Temporal Trends in Quality of Care among ICD Recipients: Insights from the NCDR®

Abstract

Background

The ICD Registry™ was established in 2006 in part to measure quality of care in patients undergoing ICD implantation; however, whether outcomes have improved since initiation of the registry is unknown. Our objective was to examine changes over time in three quality metrics available from the registry.

Methods and Results

We performed an observational study of 367,153 patients receiving new ICD implants from 4/2006-3/2010. Three quality metrics were selected: adverse events (inhospital complications or mortality), optimal medical therapy (OMT), and cardiac resynchronization therapy (CRT). OMT was defined as prescription of beta blocker and either ACE inhibitor or ARB in eligible patients. CRT eligibility was determined by QRS ≥120ms, LVEF ≤35%, and NYHA class III/IV. Observation periods were divided into four 12-month intervals. We analyzed changes over time, and used hierarchical logistic regression to adjust for potential confounders. Adverse events decreased over time (3.7% to 2.8%, P<0.001). Among eligible patients, rates of OMT and CRT increased over time (OMT: 69.0% to 74.3%, P <0.001; CRT: 80.5% to 84.2%, P<0.001). After adjustment for potential confounders, patients were significantly less likely to experience adverse events in Year 4 compared with Year 1 (OR 0.75, 95% CI 0.71-0.79), and significantly more likely to receive OMT (OR 1.29, 95% CI 1.26-1.32) and CRT (OR 1.42, 95% CI 1.35-1.49).

Conclusions

Since initiation of the ICD Registry, adverse events are decreasing, and rates of OMT and CRT among eligible patients are increasing, although there is still significant room for improvement.

INTRODUCTION

Over the past decade, implantable cardioverter-defibrillators (ICDs) have played an increasingly important role in the care of patients at high risk of sudden cardiac death (SCD). Randomized clinical trials have demonstrated that ICDs improve survival in patients with evidence of high risk of SCD due to prior myocardial infarction and/or heart failure¹. In response to this evidence, payors expanded their coverage of these devices to include primary prevention ICD implantation^{2, 3}.

As part of this coverage expansion, the ICD Registry™ was developed by the American College of Cardiology (ACC) National Cardiovascular Data Registry (NCDR®) in partnership with the Heart Rhythm Society to collect information from U.S. hospitals about the characteristics and inhospital outcomes of patients undergoing ICD implantation⁴. Participating hospitals are required to submit data to the registry for all Medicare patients receiving primary prevention ICDs, and 79% of participating hospitals submit data for all patients receiving ICDs regardless of indication⁵. Thus, the registry represents a significant investment by hospitals, physicians, and professional societies in monitoring the care of patients undergoing ICD implantation.

A major goal of the registry is to provide feedback to sites that can be used for quality improvement efforts⁶, and studies using data from the ICD Registry have identified potential areas for improvement including procedural safety^{7, 8}, concomitant medical therapy⁹, and use of cardiac resynchronization therapy (CRT) in eligible patients¹⁰. However, it is not known whether the quality of care for patients has changed over time. Having collected data about implants since 2006, the registry provides an excellent opportunity to address this gap in knowledge by evaluating trends in the care of patients undergoing ICD implantation. To accomplish this, we evaluated three quality metrics that are known to be meaningfully associated with long-term outcomes: adverse events (procedural complications or mortality)¹¹, prescription of optimal medical therapy (OMT) among eligible patients¹², and implantation of a cardiac resynchronization therapy (CRT) device among eligible patients¹². Furthermore, we performed analyses examining whether changes in quality of care varied across hospitals and whether individual performance metrics were correlated.

