Package of Resources for Assisted Contact Tracing: Implementation, Costs, and Effectiveness: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Package of Resources for Assisted Contact Tracing: Implementation, Costs, and Effectiveness (NCT05343390), a randomized na study. The primary outcome is Index Client Fidelity Assessment.

2. Clinical Question

3. Trial Design

This was a randomized, parallel-group implementation science study, registered as NCT05343390. The registry describes the design as cluster-randomized, so assignment occurred by groups rather than by treating every participant as an independent randomized unit. The study enrolled 841 participants. Its status is completed, and the registry lists the study start as 2021-11-01 and primary completion as 2023-11-30.

The comparison was between an enhanced implementation package and a standard implementation package. The enhanced strategy involved digitally guided training and continuous quality improvement; the standard strategy involved standard training and clinical support. The registry describes single masking, but does not report who was masked or how masking was carried out. The phase is listed as NA.

The study addressed the practical challenge of helping health workers provide assisted contact tracing in Malawi. Its brief summary describes health workers assisting people with HIV to recruit sexual contacts and biological children for testing. The study’s stated focus was whether digital capacity-building tools could help health workers provide these services effectively and efficiently. The primary outcome was assessed one year after training, through observation of health workers counseling actual index clients.

The registry lists nine outcome measures and seven analyses. That count does not, by itself, identify which outcomes were confirmatory or how the analyses were prioritized. Details such as the planned allocation ratio, cluster definition, number of clusters, analysis populations, and full statistical analysis plan are not reported in the registry record. Those omissions matter when interpreting a cluster-randomized study, because the grouping and analysis of participants affect how uncertainty should be estimated.

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: 33 clusters were randomized in a 1:2 ratio (Enhanced:Standard). Within these clusters Healthcare workers, Index clients, Contact clients and data from registers were observed. To clearly represent the groups within each arm, four "periods" are shown below. Periods do not reflect a time element, but rather the four populations measured within these clusters (units).

Clusters and Healthcare Workers

Clusters and Index Clients

Clusters and Contact Clients

Clusters and Register-Extracted Data

Analysis population for the primary outcome, as defined in the registry: Data were missing for 2 participants.

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: In the Enhanced arm there were 125 Healthcare workers, 124 index clients, and 67 contact clients for a total of 316. In the Standard arm, there were 181 Healthcare workers, 217 Index clients, and 127 Contact clients for a total of 525. The "Cluster and Register-Extracted data" population were not considered as enrolled participants in the study; they were extracted data from register and no data were collected from these individuals.

6. Primary Endpoint

The registry lists the primary outcome as Index Client Fidelity Assessment, measured over 1 year after training.

The primary endpoint was the Index Client Fidelity Assessment, a score based on a 15-item assessment of health workers counseling actual index clients. The score ranges from 0 to 100%, with higher scores indicating better fidelity. The endpoint was assessed one year after training, and the registry describes the outcome parameter as a mean.

Fidelity is relevant to an implementation study because it evaluates whether health workers carry out the counseling behaviors that the implementation strategy is intended to support. It is not itself a count of contacts tested or a measure of HIV diagnoses. Nor does the registry description make it a direct measure of longer-term health outcomes. It captures performance during observed counseling with actual index clients.

Comparing mean scores between the implementation strategies gives a single summary of the groups’ average assessment results. The reported mean difference of 20.6 is in the direction of higher scores for the enhanced strategy, according to the order and result presented in the registry’s superiority analysis. The confidence interval communicates uncertainty around that estimated difference. The registry states that data were missing for 2 participants; further details on missingness and how those observations were handled are not reported in the registry record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary comparison is a comparison of means for a continuous score. A mean summarizes the average score within a group. A mean difference summarizes how far apart the two group averages are, in the units of the score. In this study, the posted mean difference is 20.6 on the Index Client Fidelity Assessment, whose range is 0 to 100%. The registry labels the estimate a mean difference in final values and reports a linear regression analysis.

