IMPROVE HF: Registry to Improve the Use of Evidence-Based Heart Failure Therapies in the Outpatient Setting: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for IMPROVE HF: Registry to Improve the Use of Evidence-Based Heart Failure Therapies in the Outpatient Setting (NCT00303979), a registered phase not stated study. The primary outcome is To Evaluate Over the Aggregate IMPROVE-HF Practice Sites the Relative Changes in 7 Performance Measures at 24 Months Compared With Baseline and Determine the Number of Performance Measures That Achieved a Relative 20% or Greater Positive Change..

2. Clinical Question

IMPROVE HF was a quality-improvement initiative conducted through review of patient records in cardiology practices. Its stated purpose was to characterize management of patients with heart failure or with prior myocardial infarction and left ventricular dysfunction, and to assess the effect of practice-improvement activities on use of evidence-based heart-failure therapies.

The seven measures covered use of ACEI/ARB, beta-blockers, aldosterone receptor antagonists, anticoagulation for atrial fibrillation, cardiac resynchronization therapy, cardioverter-defibrillator therapy, and heart-failure education. The registry record describes the threshold for success as reaching a relative improvement of at least 20% in at least two of these measures. It does not report the observed changes or whether that threshold was met.

3. Trial Design

This was a registry-based quality-improvement initiative, rather than a randomized treatment trial. The record describes review of patient records and an assessment of practice-level changes associated with educational and other improvement activities. The allocation method is not stated, and the registry record does not describe a randomized control group. The primary comparison therefore uses different time points within the longitudinal cohort, rather than a concurrent randomized comparison.

The record describes three cohorts. Cohort A was longitudinal, with approximately 15,000 patients followed at baseline, 12 months, and 24 months. Cohort B involved approximately 10,000 patients reviewed at a single time point at 6 months, while Cohort C involved approximately 10,000 patients reviewed at a single time point at 18 months. The primary outcome focuses on aggregate practice performance at baseline and 24 months in Cohort A. The other cohorts appear in the descriptions of secondary comparisons; the registry record does not report their outcome results.

For the primary analysis, 167 practices contributed baseline chart-review data. Twelve practices withdrew before the next chart-review milestone, and data from 155 practices contributed to follow-up of the longitudinal cohort at 12 and 24 months after the educational workshop. These practice counts describe participation at the practice level; they should not be read as patient counts or as evidence that the analysis included the same eligible patients at every time point.

The registry lists enrollment of 34810 and a completed status. The study start is recorded as 2005-05 and primary completion as 2009-08. The record does not give a detailed schedule of intervention delivery, a full participant flow, or a statistical analysis plan beyond the methods described for the primary outcome.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Before assignment: The study included 3 mutually exclusive patient cohorts. Patients in cohort A were reviewed at baseline, 12 and 24 months. This is also called the longitudinal cohort. Patients in cohort B and C were reviewed at 6 and 18 months, respectively. These two cohorts are referred to as the single time point cohorts.

Overall Study

Analysis population for the primary outcome, as defined in the registry: 167 practices contributed data to the baseline chart review. Twelve of the practices withdrew from the study prior to the next chart review milestone. 155 of those practices' data contributed to the follow up of the longitudinal cohort time points of 12 and 24 months post educational workshop.

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

6. Primary Endpoint

The registry lists the primary outcome as To Evaluate Over the Aggregate IMPROVE-HF Practice Sites the Relative Changes in 7 Performance Measures at 24 Months Compared With Baseline and Determine the Number of Performance Measures That Achieved a Relative 20% or Greater Positive Change., measured over 24 Month.

The primary endpoint assesses relative changes in seven performance measures across the aggregate IMPROVE-HF practice sites at 24 months compared with baseline. For each measure, the record describes calculating the proportion of eligible patients who were treated at baseline and at 24 months, then expressing the change relative to the baseline proportion. The measures concern use of specified heart-failure therapies and provision of heart-failure education. Eligibility is measure-specific, so the population contributing to one measure need not be identical to the population contributing to another.

The endpoint also counts how many of the seven measures reached a relative improvement of at least 20%. The registry defines the initiative as successful if at least two measures met that threshold. This combines measure-specific changes with a threshold-based summary of whether improvement was sufficiently broad under the stated criterion. It does not, by itself, describe the size or precision of each measure’s change, or whether changes were consistent across individual practices.

These measures are used to assess documented care processes, not patient-level clinical outcomes such as survival or symptoms. The registry record specifies a 95% confidence interval and a one-sample proportion z-test for evaluating relative changes. However, it does not report the numerical estimates, confidence intervals, test results, or the number of measures meeting the threshold. Consequently, the endpoint definition is available, but its observed result cannot be interpreted from the registry record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is a collection of seven performance measures, rather than a single continuous clinical endpoint. For each measure, the registry describes calculating the proportion of eligible patients who received the relevant therapy or education at baseline and at 24 months. Eligibility is specific to each measure. Thus, the denominator for one measure may differ from the denominator for another, and a change in a measure’s eligible population could affect interpretation. The registry record does not give the measure-specific denominators or explain how eligibility was operationalized.

What does relative change mean?