METHODS

Data Source

Details of the ICD Registry have been previously described⁴. In brief, information about patients undergoing ICD implantation are collected using standardized definitions and submitted by participating hospitals to the ICD Registry via a secure website. For the purposes of training, new NCDR sites participate in a quarterly conference call where data entry practices are reviewed, and the NCDR maintains a website with a list of frequently asked questions as well as a call center where individuals and centers can receive direction about proper abstraction strategies. Participating sites may also participate in quarterly conference calls and an annual in-person meeting where NCDR leaders answer questions and review cases that represent common challenges faced by data abstractors.

Once data are submitted, they undergo quality checks to ensure data completeness, consistency, and accuracy¹³. Before entering the Enterprise Data Warehouse (EDW), submissions receive one of three color-coded scores based on data quality. Data are given a “green light” if they have passed all quality and integrity checks and are suitable for analysis. A “yellow light” means that data have passed integrity checks but are incomplete; these data are loaded into the EDW and sites are given the opportunity to resubmit their data¹³. A “red light” denotes a failed submission due to either extensive missing data, or data that are internally inconsistent. These data are neither processed nor loaded into the EDW¹³. Data accuracy is also evaluated with chart audits which constituted approximately 2% of patient records in 2010 and 4% in 2011 (Pourhamidi, F, NCDR staff, email correspondence dated July 8, 2013).

For the current study we used Version 1 data, which include procedures between 04/01/2006 and 03/31/2010. Full elements are available at the NCDR website⁶. We analyzed data for patients undergoing first-time transvenous ICD implantation and divided periods of analysis into 12-month intervals (Year 1: 04/01/2006-03/31/2007; Year 2: 04/01/2007-03/31/2008; Year 3: 04/01/2008-03/31/2009; Year 4: 04/01/2009-03/31/2010). Since the focus was on change in our metrics over time, we restricted our analysis to hospitals reporting to the ICD Registry for at least 12 out of the total of 16 quarters. If a patient had multiple procedures within a given 12-month interval, only the initial procedure was counted.

Institutional review board approval was obtained through the Yale University Human Investigation Committee, and the requirement for informed consent was waived based on the nature of the study.

Outcomes

We defined three quality metrics that were available from registry data: (1) adverse in-hospital events (including complications or death), (2) receipt of OMT, and (3) receipt of CRT among eligible patients. Using definitions similar to prior studies from the ICD Registry¹⁴, we divided complications into major (cardiac arrest, cardiac perforation, valve injury, stroke, myocardial infarction, tamponade, device infection, peripheral embolus, lead dislodgement, hemothorax, pneumothorax, arteriovenous fistula) and minor (drug reaction, hematoma, phlebitis, conduction block, peripheral nerve injury). Complications were only available for analysis if they occurred during the index hospitalization. OMT was determined according to American Heart Association/American College of Cardiology consensus guidelines for patients with heart failure and systolic dysfunction¹², and included discharge on a beta blocker and ACE inhibitor/angiotensin receptor blocker for patients with left ventricular ejection fraction (LVEF) ≤35% and no recorded contra-indications to OMT recorded by sites at the time of procedure⁹. For CRT, eligible patients were defined by LVEF ≤35%, QRS duration ≥120ms, and New York Heart Association (NYHA) Class III or IV heart failure, in accordance with consensus guidelines available during the study period¹² and as used in prior ICD Registry analyses¹⁰. We did not include patients with NYHA Class II heart failure as our period of analysis was prior to the publication of data showing a definitive benefit of CRT therapy in these patients^{15, 16}. Because new data were emerging during this time period that patients with a QRS duration of ≥150ms had a greater benefit than those with QRS 120-150 ms¹⁷, we also performed a sensitivity analysis for CRT using a QRS duration of ≥150ms rather than ≥120ms.

