Quality Improvement and Personalization for Statins: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Quality Improvement and Personalization for Statins (NCT02820870), a randomized na study. The primary outcome is Percent of Visits Where Moderate-to-high-strength Statins Are Appropriately Initiated.

2. Clinical Question

The clinical question is about a care process: whether an appropriate statin was initiated after a qualifying visit. It is not a comparison of cardiovascular events, long-term health outcomes, or the effects of statin therapy among patients already taking it. The intervention was directed at clinical practice and guideline adoption, while the outcome was assessed using electronic health record medication data.

3. Trial Design

This was a randomized, parallel-group, cluster trial conducted in a single clinic. Care teams, rather than individual visits, were assigned to intervention or usual care. Five care teams were randomized: two teams, with 15 total providers, received the intervention; three teams, with 28 providers, served as controls. The registry lists enrollment of 43. It does not further clarify how that enrollment figure relates to the provider and visit counts in the outcome description.

The intervention was a bundle of quality-improvement activities: a personalized decision-support tool, education, a performance measure, and audit and feedback. For the intervention group, the description says that providers received a generated print-out with statin recommendations for patients whose current medication therapy was guideline-discordant. The description is truncated in the registry record, so further operational details are not reported there. Usual-care teams were expected to follow VA guidelines and HEDIS measures as part of the VA national roll-out.

The primary outcome was based on visits. The registry reports 573 eligible visits in the intervention arm and 673 in the control arm during the QI period. A visit counted as an opportunity when the patient was guideline-appropriate for a moderate-to-high-strength statin but was not taking one. The outcome was met if the patient started such a statin within 30 days after the visit, based on electronic health record medication data. The registry explicitly allows for a patient to appear in more than one visit.

Masking is listed as none, and the trial status is completed. The recorded study dates run from 2016-03 to primary completion in 2016-08. The registry identifies the sponsor as the VA Office of Research and Development. Additional protocol details, including any further description of the randomization procedure, are not reported in the registry record.

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 unit of analysis was the visit during the 3-month intervention, with clustering for provider. Patients who met VA/DoD statin recommendations were excluded from the project. Each time-period (pre-intervention, intervention, post-intervention) are visits, and hence should have different sample sizes.

Before Intervention

Intervention

Post Intervention

Analysis population for the primary outcome, as defined in the registry: Five care teams were randomized. Two with 15 total providers received the intervention. Three, with 28 providers, were the control arm. During the QI period, the intervention arm had 573 eligible visits and the control arm had 673. Our outcome was the chance an appropriate statin would be started at a visit.

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.

Population: These numbers reflect the number of primary care providers who were randomized to the Intervention Group and Usual Care Group.

6. Primary Endpoint

The registry lists the primary outcome as Percent of Visits Where Moderate-to-high-strength Statins Are Appropriately Initiated, measured over 30 days.

The primary endpoint measures the percentage of eligible visits at which a moderate-to-high-strength statin was appropriately initiated within 30 days. Eligibility required that the patient be considered appropriate for that statin strength under the guideline criteria and not already be taking one. Initiation was assessed using electronic health record medication data. Thus, the endpoint represents a documented prescribing and medication-initiation process tied to a qualifying visit.

A visit-based endpoint is useful for assessing a workflow intervention because the intervention was intended to influence provider decisions at the point of care. It allows the analysis to count opportunities for appropriate initiation, including more than one eligible visit for a patient who remained guideline-discordant. The registry specifically says that the analysis was based on visits, not patients.

That choice also affects interpretation. Repeated visits from the same patient are not necessarily independent observations, and visits are additionally grouped within randomized care teams. The statistical analysis therefore needs to reflect the trial’s clustered assignment and the visit structure. The registry reports logistic regression for the primary comparison, but detailed information about how clustering and repeated visits were represented in the model is not reported in the registry record.

