using PROC PHREG in SAS software, version 9.4 (SAS Institute, Cary, NC, USA) to account for the time-dependent nature of the exposure variable of interest, time to OptiVol^® threshold crossing. With increasing follow-up time, patients who experience a threshold crossing move from the ‘no crossing’ group to the ‘crossing’ group. The extended Cox method accounts for this change in exposure status, adjusting the exposure group sample sizes appropriately at the time of each event.¹⁴ Analyses for hospitalization endpoints were run both with cause-specific models, where all competing outcomes other than hospitalization (namely, patient death) were censored, as well as competing risk (or subdistribution) models, where patients who died were left in the denominator to estimate actual observable patient hospitalization rates.¹⁵ Models were run first as a minimally adjusted analysis, only including patient age at implant and patient sex. Where HFH outcomes were examined, we also adjusted for patient health maintenance organization (HMO)/managed care organization (MCO) coverage as well. Fully adjusted models were then also developed, based on all of the available Medicare ICD Registry variables. Because data on hospitalizations from patients enrolled in HMOs or MCOs can be incomplete, we performed a sensitivity analysis in which we excluded patients with any indication of HMO or MCO coverage, as determined from Medicare annual summary data. We further explored whether either patient race or sex would modify the association observed between OptiVol^® threshold crossing and mortality or HFH by adding interaction terms to each of these models.

For Kaplan–Meier plots based on threshold crossings, we applied the method of Simon and Makuch to account for the time-dependent nature of threshold crossings. As per this method, the data for each individual were split to account for time a patient spent in the ‘no crossing’ group before a threshold crossing occurred. An individual patient’s data therefore show up in both groups, contributing their associated amount of time in each group before an outcome event or censoring occurs.^16,17

Lastly, we determined the number of days each patient spent above the programmed threshold value, if any, from CareLink^®, and calculated the percentage of their follow-up time each individual patient spent above threshold. This could have come in the form of a single, prolonged crossing event, or through several acute crossing occurrences. Patients were then grouped into quartiles (<4.1%, 4.1–8.3%, 8.3–15.1%, >15.1%) and those in the highest quartile (>15.1% of days above threshold) were compared with those in the lowest quartile.

Results

Table 1 shows the subset of patients analysed in this study compared with those in the Medicare dataset who were not examined (those without Medtronic devices, or those who could not be matched in the database). Patient characteristics were similar in those patients who were analysed vs. those not analysed, although some statistically significant differences were obtained because of the large number of patients studied.

Table 1.

Baseline demographics for Medicare CRT patients by analysis cohort

	All Medicare CRT patients(n = 14 935)	Analysis cohort (n = 1565)	Non-analysed cohort (n = 13 370)	P-value*
Age, mean ± SD, years	73.0 ± 10.5	72.8 ± 8.1	73.1 ± 10.7	0.24
Duration HF, mean ± SD, years	24.7 ± 25.4	24.1 ± 24.4	24.8 ± 25.5	0.33
LVEF, mean ± SD, %	23.1 ± 6.3	23.6 ± 6.3	23.1 ± 6.3	0.002
QRS duration, mean ± SD, ms	156.8 ± 24.9	157.8 ± 24.7	156.7 ± 24.9	0.11
SBP, mean ± SD, mm Hg	126.5 ± 22.4	126.1 ± 21.3	126.5 ± 22.5	0.52
DBP, mean ± SD, mm Hg	70.2 ± 13.7	69.9 ± 12.6	70.2 ± 13.8	0.38
Heart rate, mean ± SD, bpm	72.0 ± 18.0	71.3 ± 14.0	72.1 ± 18.4	0.03
Sex, n (%)
Female	27.3	28.2	27.2
Male	72.7	71.7	72.8	0.39
NYHA class, n (%)
I	1.2	1.1	1.2
II	11.0	10.4	11.1
III	74.2	77.0	73.8	0.04
IV	13.6	11.6	13.9
Ischaemic CM, n (%)	69.2	62.8	69.9	<0.001
Prior CABG, n (%)	42.0	39.1	42.3	0.02
BBB morphology, n (%)
LBBB	69.3	71.4	69.1
RBBB	11.0	9.7	11.1	0.13
IVCD	19.8	18.9	19.8
Atrial fibrillation, n (%)	34.7	35.2	34.6	0.63
Ventricular tachycardia, n (%)	19.6	17.0	19.9	0.01
Sudden cardiac arrest, n (%)	1.7	1.5	1.7	0.54
Diabetes mellitus, n (%)	35.7	34.9	35.8	0.49
Prior MI, n (%)	50.9	45.8	51.4	<0.001
Chronic kidney disease, n (%)	32.0	26.8	32.6	<0.001
End stage renal disease, n (%)	3.1	1.8	3.2	<0.001
Smoker status, n (%)
Never	42.3	44.2	42.1
Former	48.6	47.0	48.8	0.28
Current	9.1	8.9	9.1
Medications, n (%)
b-blocker	78.9	80.3	78.7	0.17
ACEI or ARB	74.3	77.8	73.9	<0.001
Digoxin	41.7	40.8	41.8	0.44
Diuretic	78.7	79.9	78.5	0.21
Amiodarone	13.6	10.5	13.9	<0.001
Coumadin	31.8	33.7	31.6	0.09