The record defines relative change as the 24-month treated proportion minus the baseline treated proportion, divided by the baseline treated proportion. In general, this expresses change in relation to the starting value rather than as a simple percentage-point difference. This distinction can matter: the same absolute change can correspond to different relative changes depending on the baseline. The registry record does not provide the baseline proportions, follow-up proportions, or observed relative changes, so their magnitudes cannot be assessed.

The stated threshold uses relative improvement of at least 20%. The record says the number of measures meeting or exceeding that threshold was determined, and that the initiative was considered successful if at least two of the seven measures met it. This is a threshold rule for summarizing the set of outcomes. It does not replace examination of the individual estimates, their uncertainty, or differences across practices. The record does not report how many measures met the threshold.

What does the confidence interval convey?

A confidence interval describes statistical uncertainty around an estimate under the assumptions of the method used. A 95% confidence interval is the interval procedure specified by the registry for the relative change in each performance measure. In general, narrower intervals indicate greater precision, while wider intervals indicate more uncertainty. An interval’s position relative to a null value can also inform interpretation, but the registry record does not give the numerical intervals or state the reference value used to assess them. No conclusion about precision or statistical compatibility can therefore be drawn from the record.

Why use a one-sample proportion z-test?

The registry says each relative change was evaluated using a one-sample proportion z-test. In general, a z-test compares an observed statistic with a reference value under a specified null hypothesis, using an approximation to its sampling distribution. The registry record does not provide the full test specification, including the precise null value, test direction, or any adjustments to the calculation. Those details should not be inferred from the name of the test alone. The test result also cannot be reconstructed without the necessary outcome counts and denominators.

What is the role of aggregation?

The primary objective is stated for aggregate IMPROVE-HF practice sites. Aggregation provides a summary across participating practices, but can conceal variation: a similar overall proportion may arise from different patterns among individual practices. The registry record does not explain whether aggregation was weighted by the number of eligible patients, how practice-level differences were handled, or whether the analysis accounted for clustering of patients within practices. These are important details for evaluating uncertainty because observations from the same practice may be related. The exact analytic treatment of that dependence is not reported in the registry record.

How should planning illustrations be read?

Any sample-size illustrations associated with this analysis are planning calculations, not observed study results and not observed power. The stated assumptions are a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation; the effect to detect is described as a stated fraction of the observed effect. Because this initiative is not described as a randomized two-arm comparison, those assumptions should not be mistaken for the design or analysis of its primary endpoint. The record does not report a sample-size justification for the initiative or establish that these planning assumptions were used when it was designed.

In general, sample-size planning estimates how many participants per arm might be needed under specified assumptions for a future comparison. Such calculations depend on the assumed effect and analysis conditions. They do not show that an existing study had a particular probability of finding a result, and they do not substitute for observed estimates and confidence intervals. The registry record does not provide numerical planning outputs in the registry record for this description, so none can be interpreted.

Overall, the methods description identifies a relative-change calculation, confidence intervals, a one-sample proportion z-test, and a threshold summary across seven measures. It does not provide enough detail to independently reproduce the analyses. The statistical plan beyond those stated elements, including handling of practice clustering, missing records, multiplicity, and sensitivity analyses, is not reported in the registry record.

10. Multiplicity, Missing Data and Analysis Populations

The primary outcome contains seven performance measures, and the registry record lists five outcome measures and 31 analyses. Multiple analyses create a general multiplicity concern: when many outcomes or tests are examined, the chance of at least one apparently unusual result can increase. Whether that concern was addressed depends on the inferential purpose and the analysis plan. The registry record does not report a multiplicity adjustment, a prespecified hierarchy of tests, or a plan for interpreting results across the measures. It is therefore not possible to determine how the reported tests, if any, were intended to support conclusions across the outcome set.

The record describes a threshold requiring at least two measures to reach a relative improvement of at least 20%, but does not say whether this threshold was prespecified in a protocol or how uncertainty around the threshold-based summary was evaluated. It also does not report the observed count of measures meeting the criterion. That rule should not be confused with a statistical adjustment for testing multiple outcomes.

Missing-data handling is not reported in the registry record. The record notes that 12 practices withdrew before the next chart-review milestone and that 155 practices contributed data to the longitudinal follow-up at 12 and 24 months, following baseline data from 167 practices. It does not state how practice withdrawal, unavailable records, incomplete measures, or changes in eligibility affected denominators or estimates. The record also does not describe imputation, exclusions, or sensitivity analyses.

The primary outcome is defined among patients eligible for each measure, but the registry does not report the analysis population in enough detail to establish whether eligibility and observation were handled consistently across time points. The number of practices contributing does not establish how many patient records were analyzed for each measure. No assumptions about missingness or the handling of incomplete observations can be made from the registry record.

11. What the Result Does, and Does Not, Mean

The registry record documents the study’s objectives, cohort structure, primary endpoint definition, and selected statistical methods. It does not report the numerical primary results. As a result, it is not possible to say whether the seven performance measures increased or decreased, how large any changes were, how precise the estimates were, or whether the stated threshold of at least two measures with a relative improvement of at least 20% was reached.