The endpoint does not measure whether a patient continued taking a statin, achieved a particular lipid level, or experienced a cardiovascular outcome. Nor does it describe the absolute initiation percentage in either group in the information reported for the primary analysis. It answers a narrower question about initiation after an eligible visit.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

Logistic regression is a method for analyzing a binary outcome, such as whether an eligible visit was followed by statin initiation within the specified timeframe. It models the log odds of the outcome and can express a comparison between groups as an odds ratio. An odds ratio above the neutral value indicates higher odds for the numerator group relative to the denominator group, while an odds ratio below it indicates lower odds. The registry reports an odds ratio of 3.00, a 95% confidence interval from 1.84 to 4.9, and p < 0.01. Group coding and fuller model details are not reported in the registry record.

What does the odds ratio describe?

The odds ratio summarizes a relative comparison of odds, not a ratio of percentages and not a percentage-point difference. A value of 3.00 does not mean that three times as many visits resulted in initiation, nor does it give the absolute probability of initiation in either group. Those quantities depend on the underlying group outcome rates, which are not reported in the primary result information. An odds ratio can be useful for comparing a binary outcome, but it should be kept distinct from an absolute effect measure.

For a practical reading, the estimate needs to be interpreted together with the trial design, outcome definition, and uncertainty interval. The 95% confidence interval gives a range of values compatible with the observed data under the model and its assumptions. It communicates uncertainty about the estimated contrast; it is not a range containing 95% of individual visit outcomes. A confidence interval also does not resolve questions about bias, implementation, or whether the findings generalize beyond the study setting.

Why does clustering matter?

Randomization occurred at the care-team level. Visits within the same care team may share workflows, provider practices, and implementation conditions. As a result, outcomes from visits in one team can be more alike than outcomes from visits in different teams. This similarity is called within-cluster correlation. When it is present, the amount of independent information is less than a calculation that treats all visits as unrelated would imply.

Over-dispersion is one way a model can be less precise than a basic model expects, often because observations show more variability or dependence than the model allows. Accounting for cluster assignment can change standard errors and confidence intervals, even if the estimated contrast itself is similar. The registry identifies the analysis as logistic regression, but does not report the specific approach to clustering, the variance estimator, or whether additional levels of dependence from repeated patient visits were addressed. Those details are not reported in the registry record.

This is why a simple re-analysis at the individual-visit level would be misleading for this trial. The randomized units were care teams, not individual visits, and the registry explicitly notes the possibility of repeated visits for the same patient. The reported sponsor analysis is the only analysis presented for this comparison in the registry record, so its reported estimate should not be replaced by an unclustered calculation from visit totals.

How should the p-value be read?

The reported p-value, p < 0.01, is a measure of how unusual the data would be under a specified null model and the assumptions used in the analysis. A small p-value is evidence against that null model, but it does not quantify the probability that the null hypothesis is true. It also does not show the probability that the result will reproduce in another clinic. A p-value does not indicate clinical importance, and it cannot substitute for absolute outcome rates or a full assessment of design and analysis.

The registry does not report the complete statistical plan, including the prespecified null hypothesis details, model covariates, or any planned sensitivity analyses. It also does not report whether any interim analyses were conducted. These omissions limit the ability to assess how the reported result fits with the full analysis plan.

What can planning calculations tell readers?

Sample-size calculations based on an observed effect are planning illustrations, not observed power for the completed trial. They can illustrate what size of future study might be needed under stated assumptions, but their output depends on those assumptions and on treating the observed effect as a planning target. Such calculations do not show the probability that the completed trial detected an effect. The registry’s planning assumptions describe a two-sided significance level of 0.05, equal allocation, a normal approximation, and an effect expressed as a fraction of the observed effect. The calculations should therefore be read as illustrations rather than as a description of the trial’s original design or as evidence of observed power.

More generally, statistical interpretation depends on the chain from randomization to measurement to modeling. In this trial, teams were randomized, the outcome was recorded at eligible visits, and the reported model was logistic regression. The record does not provide all details needed to reconstruct the analysis. Readers can interpret the reported odds ratio and its interval, while recognizing that the exact treatment of team-level clustering, repeated visits, missing data, and model adjustments is not reported in the registry record.