NYHA, New York Heart Association; CRT, cardiac resynchronization therapy; SD, standard deviation; HF, heart failure; LVEF, Left Ventricular Ejection Fraction; ACEI, Angiotensin-converting enzyme inhibitor; ARB, Angiotensin receptor blocker; LBBB, Left bundle-branch block; RBBB, Right bundle-branch block; IVCD, Intra-ventricular conduction delay; MI, Myocardial infarction; CM, Cardiomyopathy; CABG, Coronary artery bypass graft; SBP, Systolic blood pressure; DBP, Diastolic blood pressure

^*P-value comparing analysed cohort to non-analysed cohort.

In our retrospective cohort study, we linked 1565 patients between the CareLink^® database, Medicare ICD registry, and Medicare claims data. Over the median 6.3-year follow-up period, we observed a mortality rate of 8.8 deaths per 100 person-years (706 deaths/8037 person-years), and a HFH rate of 9.2 hospitalizations per 100 person-years (608 cases/6581 person-years). There were a total of 1514 patients (97%) who experienced an OptiVol^® threshold crossing event during follow-up. Median time to threshold crossing was 10.5 months. A threshold crossing value of 60 Ohm-days was programmed in 99.2% of devices.

As shown in Table 2, an OptiVol^® threshold crossing at any point in the life of the device was associated with an 87% increase in patient mortality [hazard ratio (HR) 1.87, 95% confidence interval (CI) 1.27–2.75] after adjusting for the presence of patient sex, patient age at implant, ischaemic cardiomyopathy, CKD, smoking status, and the prescription of digoxin, compared with those patients who have not had a threshold crossing by that same time. (HRs for all statistically significant variables are given in the Supplementary material online.) We examined the effect of race in this analysis, but it was found not to be a significant predictor of mortality and was therefore dropped from the models. Figure 2 shows that unadjusted Kaplan–Meier survival probabilities comparing mortality between the ‘Threshold Crossing’ group to the ‘No Crossing’ group were statistically significant (log-rank test, P < 0.0001). Of note, individual patients contributed follow-up time to both curves if they experienced a threshold crossing: the time until threshold crossing was accounted for in the ‘no-crossing’ group, and any time after such a crossing was accounted in the ‘crossing’ group until either patient death or censoring occurred.

Table 2.

Associations between OptiVol^® crossing status and patient mortality and hospitalization

	Minimally adjusteda cause-specific HR (95% CI)	Minimally adjusteda subdistribution HR (95% CI)	Fully adjustedb cause-specific HR (95% CI)	Fully adjustedb subdistribution HR (95% CI)
Mortality	2.03 (1.38–3.00)	N/A	1.87 (1.27–2.75)	N/A
HF-related hospitalization (including HMO/MCO patients)	1.84 (1.38–2.45)	1.86 (1.37–2.52)	1.70 (1.28–2.27)	1.75 (1.30–2.37)
HF-related hospitalization (excluding HMO/MCO patients)	1.75 (1.31–2.35)	1.77 (1.30–2.42)	1.63 (1.22–2.19)	1.68 (1.24–2.29)

HMO, health maintenance organization; MCO, managed care organization; HR, hazard ratio; CI, confidence interval.

^aAdjusted for patient age at implant, sex, and for hospitalization events, any HMO/MCO coverage (Y/N).

^bAdjusted for patient age at implant, sex, ischaemic cardiomyopathy, chronic kidney disease, digoxin, smoking status, and for hospitalization events, any HMO/MCO coverage (Y/N).

Figure 2.

Patient mortality by threshold crossing status—Kaplan–Meier plot showing mortality-free survival via the method of Simon and Makuch. Patients with threshold crossings contribute follow-up time to both curves to account for the time-dependent nature of threshold crossings.

For hospitalization outcomes, the extended Cox model showed that an OptiVol^® threshold crossing was associated with a 70% higher rate of HFH (HR 1.70, 95% CI 1.28–2.27) with adjustment for these same covariates, including whether patients had private health coverage through an HMO or MCO organization, which was added as an additional variable in the model. In the Kaplan–Meier plot shown in Figure 3, the log-rank statistic comparing the ‘Threshold Crossing’ group to the ‘No Crossing’ group was again statistically significant (P < 0.0001). When excluding the HMO/MCO covered patients from the dataset, the association calculated by the extended Cox model was slightly attenuated (HR 1.63, 95% CI: 1.22–2.19). We again examined the effect of race, but it was not found to be a significant predictor of HFH, and was therefore dropped from the models.