Statistical Analysis

Patient characteristics and outcomes were compared among different years using one-way ANOVA or non-parametric test for continuous variables where appropriate and the Pearson Chi-squared test for categorical variables. For each year of data we calculated the proportion of patients experiencing adverse events, and the proportions of eligible patients receiving OMT and CRT. Trends in outcomes for the study cohort over the four-year period of analysis were examined using the Cochran-Armitage test. To evaluate the independent effect of different time periods, we used hierarchical logistic regression models with and without adjustment for potential confounders. Odds ratios and 95% confidence intervals were reported for outcomes in year 4 versus year 1 (reference group). We then performed several exploratory hospital-level analyses for each quality metric (adverse events, OMT among eligible, CRT among eligible), excluding hospitals that performed fewer than 50 procedures in each year in order to ensure stable estimates. We first analyzed the distribution of absolute change in each quality metric among hospitals between Year 1 and Year 4. Subsequently, we examined the correlation between quality metrics among hospitals in the aggregate time period (i.e. combining data from Year 1 through Year 4). A two-tailed P value less than 0.05 was considered statistically significant for all tests. All statistical analyses were performed with the use of SAS software, version 9.2 (SAS Institute).

RESULTS

Study sample

From an initial sample of 538,670 patients, we excluded 8843 who received an epicardial lead, 142,484 who had a previous ICD, and 20,190 who underwent ICD implantation in hospitals reporting data for fewer than 12 quarters, leaving an analytic sample of 367,153 patients. Patient and hospital characteristics over time are shown in Table 1. Overall, there were only modest changes in these characteristics during the four-year study period.

Table 1.

Patient characteristics over time

Year	4/06-3/07 (N=93414)	4/07-3/08 (N=92046)	4/08-3/09 (N=92704)	4/09-3/10 (N=88989)	P for Trend
Demographics
Age (mean ± SD)	67.7 ± 12.8	67.5 ± 13.1	67.1 ± 13.2	66.9 ± 13.3	<0.01
Female (%)	26.7	27.1	27.2	27.7	<0.01
Nonwhite race (%)	18.0	18.7	19.6	20.0	<0.01
Medical history
Cerebrovascular disease (%)	14.4	14.4	14.2	14.3	0.32
Chronic lung disease (%)	21.7	23.1	23.3	23.0	<0.01
Diabetes (%)	36.6	37.3	37.5	37.8	<0.01
Hypertension (%)	73.3	76.0	77.5	78.7	<0.01
Renal failure – dialysis (%)	4.2	4.2	4.2	4.0	<0.05
Congestive heart failure (%)	77.8	77.2	77.3	77.1	<0.01
Previous myocardial infarction (%)	54.7	53.0	51.0	49.0	<0.01
NYHA Class III/IV (%)	51.6	51.4	51.2	51.4	<0.01
Sustained ventricular tachycardia (%)	10.7	11.3	11.9	12.2	<0.01
Prior pacemaker (%)	11.3	10.8	10.4	10.1	<0.01
Diagnostics
%EF, when assessed (mean ± SD)	27.0 ± 10.4	27.7 ± 10.8	28.1 ± 11.1	28.5 ± 11.4	<0.01
Creatinine level (mean ± SD)	1.40 ± 1.18	1.39 ± 1.15	1.36 ± 1.14	1.35 ± 1.14	<0.01
Sodium level (mean ± SD)	138.5 ± 3.6	138.5 ± 3.5	138.4 ± 3.5	138.4 ± 3.4	<0.01
Systolic blood pressure (mean ± SD)	129.8 ± 12.9	130.2 ± 22.3	130.5 ± 22.3	130.8 ± 22.5	<0.01
QRS width, milliseconds (mean ± SD)	125.7 ± 34.2	124.8 ± 34.0	124.3 ± 33.5	124.4 ± 33.3	<0.01
Hospital characteristics
Private/community (%)	84.7	84.4	84.0	83.3	<0.01
Teaching (%)	55.5	54.7	54.5	54.3	<0.01
Number of beds (mean ± SD)	489.8 ± 259.7	479.9 ± 253.4	479.8 ± 252.3	479.8 ± 253.2	<0.01
Hospital volume (mean ± SD)	599.2 ± 386.8	589.1 ± 383.3	585.0 ± 384.6	585.8 ± 383.9	<0.01