What does the confidence interval communicate?

A confidence interval presents statistical uncertainty around an estimate. The reported 95% confidence interval extends from 16.1 to 25.1. Under the assumptions of the statistical procedure, the interval is a measure of how precisely the mean difference has been estimated. A narrower interval indicates greater precision than a wider interval, all else equal. The interval does not give the spread of individual scores, and it does not mean that every participant’s outcome is near the estimated mean difference.

The reported p-value is < 0.001. In general, a p-value measures how unusual data at least as incompatible with a specified null hypothesis would be if that null hypothesis and the model assumptions were true. It is not a measure of the size or importance of the effect. It also does not quantify the probability that the finding is due to chance, nor does it establish that the model assumptions are appropriate.

Why does cluster randomization matter?

In a cluster-randomized trial, groups of participants are assigned together. Outcomes from participants in the same cluster may be correlated because they share staff, facilities, training, or other features of their setting. When that correlation is present, the amount of independent information is not captured by simply counting every participant as an independent unit. A suitable analysis needs to account for the cluster structure when estimating uncertainty.

The registry explicitly notes that a simple individual-level re-analysis would be misleading and that only the sponsor’s own analysis is shown. That is why the sponsor’s reported estimate and its uncertainty take precedence over an unadjusted re-calculation. The registry does not report the cluster definition, the number of clusters, the intracluster correlation, or the detailed regression specification, so the exact handling of clustering cannot be assessed from the record alone.

What does regression contribute?

Linear regression is a method for describing the relationship between an outcome and one or more explanatory variables. For a continuous endpoint, a regression coefficient can express an estimated difference in average outcome between groups. The registry reports linear regression and a mean difference in final values. It does not state whether the model included covariates, how variables were coded, or what variance estimator was used. Those details should not be inferred from the method label alone.

The registry’s sample-size section describes planning illustrations based on the observed effect, with a two-sided significance level of 0.05, equal allocation, and a normal approximation. It says the effect to detect is a stated fraction of the observed effect. Such figures are planning illustrations, not observed power and not evidence of the trial’s original design assumptions. The registry does not report the trial’s sample-size justification in the information available for this analysis.

How should missing data be considered?

The primary outcome description says data were missing for 2 participants. Missing observations can affect an estimate if the chance of a score being missing is related to the outcome or to the intervention group. In general, analyses may use different assumptions and methods to address missing data. The registry does not report reasons for missingness, how missing data were distributed across groups, or the method used in the primary analysis. The reported estimate should therefore be interpreted with that uncertainty in mind.

The registry lists nine outcome measures and seven analyses. When a study examines multiple outcomes or performs multiple analyses, the chance of at least one apparently unusual result can rise if no adjustment or prespecified hierarchy is used. The registry record does not report a multiplicity strategy, an interim-analysis plan, or the full statistical analysis plan. These absences do not show that no such procedures existed; they mean that those details are not reported in the registry record.

10. Analyses Reported by the Sponsor

The sponsor’s own analysis is the central estimate for this cluster-randomized study. It reports linear regression, a mean difference in final values of 20.6, a 95% confidence interval from 16.1 to 25.1, and p < 0.001. Because the registry warns that a simple individual-level re-analysis would be misleading, an unadjusted recalculation from summary statistics should not be substituted for the sponsor’s result.

No separate recalculated mean difference or recalculated confidence interval is reported for this comparison. The posted result should therefore not be described as agreeing or disagreeing with an independent recalculation. The sponsor’s analysis takes precedence because the design is cluster-randomized and an analysis must account for that design to estimate uncertainty appropriately.