Figure 3.

Patient hospitalization for heart failure by threshold crossing status—Kaplan–Meier plot showing heart failure hospitalization-free survival via the method of Simon and Makuch. Patients with threshold crossings contribute follow-up time to both curves to account for the time-dependent nature of threshold crossings.

We further explored whether there was any modification of the association between OptiVol^® threshold crossing and the outcomes by either race or patient sex. For mortality outcomes, we found no significant interactions between threshold crossing status and either sex (P = 0.49) or race when its main effect was left in the model (P = 0.99). In contrast, the patient’s sex significantly modified the association of threshold crossing with HFH (P = 0.01–0.03) based on a stronger association between OptiVol^® status and HFH in men (regardless of whether HMO/MCO patients were included or excluded, or the survival model used). There was not a significant interaction for race (P = 0.90) for hospitalization outcomes when leaving the main effect in the model. The results stratified by sex are presented in Table 3.

Table 3.

Interaction between OptiVol^® crossing status and patient sex

	Fully adjusteda cause-specific HR (95% CI)	Fully adjusteda overall interaction P-value	Fully adjusteda subdistribution HR (95% CI)	Fully adjusteda overall interaction P-value
Mortality
Male	1.76 (1.16–2.66)	0.49	N/A	N/A
Female	2.46 (0.99–6.10)
HF-related hospitalization (including HMO/MCO patients)
Male	1.92 (1.42–2.61)	0.02	1.96 (1.42–2.70)	0.03
Female	1.18 (0.78–1.77)		1.26 (0.83–1.90)
HF-related hospitalization (excluding HMO/MCO patients)
Male	1.86 (1.36–2.54)	0.01	1.89 (1.36–2.62)	0.03
Female	1.12 (0.74–1.69)		1.20 (0.79–1.82)

HMO, health maintenance organization; MCO, managed care organization; HR, hazard ratio; CI, confidence interval.

^aAdjusted for patient age at implant, sex, ischaemic cardiomyopathy, chronic kidney disease, digoxin, smoking status, and for hospitalization events, any HMO/MCO coverage (Y/N).

Since patients are at risk of experiencing multiple threshold crossings over their follow-up time, we also examined the effect of each individual patient’s time above the pre-programmed threshold as a percentage of their total follow-up time. These percentages were categorized into quartiles, and their survival plots are shown in Figures 4 and 5, showing significant differences across the quartiles (log-rank test, P < 0.0001). In a multivariable Cox model with the same covariates described above, but removing the threshold crossing variable, the quartile of percent follow-up time above threshold was statistically significant for both mortality and HFH outcomes (both P < 0.0001). In this model, patients with > 15.1% of follow-up days above the OptiVol^® threshold (highest quartile) had more than a 4-fold increased rate of mortality (HR 4.2, 95% CI: 3.3–5.3) and more than a 3-fold increased rate of HFH (HR 3.2, 95% CI: 2.4–4.2) compared with those patients in the lowest quartile (<4.1% of days above threshold).

Figure 4.

Patient mortality by quartile of percent follow-up time above OptiVol^® threshold—Kaplan–Meier plot showing mortality-free survival. Patients were grouped into quartiles depending on the percentage of their follow-up time they spend above the programmed OptiVol^® fluid index threshold.

Figure 5.

Patient hospitalization for heart failure by quartile of percent follow-up time above OptiVol^® threshold—Kaplan–Meier plot showing heart failure hospitalization-free survival. Patients were grouped into quartiles depending on the percentage of their follow-up time they spend above the programmed OptiVol^® fluid index threshold.

Discussion

In this large cohort of patients with CRT devices linked to Medicare data followed for a median of over 6 years, patients who had more than 15.1% of follow-up days above threshold (representing 25% of the patients in the entire cohort) had more than a 4-fold increased rate of mortality and more than a 3-fold rate of HFH after adjustment during more than 6 years of follow-up compared with the 25% of patients who had <4.1% of days above threshold. In addition, a single OptiVol^® threshold crossing was associated with significantly increased rates of both patient mortality (87%) and HFH (70%). These results were robust with respect to the statistical analysis used and whether patients with HMO/MCO coverage were excluded. Although the impact of OptiVol^® threshold crossing had a clinically significant impact on survival in both genders, we did find that an OptiVol^® threshold crossing was more strongly associated with HFH in men than in women. Previous studies either lacked hospitalization outcome information, covariate information, or were insufficiently powered to detect this interaction. However, this observed difference between the strength of the threshold crossing-hospitalization association for men and women is important and likely warrants further study.