10. Analyses Reported by the Sponsor

The registry reports the sponsor’s logistic-regression analysis as an odds ratio of 3.00, with a 95% confidence interval from 1.84 to 4.9 and p < 0.01. The reported analysis is the relevant estimate for the cluster-randomized comparison. The trial description states that care teams were randomized, and a simple individual-level recalculation from visit totals would not preserve that design.

No simple unadjusted recalculation is presented for this result. The registry specifically notes that such a re-analysis would be misleading because of the cluster-randomized design and that only the sponsor’s own analysis is shown. There is therefore no recalculated estimate to compare numerically with the sponsor’s result. The available information also does not give the group-specific counts of visits meeting the endpoint, which would be needed for a basic reconstruction of the outcome rates.

The registry does not report the model’s detailed specification, including how clustering was handled. The sponsor’s estimate should consequently be read as the reported analysis, with its confidence interval and p-value, rather than treated as independently verified from the visit counts. The record does not give a separate recalculated result or a documented discrepancy between a recalculation and the sponsor’s analysis.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists the primary endpoint and secondary outcomes concerning appropriate statin initiation after the intervention, provider response forms returned, and responses on those forms. It does not report whether statistical testing across these outcomes was adjusted for multiplicity, whether an outcome hierarchy governed interpretation, or how multiple analyses were handled. Without that information, the reported p-value should be read as the result for the stated primary comparison, not as evidence that the full set of outcomes was controlled for repeated testing.

The registry reports the primary method as logistic regression, but the full statistical analysis plan is not reported in the registry record. Details such as model covariates, the method for accounting for team-level clustering, treatment of repeated visits from the same patient, and planned sensitivity analyses are not reported. The record also does not state whether the reported comparison followed an intention-to-treat approach or describe other analysis populations.

Missing-data procedures are not reported in the registry record. The primary outcome was based on electronic health record medication data, but the record does not describe completeness checks, how unavailable or ambiguous medication data were handled, or whether any visits were excluded for missing information. The actual number of visits meeting the endpoint is also not reported in the primary result information.

The registry does not report whether interim analyses were planned or conducted. It also does not report a statistical plan for the secondary outcomes, including whether they were exploratory or confirmatory. These absences do not change the reported odds ratio, confidence interval, or p-value, but they limit assessment of the analysis strategy as a whole. Interpretation should therefore remain focused on the reported primary comparison and acknowledge that multiplicity, missing-data handling, and analysis-population definitions are not documented in the registry record.

12. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 0% reporting threshold.

The registry lists zero serious adverse events and zero other adverse events in the intervention group, among 15 at risk, and zero serious adverse events and zero other adverse events in the usual-care group, among 28 at risk. These figures are the adverse-event counts recorded in the registry. Death counts are not reported in the record.

Zero recorded events should not be interpreted as proof that no adverse events could occur or that the groups had identical safety profiles. The registry does not report further detail about adverse-event collection, monitoring, definitions, or the period over which these figures were assessed. It also does not list serious adverse-event terms.

The trial focused on a quality-improvement process and a visit-based statin-initiation endpoint. The adverse-event figures do not provide a broad assessment of the risks associated with statin use, and they should not be treated as an estimate of medication safety outside the study. They are best read narrowly as the event counts recorded for the study arms, with the limitations of the registry’s safety reporting in mind.

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

The reported result concerns a specific care process: whether a moderate-to-high-strength statin was initiated within 30 days after an eligible visit. The sponsor’s logistic-regression analysis reports an odds ratio of 3.00, with a 95% confidence interval from 1.84 to 4.9 and p < 0.01. This provides evidence of a difference in the modeled odds for the comparison as reported, with uncertainty reflected in the confidence interval.

The result does not provide the absolute percentage of eligible visits leading to initiation in each arm, because those outcome counts or percentages are not reported in the primary result information. It therefore does not establish the absolute size of the change in practice. Nor does it describe patient adherence after initiation, longer-term medication use, lipid levels, cardiovascular events, or other clinical outcomes.

The intervention combined several activities, so the result cannot identify which component, if any, accounted for the observed contrast. It also does not show whether the same result would occur in other clinics or under different staffing, workflow, or implementation conditions. The trial was conducted in a single clinic, with care teams as the randomized units, and the registry does not report full model details.