Even if changes in the performance measures were available, they would describe documented care processes rather than directly establish effects on clinical outcomes. The measured indicators concern therapies and heart-failure education among eligible patients. They do not, by themselves, quantify survival, symptoms, or other patient-level outcomes. The record does not report such outcome results in the registry record for this description.

The design also limits causal interpretation. The principal comparison is between baseline and 24 months in a longitudinal cohort, with no randomized control comparison described. A before-and-after difference can reflect the initiative alongside other changes over time. The record does not provide a method for separating those influences. Practice withdrawal and unreported details of patient-record availability also leave uncertainty about the composition of the analyzed populations over time.

The threshold summary can indicate whether a specified number of measures crossed a specified relative-change criterion, but it cannot replace the individual estimates and their confidence intervals. Without those results, neither the threshold criterion nor the precision of any estimate can be assessed. The registry record also does not report how variability between practices was represented statistically.

Accordingly, this record supports understanding what the initiative set out to evaluate and how the primary endpoint was described. It does not support a conclusion that the initiative did or did not improve care, nor does it support a claim about the size, certainty, or clinical consequences of any change. Those judgments would require results and analytic details that are not reported in the registry record.

12. Why This Trial Matters Statistically

This study is statistically informative as an example of evaluating practice-level quality improvement through repeated chart review. Rather than comparing randomized treatment groups, its central question concerns whether aggregate care-process measures changed from baseline to follow-up after a package of improvement activities. That design can describe change over time, while also raising the causal and measurement questions that arise when there is no concurrent randomized control group.

The endpoint illustrates why denominators and eligibility matter. Each of the seven measures applies to patients eligible for that particular therapy or educational measure. The resulting proportions may therefore represent different patient subsets. Changes in proportions can also be difficult to interpret if eligibility, documentation, or the mix of contributing records changes over time. The registry record does not provide the detailed denominators or missing-data methods needed to assess those issues.

The planned analysis also distinguishes relative change from absolute change. Relative change scales the difference to the baseline proportion, while the threshold summary focuses on how many measures cross a specified criterion. Both choices can be useful for a defined question, but neither alone conveys the full pattern of estimates, their uncertainty, or variation across practices. Confidence intervals and measure-level results are necessary for that fuller assessment; they are not reported in the registry record.

Finally, the record’s stated use of a z-test and confidence intervals provides only part of the inferential picture. With multiple measures and repeated practice observations, interpretation also depends on multiplicity, dependence within practices, participant eligibility, and missing data. The registry record does not report how those issues were handled. These features make the study useful for teaching how to distinguish an outcome definition and statistical plan from the evidence needed to judge observed results.

13. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

The registry record leaves several design and reporting limitations specific to this initiative. The primary analysis compares aggregate performance at baseline and 24 months, but no randomized or concurrent control group is described. Allocation is not stated. Consequently, the record does not identify a design-based way to separate changes associated with the improvement activities from other changes over time.

Practice participation changed between baseline and follow-up: the record says 167 practices contributed baseline data, 12 withdrew before the next chart-review milestone, and 155 practices contributed to follow-up at 12 and 24 months. It does not explain why practices withdrew or how their absence was handled analytically. The record also does not establish that the same patient records or eligible populations contributed at each time point.

Measure-specific denominators, baseline and follow-up proportions, and numerical outcome estimates are not reported in the registry record. The primary outcome therefore cannot be checked or interpreted for magnitude, direction, or precision. Nor does the record explain the weighting used to create aggregate practice results or how dependence among patients within a practice was accounted for.

Further limitations include the absence of reported details on missing-data procedures, multiplicity, sensitivity analyses, and the full statistical analysis plan. The record describes a one-sample proportion z-test and 95% confidence intervals, but does not give enough information to reproduce the analysis. It also does not report the count of measures meeting the improvement threshold. Finally, the endpoint assesses care processes; the available record does not provide results for patient-level clinical outcomes or a basis for assessing clinical consequences.

14. Related Tutorials

15. Related Statistical Calculators

16. Sources

• ClinicalTrials.gov: IMPROVE HF: Registry to Improve the Use of Evidence-Based Heart Failure Therapies in the Outpatient Setting — NCT00303979.
• Registry record last updated 2019-01-31; status last verified 2019-01.
• Results first posted 2012-12-27.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

17. Record Summary

IMPROVE HF is a completed registry-based quality-improvement initiative listed under NCT00303979. The record describes patient-record review in cardiology practices and an assessment of changes in seven heart-failure performance measures. The primary objective compares aggregate performance at 24 months with baseline and counts measures reaching a relative improvement of at least 20%.

The longitudinal cohort was assessed at baseline, 12 months, and 24 months; two other cohorts were reviewed at single time points. The primary-outcome description states that 167 practices contributed baseline data and 155 contributed longitudinal follow-up after 12 practices withdrew before the next chart-review milestone. The record describes relative-change calculations, 95% confidence intervals, and one-sample proportion z-tests.

Numerical primary results, measure-specific denominators, and key details about missing data, multiplicity, and practice-level statistical handling are not reported in the registry record. The record therefore defines the question and methods but does not provide the results needed to determine what changes occurred.