The association between OptiVol^® threshold crossing, indicating accumulating fluid, and worsening HF is perhaps an obvious one, but has not been validated before in a real-world setting in the context of a large cohort. Using sophisticated analytic methods, the present study offers important new findings including associations with long-term clinical outcomes. First, the Medicare data provided data on HFH in addition to mortality. Second, the Medicare registry data also provided the opportunity to adjust for significant patient covariates and comorbid conditions, which resulted only in small differences in HR estimates vs. the minimally adjusted analysis. Third, the present study is an improvement on previous work in that it considered OptiVol^® threshold crossing as a time-dependent variable. Previous publications examined only those patients with an OptiVol^® threshold crossing during the first 6 months after device implant, effectively treating a patient with a crossing 1 day after implant as contributing the exact same follow-up time to the ‘crossing’ group as a patient with a crossing 6 months after implant.⁹

Our data matching scheme linked about 20.4% of the available 7670 Medicare CRT patients. This proportion was limited in two major ways: first, few device models implanted in the time frame of the Medicare ICD registry were available with the OptiVol^® feature, which was introduced beginning in April 2005. As a result, 4035 (52.6%) devices in the Medicare registry were of a device model which had the OptiVol^® feature. Second, at the time of the Medicare ICD Registry, enrolment in the CareLink^® network was not automatic, and it is estimated that 39% of Medtronic CRT-Ds implanted in this timeframe were actually enrolled in the CareLink^® network (Medtronic data on file, 2016). We therefore estimate that approximately 1600 devices of a type with OptiVol^® available which were also enrolled in the CareLink^® Network would exist in the available Medicare data. Based on this estimate, our actual linkage (1565 patients, or 20.4% of available Medicare data) is deemed reasonable.

With respect to the clinical impact, our data show that both a single OptiVol^® crossing and time above OptiVol^® threshold are associated with markedly elevated rates of patient mortality and patient hospitalization in real-world practice. Those patients with >15.1% of days above threshold are of particular concern, as these patients have more than a 4-fold increased risk of mortality and more than a 3-fold risk of HFH after adjustment during more than 6 years of follow-up compared with the 25% of patients who have <4.1% of days above threshold. The patients, in particular, may benefit from more intensive medical management, including re-assessment of adherence to medical therapy, dietary recommendations, and CRT pacing percentage. The high predictive value of intrathoracic impedance monitoring in this cohort suggests that interventions designed to improve clinical outcomes based on intrathoracic impedance monitoring could be developed and tested in randomized clinical trials with a similar design to that of pulmonary artery pressure monitor trials.¹¹ In summary, the findings of the present study suggest that Optivol^® has the clinical predictive value necessary to improve clinical outcomes for HF patients, and that the pressing need is to develop an optimal intervention strategy based on Optivol^® findings.

Limitations

Limitations of this analysis include our inability to account for what specific clinics do with threshold crossing information. Some clinics may be more likely to actively intervene when a threshold crossing is reached, by using the information to inform management without an actual in-person evaluation, while some may be more likely to hospitalize patients. The net effect is that our HR for hospitalization after a threshold crossing may be biased downwards in clinics with aggressive outpatient management practices and biased upwards in clinics with aggressive hospitalization practices. Even so, this analysis represents ‘real-world’ usage and provides a clear, aggregate picture of the effect of a threshold crossing. In addition, the impressive findings with respect to survival reflect a hard endpoint not influenced by variations among clinics.

Of note, both the Medicare and CareLink^® databases were deidentified prior to our joining them together; however, based on the combination of five variables used to join the data sources together, we have shown that the likelihood of proper Medicare-CareLink^® linkage is excellent based on the fact that the actual sample size of our cohort matches what was expected given the constraints of OptiVol^® functionality and CareLink^® enrolment at the time.

Conclusions

In conclusion, both the occurrence of a single OptiVol^® threshold crossing and the time above threshold are very strong predictors of patient survival and HFH after adjustment for key covariates in a large Medicare cohort of CRT-D patients. This represents the first time that these OptiVol^® findings have been associated with long-term clinical outcomes including all-cause mortality and HFH after adjustment for a complete range of covariates. Gender-specific associations between OptiVol^® findings and HFH warrant further study.

Supplementary material

Supplementary material is available at Europace online.

Funding

Funding was provided by grants from Medtronic PLC (J.B.), the University of Virginia (K.B.), and the National Institutes of Health (1 R03 HL135463) (K.B.).

Conflict of interest: Mr. Brown and Dr. Warman are full-time employees of Medtronic. Dr. Alonso and Dr. Bilchick have no conflicts to disclose.