Unadjusted Outcomes

There were significant declines in adverse events over the study period (Table 2), which decreased from 3.7% to 2.8% overall (trend P<0.001). Major complications decreased from 2.3% to 2.0% (P for trend <0.001), and minor complications decreased from 1.4% to 0.8% (P for trend <0.001). In-hospital death declined only slightly from 0.4% to 0.3% (P for trend <0.05). The proportion of eligible patients receiving OMT steadily increased from 55,316/80,120 (69.0%) in Year 1 to 53,805/72,416 (74.3%) in Year 4 (P for trend <0.001). Similarly, the proportion of patients receiving CRT among those eligible increased from 23,897/29,687 (80.5%) in Year 1 to 23,040/27,367 (84.2%) in Year 4 (P for trend <0.001). Using a QRS cutoff of 150ms instead of 120ms for CRT eligibility, this trend was similar: 14,183/16,579 (85.5%) in Year 1 and 13,384/15,186 (88.1%) in Year 4 (P for trend <0.001).

Table 2.

Overall trends in quality of care metrics (unadjusted)

Year	4/06-3/07	4/07-3/08	4/08-3/09	4/09-3/10	P for trend
Adverse events	3.7	3.4	3.1	2.8	<0.01
Major complication* (%)	2.3	2.3	2.1	2.0	<0.01
Minor complication† (%)	1.4	1.0	0.9	0.8	<0.01
In-hospital death (%)	0.4	0.5	0.4	0.3	<0.05
CRT among eligible (%)	80.5	81.1	82.0	84.2	<0.01
Optimal medical therapy§ (%)	69.0	71.5	73.0	74.3	<0.01

^*Cardiac arrest, cardiac perforation, valve injury, coronary venous dissection, hemothorax, Pneumothorax, deep phlebitis, transient ischemic attack, stroke, myocardial infarction, pericardial tamponade, arteriovenous fistula, device-related infection, peripheral embolus, lead dislodgement

^†Drug reaction, conduction block, hematoma, superficial phlebitis, peripheral nerve injury

^§Among patients with LVEF <=35%, prescription of beta blocker and ACE inhibitor/angiotensin receptor blocker at discharge, unless contraindication to therapy

Adjusted Outcomes

The unadjusted and adjusted odds ratios (Year 4 vs. Year 1) for adverse events, OMT, and CRT are shown in Table 3. After adjustment for potential confounders, patients were significantly less likely to experience adverse events in Year 4 compared with Year 1 (odds ratio [OR] 0.75, 95% confidence interval [CI] 0.71-0.79), and significantly more likely to receive OMT (OR 1.29, 95% CI 1.26-1.32) and CRT (OR 1.42, 95% CI 1.35-1.49).

Table 3.

Unadjusted and adjusted odds ratios, year 4 vs. year 1 of observation

	Unadjusted OR* (95% CI)	Adjusted OR† (95% CI)
Adverse events	0.75 (0.71-0.78)	0.75 (0.71-0.79)
OMT among eligible	1.33 (1.30-1.36)	1.29 (1.26-1.32)
CRT among eligible	1.36 (1.30-1.42)	1.42 (1.35-1.49)

^*OR = Odds ratio for Year 4 (4/09-3/10) vs. Year 1 (4/06-3/07); CI = Confidence interval

^†Multivariable model adjusted for age, sex, race, insurance payor, New York Heart Association class, cardiac arrest, atrial fibrillation, ventricular tachycardia, ischemic heart disease, prior myocardial infarction, cerebrovascular disease, chronic lung disease, diabetes, hypertension, dialysis, length of stay, indication (primary vs. secondary prevention)

C-statistics for adjusted model: 0.651 (adverse events), 0.619 (OMT among eligible), 0.714 (CRT-D)

