Comprehensive Post-Acute Stroke Services: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Comprehensive Post-Acute Stroke Services (NCT02588664), a randomized na study. The primary outcome is Stroke Impact Scale (SIS-16).

2. Clinical Question

3. Trial Design

This was a completed, randomized, parallel-group study with a cluster-randomized design. Hospitals, rather than individual participants, were assigned to the COMPASS Intervention or Usual Care. The registry record lists 6,024 participants enrolled and identifies Wake Forest University Health Sciences as the sponsor. The start date was 2016-07-25 and the primary completion date was 2018-07-25.

In a cluster-randomized study, people treated within the same hospital may have more similar outcomes than people treated at different hospitals. This can happen because they share care teams, practices, or local context. As a general statistical principle, analyses should account for this within-hospital similarity; treating every participant as statistically independent can make uncertainty appear smaller than it is. The registry record does not report the sponsor’s specific modeling approach or the number of participating hospitals.

The registry describes the masking as single, but who was masked is not reported in the registry record. The phase is listed as not applicable. The usual-care arm continued the hospitals’ usual post-acute stroke care, while hospitals assigned to COMPASS changed the structure and process of that care. The registry record does not report further details about how consistently the model was implemented across hospitals.

The study summary says the intervention was intended to improve functional outcomes, reduce caregiver stress, and reduce readmission rates. The listed secondary outcomes include caregiver strain, self-reported general health, modified Rankin score, physical activity, depression, cognition, medication adherence, and falls. The record lists 27 outcome measures, but reports 0 analyses.

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: Excluded subsequent stroke (or TIA) events within the study period (N=142)

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Of the 5,882 that were enrolled in the study, 3476 provided a response to the survey question. However, we utilized inverse probability weight to perform the analysis in a way to account for missing data. The final analysis included 5,882 for this outcome measure

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: Subsequent events (i.e. secondary stroke events) were excluded within the study period.

6. Primary Endpoint

The registry lists the primary outcome as Stroke Impact Scale (SIS-16), measured over post-stroke day 90.

The primary endpoint was the Stroke Impact Scale (SIS-16), a 16-item survey about difficulty performing basic physical activities. It was assessed at post-stroke day 90. The scale ranges from 0-100, with higher scores corresponding to more favorable outcomes. The registry describes the measure as a mean with standard deviation.

Functional status is relevant to a study of post-acute stroke care because it captures participants’ reported ability to carry out physical activities. A continuous score can describe the distribution of function across participants, rather than reducing outcomes to a simple yes-or-no category. In general, a group mean gives a measure of the average score, while a standard deviation describes how spread out individual scores are around that mean.

The primary-outcome description says 3,476 participants responded to the survey question. It also says inverse probability weighting was used to account for missing data and that the final analysis included 5,882 participants. The registry record does not report the group-specific means, the estimated between-group difference, a confidence interval, or a p-value. Without those results, the SIS-16 outcome’s direction, magnitude, and precision cannot be assessed from the registry record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

Why the randomization unit matters

The study used cluster randomization: hospitals were assigned to care groups. General statistical principle: when participants belong to randomized clusters, their outcomes may be correlated within a cluster. For example, care processes shared within a hospital can affect multiple participants. If an analysis ignores that correlation, it may overstate the amount of independent information and understate uncertainty. The analysis should reflect the actual randomization unit. The registry record does not report the sponsor’s detailed statistical model, cluster adjustment, or uncertainty calculation.

What a mean and standard deviation describe

The primary endpoint is continuous: it is a score on a scale, analyzed as a mean with standard deviation. In general, the mean describes the average score in a group. The standard deviation describes the spread of scores around that average; it is not the uncertainty in the estimated mean or in a comparison between groups. A treatment comparison would ordinarily be expressed as an estimated difference between group means, accompanied by a measure of uncertainty. Those results are not reported in the registry record.

How to interpret a confidence interval

In general, a confidence interval gives a range of effect values compatible with the observed data under the assumptions of the statistical model. A narrow interval indicates greater precision than a wide interval, while the interval’s position relative to the null value helps describe compatibility with no difference. A confidence interval does not guarantee that the true effect lies within its particular limits, nor does it establish clinical importance by itself. No confidence interval for the primary comparison is reported in the registry record.