The record identifies the method as linear regression but does not report the full model specification or the method used to account for clustering. The stated result can be reported as the sponsor’s analysis, while recognizing that the available registry entries do not permit a full reconstruction of the model. This distinction is important: a result can be presented faithfully without claiming that every modeling decision can be independently checked from the record.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists nine outcome measures and seven analyses. The primary endpoint is the Index Client Fidelity Assessment, but the record does not report a complete hierarchy distinguishing confirmatory outcomes from exploratory outcomes. It also does not report whether the analyses were prespecified in a statistical analysis plan or whether any procedure was used to account for testing multiple outcomes or hypotheses.

Multiplicity matters because interpreting many statistical tests separately can make at least one apparently strong result more likely, even when there is no true difference for some outcomes. General approaches include defining a primary outcome in advance, specifying a testing order, or adjusting significance thresholds. Which approach was used in this study is not reported in the registry record. The reported p-value should be interpreted as the result of the posted analysis, not as evidence that multiplicity was or was not controlled.

For the primary outcome, the registry says data were missing for 2 participants. The record does not give the reasons, group allocation, or characteristics of those participants, and it does not describe the missing-data method. Without that information, it is not possible to assess from the registry record whether missingness could have affected the estimate or how robust the analysis was to different assumptions.

The record does not report the analysis populations, such as whether all randomized clusters or participants were included according to original assignment. It also does not report interim analyses or stopping rules. These details are relevant to understanding the handling of participants and the interpretation of statistical testing, but their absence from the registry record should not be taken as evidence that they were not addressed elsewhere. The available primary estimate remains the sponsor’s cluster-aware analysis, with these reporting limitations acknowledged.

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.

Most frequent serious adverse event terms

Serious events with the highest arm-level rate: surgery complication (0.6%).

The registry reports safety figures separately for healthcare workers, index clients, and contact clients in each strategy group. In the enhanced strategy, the listed serious adverse event counts are 0 among 125 healthcare workers, 0 among 124 index clients, and 0 among 67 contact clients. The corresponding death counts are also 0 in each of those groups.

In the standard strategy, the registry lists 1 serious adverse event among 181 healthcare workers, 0 among 217 index clients, and 0 among 127 contact clients. It lists 1 death among standard-strategy healthcare workers and 0 deaths in the other listed groups. The record gives “Surgery complication” as the top serious event term and reports 0.6%. It does not establish from that entry alone how the term relates to a particular group or participant.

These are descriptive counts, not a randomized comparison designed to establish a difference in safety. The groups are listed by participant category, and the record does not explain how the safety populations relate to the primary outcome population. Sparse event counts also provide limited information about uncommon harms. No causal conclusion should be drawn from these figures alone, and the registry does not report further event narratives or attribution details.

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

The posted primary result indicates that the average Index Client Fidelity Assessment score was higher in the enhanced implementation strategy group than in the standard strategy group, by a reported mean difference of 20.6. The 95% confidence interval, from 16.1 to 25.1, describes the statistical uncertainty reported by the sponsor’s analysis. The endpoint concerns observed counseling fidelity one year after training, with higher scores indicating better fidelity.

This finding is about a measured implementation outcome. It is not a direct estimate of the number of contacts who were tested, the number of new HIV diagnoses, or the effect of the strategies on health outcomes. The registry’s study rationale discusses assisted recruitment and testing, but the primary endpoint is a fidelity score. The score therefore speaks to observed counseling performance as measured by the assessment, not every step or consequence of contact tracing.

The result also does not establish that the enhanced strategy would have the same effect in every setting. The trial concerns a particular implementation comparison in Malawi, and the registry record does not provide enough detail to assess how the result might vary across settings or groups. Nor does the reported p-value establish the practical importance of the difference; statistical evidence and practical relevance are distinct questions.

The study was cluster-randomized, and the registry specifically warns against a simple individual-level re-analysis. Interpretation should therefore rely on the sponsor’s own analysis rather than an analysis that ignores the grouping structure. Details of the cluster-aware model are not reported in the registry record, and the primary outcome had data missing for 2 participants. Those reporting limitations qualify how fully the result can be independently assessed.