Hospital-level Analyses

For hospitals performing over 50 procedures in each year of observation, the median rate of inhospital adverse events decreased by 0.79% (IQR −2.72% to 0.92%), the median proportion of patients receiving OMT increased by 4.83% (IQR −1.77% to +12.86%), and the median proportion of patients receiving CRT increased by 3.73% (IQR −1.70% to 9.91%). As shown in Figure 1, while the majority of hospitals improved their metrics, in some hospitals metric performance declined. There was no correlation between in-hospital adverse events and either CRT (r=0.03, P=0.46) or OMT (r=0.04, P=0.27) among eligible (Figure 2). There was a weak positive correlation between CRT and OMT among eligible (r=0.17, P<0.001).

Figure 1.

Distribution of change in quality metrics over time. For hospitals performing over 50 procedures per year, we analyzed the distribution of absolute change between Year 1 and Year 4 among three quality metrics: adverse events (complications or mortality) (top left), optimal medical therapy (OMT) among eligible (top right), and cardiac resynchronization therapy (CRT) among eligible (bottom left). The majority of hospitals had fewer adverse events in Year 4 compared with Year 1, and more eligible patients received OMT and CRT; however, as seen in the figures, performance among some hospitals worsened. The lower overall number of hospitals in the CRT histogram represents fewer registry hospitals meeting criteria for ≥50 annual CRT procedures.

Figure 2.

Correlation among quality metrics. For hospitals performing over 50 procedures per year, we analyzed correlations between three quality metrics: adverse events (complications or death), optimal medical therapy (OMT) among eligible, and cardiac resynchronization therapy (CRT) among eligible. Among these hospitals there was no significant correlation between low rates of adverse events and high proportions of either OMT (top left) (R=0.04, P=0.27) or CRT (top right) (R=0.03, P=0.46) among eligible patients. There was a weak positive correlation between rates of OMT and CRT among eligible patients (bottom left) (R=0.17, P<0.001).

DISCUSSION

In a large sample of U.S. patients undergoing new ICD implants over a four-year time period, we found a significant improvement in quality of care as evidenced by a decrease in adverse events and significant increases in the proportions of eligible patients receiving OMT and CRT. As each of these quality metrics has been independently associated with survival^{11, 12}, the observed improvements are likely to benefit both the short and long-term outcomes of new ICD recipients.

Although these findings are encouraging, there remains significant room for improvement. Among eligible patients in Year 4, 26% did not receive OMT, and 16% did not receive CRT. In addition, there was considerable variation in change for quality metrics among hospitals such that many hospitals had worse outcomes in Year 4 compared with Year 1. Although this may reflect the play of chance, it raises some concern that not all hospitals are improving at the same rate. Similarly, we found that there was very little correlation among different quality metrics (i.e. fewer in-hospital adverse events were not at all correlated with increasing rates of OMT or CRT), suggesting the factors that allow a hospital to perform well in one area may not carry over to other domains. This is consistent with prior research which has shown high correlation among process measures (for example, beta blocker and aspirin at discharge among survivors of acute myocardial infarction), but less correlation between process measures and outcomes such as mortality¹⁸. Thus, efforts to improve care of patients undergoing ICD implantation may require a broad range of approaches targeting physicians, hospital processes, and staff education. While there may be a “ceiling effect” (e.g. we would not expect complications to ever reach 0%), evidence continues to evolve (for example, recent data support continuing periprocedural oral anticoagulation to reduce the incidence of device pocket hematoma) ¹⁹, which suggests continued opportunities for improvement.