What inverse probability weighting is for

The primary-outcome description says inverse probability weighting was used to account for missing survey data. In general, this method gives different weights to observed participants according to their estimated probability of having an observed response. The intent is to make the analyzed responses better represent the target population when response is related to measured characteristics. Its reliability depends on assumptions, including that relevant predictors of response are measured and that the probability model is adequate. The registry record does not report the weighting model, predictors, diagnostics, or sensitivity analyses. The record states that 3,476 participants responded and that the final analysis included 5,882 participants; it does not describe the individual weights.

Why planning illustrations are not trial results

In general, a sample-size calculation asks how many participants would be needed under stated assumptions to detect a chosen effect with a specified probability. Such a calculation is a planning exercise, not an observed result and not “observed power.” The record’s stated illustration assumptions include a two-sided significance level of 0.05, equal allocation, a normal approximation, and effects expressed as fractions of the observed effect. These assumptions do not reveal the trial’s original design justification. The registry record does not report the trial’s sample-size justification, and no outcome effect estimate is reported from which readers could assess the illustration’s empirical basis.

For cluster-randomized studies, participant counts alone do not describe the information available for a comparison: the number and size of clusters and the within-cluster similarity also matter. The registry record does not report the number of hospitals or the details needed to assess those design features. Any planning illustration should therefore be kept distinct from evidence about the observed comparison.

What the analysis record can establish

The registry lists 27 outcome measures and 0 analyses. The record identifies the primary endpoint and describes an approach to missing survey responses, but it does not report the estimated contrast, standard error, confidence interval, or p-value. It also does not report an interim-analysis plan or detailed statistical analysis plan. Without these details, readers cannot independently assess the primary estimate, its precision, or whether the analysis followed a prespecified plan.

10. Confidence Intervals for Each Arm

No two-arm comparison is shown for this trial, but each arm's mean can be given with a t-based 95% confidence interval computed from the posted mean, spread and number of participants.

R code and output

arm <- c("Usual Care", "COMPASS Intervention")
n   <- c(1832, 1644)
m   <- c(79.9, 80.6)
sp  <- c(21.4, 21.1)
sd  <- sp      # posted standard deviations
for (i in seq_along(n)) {
  se <- sd[i] / sqrt(n[i]); h <- qt(0.975, n[i] - 1) * se
  cat(sprintf("CARM|%s|%d|%.3f|%.3f|%.3f|%.3f\n", arm[i], n[i], m[i], sd[i], m[i] - h, m[i] + h))
}

CARM  Usual Care  1832  79.900  21.400  78.919  80.881
CARM  COMPASS Intervention  1644  80.600  21.100  79.579  81.621

11. Multiplicity, Missing Data and Analysis Populations

The record lists 27 outcome measures. In addition to the SIS-16 primary endpoint, listed secondary outcomes include caregiver strain, self-reported general health, modified Rankin score, physical activity, depression, cognition, medication adherence, and falls. When a study examines multiple outcomes or comparisons, the chance of at least one apparently unusual result can increase. General approaches to multiplicity include prioritizing a primary endpoint, adjusting inference across a defined family of tests, or treating secondary findings as exploratory. The registry record does not report a multiplicity strategy or analyses for the listed outcomes.

Missingness is relevant to the primary endpoint: 3,476 participants responded to the survey question, and the record states that inverse probability weighting was used. The final analysis included 5,882 participants. In general, weighting can address differences in response if the model and assumptions are appropriate, but it cannot ensure that unmeasured differences between respondents and nonrespondents have been removed. The weighting model, assumptions, diagnostics, and sensitivity analyses are not reported in the registry record.

The primary-outcome description identifies a final analysis population, but the record does not report further analysis-population definitions, such as how participants were handled if they did not follow their assigned care pathway. It also does not report a detailed protocol or statistical analysis plan. These omissions limit assessment of prespecification, missing-data robustness, and multiplicity. No outcome analysis is reported in the registry record.

12. Safety Results

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

Deaths

Other (non-serious) adverse events

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

The safety figures list 3,137 participants at risk in Usual Care and 2,689 in the COMPASS Intervention group. The serious-adverse-event figures show 0 affected participants in each group. Deaths are listed separately: 56 in Usual Care and 55 in the COMPASS Intervention group.

These are counts, not rates. The registry record does not give a safety analysis with follow-up time, event definitions, or an assessment of whether events were considered related to the intervention. The zero serious-adverse-event counts should therefore be read as no affected participants recorded in those fields, not as evidence that serious events were impossible or that the groups had identical safety. The death counts are also descriptive and do not, by themselves, establish cause or a between-group difference.