Accordingly, the estimate should be understood as a finding about a bundled quality-improvement approach and a short-term, visit-based initiation measure in this study setting. It is not a general conclusion about statin treatment effects, nor does it establish a patient-level clinical benefit. The registry result supports a statistical comparison on the stated process endpoint; its scope should not be extended beyond what that endpoint and design measured.

14. Why This Trial Matters Statistically

This trial is statistically notable because its randomization and outcome measurement occur at different levels. Care teams were randomized, while the primary outcome was recorded for eligible visits. Some patients could contribute more than one visit. These features make the choice of analysis important: the model needs to respect team assignment and the possibility that observations are related within teams or patients.

The registry’s warning against a simple individual-level re-analysis highlights a core design principle. A large number of recorded visits does not by itself mean that every visit provides independent information about the intervention. When assignment occurs by cluster, the cluster structure affects how uncertainty should be assessed. Reporting the sponsor’s logistic-regression estimate, confidence interval, and p-value gives readers a result to interpret, while the missing model details limit a full assessment of how that structure was handled.

The endpoint also illustrates the difference between a process outcome and a clinical outcome. Appropriate initiation within 30 days is a measurable care-process target, but it does not on its own establish later adherence or patient health outcomes. The odds ratio captures a relative comparison on this binary process measure, whereas absolute initiation rates are not reported in the primary result information.

Finally, the intervention was a bundle, not a test of a single isolated component. The findings therefore relate to the combined quality-improvement strategy as implemented in a single clinic. Together, the cluster design, visit-based outcome, bundled intervention, and incomplete reporting of analysis details make this a useful example of why effect estimates must be interpreted in relation to both the unit of randomization and the unit of measurement.

15. 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 trial was conducted in a single clinic and randomized five care teams. That setting and cluster structure limit how broadly the result can be generalized and make the details of cluster-adjusted analysis important. The registry identifies logistic regression but does not report the specific handling of team-level dependence, repeated visits, or any covariate adjustment. A simple re-analysis treating visits as independent would be misleading, as the registry itself notes.

The primary outcome is visit-based, and the same patient could contribute more than one eligible visit. This is appropriate to the stated process question but means the outcome is not a count of unique patients. The registry does not report the group-specific number of visits meeting the endpoint or the absolute initiation percentages. Consequently, readers cannot assess absolute outcome rates from the reported primary result alone.

The intervention included decision support, education, a performance measure, and audit and feedback. Because these components were implemented together, the reported result cannot isolate the effect of any one component. The intervention-arm description is also truncated in the registry record, and further implementation details are not reported there.

Other limitations arise from incomplete statistical reporting. Multiplicity procedures, missing-data handling, analysis populations, interim analyses, and the full statistical analysis plan are not reported in the registry record. The registry also does not provide detailed adverse-event collection methods or death counts. These reporting gaps do not negate the reported analysis, but they constrain evaluation of its assumptions and the completeness of the trial account.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Quality Improvement and Personalization for Statins — NCT02820870.
• Registry record last updated 2019-12-20; status last verified 2019-12.
• Results first posted 2019-12-20.
• 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.

19. Record Summary

The registry record describes a completed, randomized, parallel, unmasked cluster trial of a statin-focused quality-improvement intervention in a single clinic. The intervention combined personalized decision support, education, performance measurement, and audit and feedback; the comparator was usual care in the context of VA guideline and HEDIS expectations.

The primary endpoint was appropriate moderate-to-high-strength statin initiation within 30 days of an eligible visit, assessed using electronic health record medication data. The analysis was visit-based, and a patient could contribute repeated eligible visits. The sponsor reports logistic regression with an odds ratio of 3.00, a 95% confidence interval from 1.84 to 4.9, and p < 0.01.

The result is specific to a process outcome and the trial setting. Absolute endpoint rates, several analysis details, multiplicity procedures, and missing-data methods are not reported in the registry record. The cluster-randomized design makes the sponsor’s reported analysis more relevant than a simple individual-level recalculation.