The factors responsible for the observed improvements in quality of care are unknown. The observed trends may reflect changes in care related to improved technology and operator proficiency over time (in the case of complications) and more widespread adoption of evidence-based practices (in the case of OMT and CRT). It is also possible that the feedback provided by the NCDR to sites has facilitated hospitals’ quality improvement efforts. This feedback includes quarterly outcome reports with benchmark comparison metrics, site audits, monthly site manager calls, and an annual educational conference. While engagement with this process among sites may vary, quarterly ICD registry conference calls involve between 150-200 sites, approximately 1200 sites access the ICD Registry's outcomes dashboard (where the quality reports are contained), and 921 participants attended the annual in-person meeting in 2013 (Pourhamidi, F, NCDR staff, email correspondence dated July 8. 2013). However, given the absence of an appropriate comparison group, we cannot empirically measure the impact of these efforts. Finally, we cannot exclude the possibility that the differences over time may reflect changes in participating hospitals’ data abstraction processes; however, as noted previously, the registry has a robust data quality program in place that includes chart review from selected sites to determine the accuracy of submitted data¹³ .

Our findings must be interpreted in the context of our study design. We cannot extend our results to hospitals that either did not submit data to the registry or whose data did not meet NCDR quality standards. In addition, like any large registry, we have limited information regarding clinical decision-making and why a particular therapy was not used in a specific case. For example, it is possible that for CRT, the option to implant CRT may have been offered to and declined by patients, attempted and unsuccessful, or not offered because of other considerations (e.g. imminent referral for mechanical circulatory support or transplant). For OMT, we did not include data on other therapies that may have been beneficial (such as potassium sparing diuretics) as these were not available from the registry. Furthermore, as doses are not recorded in the registry, we cannot determine if the OMT medications were truly optimized. In addition we do not know whether use of OMT medications increased due to conditions other than heart failure (for example, hypertension which increased in prevalence over time). Both the OMT and CRT metrics were process measures – we were unable to evaluate in our dataset whether increasing rates of OMT and CRT were actually associated with reduced long-term mortality – although our selection of these metrics was based on efficacy data from randomized trials¹². Furthermore, as our study is observational in nature, practice patterns may have changed over time based on emerging data prior to the revision of guidelines. Finally, we were not able to use Version 2 registry data to capture in-hospital outcomes beyond March 2010, and also do not have information on important outcomes that may have occurred following hospital discharge (such as device-related infections, or OMT at 6 months). Investigating inhospital trends beyond March 2010, as well as linking registry data with other sources that capture outcomes beyond the initial hospitalization, are worthwhile directions for future investigations.

In conclusion, since initiation of the ICD Registry in 2006, adverse events among ICD recipients have decreased, and rates of OMT and CRT among eligible patients have increased. While these trends are encouraging, there remains significant variation among hospitals, highlighting the importance of continued quality improvement efforts. Determining factors that promote improvements in quality of care is an important avenue of further research.

Clinical Summary.

Implantable cardioverter-defibrillators (ICDs) improve survival in patients at high risk of sudden cardiac death, and their use has increased over time. In conjunction with a coverage expansion in 2006 to cover primary prevention, the ICD Registry™ was developed in order to capture the characteristics and in-hospital outcomes of ICD recipients. A major goal of the registry is to provide feedback to hospitals that can be used for quality improvement efforts, and as the largest database of U.S. ICD recipients to date, it provides a unique opportunity to examine quality metrics. The purpose of our study was to examine trends in three quality metrics known to be associated with long-term outcomes in patients receiving ICDs: adverse in-hospital events (procedural complications or mortality), prescription of optimal medical therapy (OMT) as defined by receipt of ACE inhibitor and beta blocker in eligible patients at discharge, and use of cardiac resynchronization therapy in eligible patients. We found that from 4/2006-3/2010 there was significant improvement in all three quality metrics: adverse events declined (3.7% to 2.8%, P<0.0001), and use of OMT and CRT increased (OMT: 69.0% to 74.3%, P <0.001; CRT: 80.5% to 84.2%, P<0.001). After adjusting for patient characteristics in each year, these trends persisted. However, we found that among hospitals there was little correlation in good performance between the three metrics. Our results show an encouraging trend in quality metrics among U.S. ICD recipients over a four year time period, although there remains room for improvement.