No top serious-event terms are listed. Further details about how adverse events were collected or reviewed are not reported in the registry record. Safety interpretation is limited to the posted counts and should not be extended to unreported event types or causal conclusions.

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

The registry record establishes that this was a randomized, hospital-level comparison of a comprehensive post-acute stroke care model with usual care. It identifies the SIS-16 at post-stroke day 90 as the primary endpoint and describes use of inverse probability weighting for missing survey data. These details clarify the research question and some aspects of the planned outcome assessment.

The record does not report the primary outcome estimate or any outcome analyses. It therefore does not show whether the intervention group had higher or lower SIS-16 scores, whether any difference was statistically distinguishable from no difference, or how precise an estimate might be. A scale range and direction are not substitutes for the observed group results.

The study summary describes intended aims that include functional outcomes, caregiver stress, and readmissions. Those aims should not be confused with findings. The record lists secondary outcomes, but does not report their results. Likewise, the safety counts are descriptive and do not establish that the intervention caused or prevented events.

In general, interpretation of a cluster-randomized trial depends on analysis that reflects hospital-level assignment and on transparent reporting of estimates, uncertainty, missing-data methods, and outcome-selection procedures. The registry record does not provide enough statistical detail to assess all of these features. The appropriate conclusion is limited: the record describes the trial and its primary endpoint, but does not provide reported evidence for the intervention’s effect on that endpoint.

14. Why This Trial Matters Statistically

This trial raises statistical issues that commonly arise when care is changed at the hospital level. Cluster randomization can reduce contamination between treatment approaches within a hospital, but it also means that participants within the same hospital may not contribute fully independent information. Appropriate analysis must reflect that structure. The registry record notes the cluster design but does not report the specific sponsor analysis method.

The study also illustrates the distinction between outcome response and analysis population. The primary-outcome description reports 3,476 survey responses and a final weighted analysis including 5,882 participants. Reporting how weighting was constructed and evaluating its assumptions are important for understanding how missing outcome data affect the estimate. Those details are not reported in the registry record.

Finally, a clearly defined endpoint does not by itself answer the clinical question. The SIS-16 is specified with its assessment time, scale, and direction, but the group comparison and its precision are absent. Listing 27 outcome measures without reported analyses or a stated multiplicity approach also leaves the outcome evidence unavailable for evaluation. The statistical importance of the record is therefore in the design and reporting questions it raises, rather than in an interpretable estimate of the intervention effect.

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 principal limitation for interpreting the study results is that the registry record lists 0 analyses and does not report the primary SIS-16 comparison. No group means, effect estimate, confidence interval, or p-value are available in the record. The result’s magnitude and precision therefore cannot be assessed.

The primary endpoint relied on survey responses, and the record states that inverse probability weighting addressed missing data. It does not describe the response model, the assumptions behind the weights, diagnostic checks, or sensitivity analyses. The record also does not explain how the final analysis population was defined beyond stating that it included 5,882 participants.

Because hospitals were randomized, the number of hospitals and the sponsor’s method for accounting for clustering would be important to assess. Those details are not reported in the registry record. The record also does not report who was masked, a detailed statistical analysis plan, the sample-size justification, interim analyses, or a multiplicity strategy. These are reporting limitations; they do not, by themselves, establish that the study was conducted improperly.

The safety information is limited to counts, without reported event definitions, follow-up details, or causal assessments. The registry record’s description of intended benefits and listed secondary outcomes should not be treated as evidence of results. Overall, the available record supports description of the design and endpoint but does not allow a complete statistical appraisal of the trial findings.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Comprehensive Post-Acute Stroke Services — NCT02588664.
• Registry record last updated 2021-06-11; status last verified 2020-12.
• Results first posted 2021-06-11.
• 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

NCT02588664, titled “Comprehensive Post-Acute Stroke Services,” is a completed randomized, parallel-group, cluster-randomized study sponsored by Wake Forest University Health Sciences. It enrolled 6,024 participants and compared the COMPASS Intervention with Usual Care. The intervention changed the structure and process of post-acute stroke care; the study summary describes integration of Early Supported Discharge and Transitional Care Management.

The primary endpoint was the mean SIS-16 score at post-stroke day 90, with higher scores indicating more favorable outcomes. The primary-outcome description reports inverse probability weighting for missing survey data and a final analysis including 5,882 participants. The registry record lists 27 outcome measures and 0 analyses. It does not report the primary outcome estimate or its uncertainty, so the intervention effect cannot be evaluated from the registry record.