Finally, a result for the primary endpoint does not determine the findings for all nine outcomes or seven analyses listed in the registry. The record does not report how multiplicity was addressed or provide the full statistical analysis plan. The result is best understood as the sponsor-reported estimate for the specified primary score, at the stated assessment time, under the analysis used for this cluster-randomized study.

14. Why This Trial Matters Statistically

This trial raises a central statistical issue in implementation research: the unit used to assign an intervention may not be the same as the unit used to measure an outcome. When assignment is clustered, outcomes among people in the same group may be related. Treating all participants as independent can produce uncertainty estimates that do not reflect the design. The registry’s explicit warning against a simple individual-level re-analysis makes that issue especially important for interpreting the posted result.

The endpoint also illustrates how implementation studies can evaluate the delivery process rather than only patient-level clinical outcomes. Here, the primary measure is an observed counseling fidelity score, taken one year after training. That makes it possible to ask whether an implementation strategy is associated with differences in the behavior the training aimed to support. It does not, by itself, answer whether downstream testing or health outcomes changed.

Reporting the mean difference alongside a confidence interval helps distinguish the estimated size of the group difference from its statistical precision. The interval from 16.1 to 25.1 gives more information than a p-value alone, while still depending on the model and its assumptions. The missing details about clustering, model specification, multiplicity, and missing-data handling show why estimates should be read together with design and reporting information.

The trial also highlights the importance of separating planning calculations from observed evidence. Illustrations based on a fraction of an observed effect can inform prospective planning, but they are not the trial’s observed power and do not replace a design rationale. The record does not report the original sample-size justification. Keeping those distinctions clear helps readers interpret a result without overstating what the statistical information establishes.

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 registry record leaves important design and analysis details unreported. Although it identifies the study as cluster-randomized, it does not describe the clusters, the number of clusters, or the allocation process in detail. Those features are necessary to understand the independent units of assignment and the precision of a cluster-aware estimate. The record reports linear regression but not the full model, covariate adjustment, or variance-estimation method.

The primary outcome had data missing for 2 participants, but the reasons, group distribution, and analytical handling of those missing observations are not reported in the registry record. The record also does not identify the analysis population. These gaps limit assessment of whether the primary estimate depends on particular assumptions about inclusion or missingness.

Nine outcome measures and seven analyses are listed, but the record does not report the full statistical analysis plan, a multiplicity strategy, or which outcomes were confirmatory. Interim analyses and stopping rules are not reported. The study’s sample-size justification is also not reported; planning illustrations based on the observed effect should not be treated as the original design rationale or as observed power.

The primary endpoint is a score from observed counseling, and the registry description does not provide further information on assessor procedures or measurement properties beyond the 15-item assessment and its 0 to 100% range. The result therefore addresses that defined score, not every component of contact tracing. Safety reporting is descriptive and provides limited detail on event attribution and narratives. These limits do not negate the posted estimate, but they constrain how much can be independently assessed from the registry record.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Package of Resources for Assisted Contact Tracing: Implementation, Costs, and Effectiveness — NCT05343390.
• Registry record last updated 2025-02-06; status last verified 2024-04.
• Results first posted 2025-02-06.
• 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

NCT05343390 is a completed, randomized, parallel-group, cluster-randomized implementation science study in Malawi, with 841 participants. It compared digitally guided training and continuous quality improvement with standard training and clinical support. The primary outcome was the Index Client Fidelity Assessment, a 15-item score ranging from 0 to 100%, assessed one year after training, with higher scores indicating better fidelity.

The sponsor’s linear regression analysis reported a mean difference in final values of 20.6, with a 95% confidence interval from 16.1 to 25.1 and p < 0.001. Data were missing for 2 participants on the primary outcome. The registry cautions that an individual-level re-analysis would be misleading and reports only the sponsor’s analysis. Further details on the cluster structure, full statistical plan, missing-data handling, and multiplicity are not reported in the registry record.