Patient Actor Training to Improve HIV Services for Adolescents in Kenya: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Patient Actor Training to Improve HIV Services for Adolescents in Kenya (NCT02928900), a randomized na study. The primary outcome is Retention in HIV Care.

2. Clinical Question

3. Trial Design

The registry describes a randomized, stepped-wedge trial at 24 HIV care and treatment facilities in Kenya. The intervention is introduced at study sites, with the control period defined as the period before a site receives training and the intervention period defined as the period after receipt. The record also labels the model as crossover. The available description does not specify the order or timing of each site’s transition, so further details about rollout should not be inferred.

The intervention is behavioral clinician training using standardized patient actors. Its intended focus is communication and interpersonal skills, including empathy, among health care workers serving adolescents and youth affected by HIV. The control period is standard of care without the training intervention. The registry lists random allocation and no masking. Details about who was aware of allocation, or whether outcome assessors were masked, are not reported in the registry record.

The registry records enrollment as 24, while its study summary describes 24 facilities. It does not clarify the enrollment figure’s unit in the information available for this analysis. The primary outcome is assessed by medical record review, and the described final analysis accounts for clustering by facility, time in the stepped-wedge design, and newly enrolled status. The specific statistical model, the allocation sequence, and the number and timing of measurement periods are not reported in the registry record.

The study is marked completed. Its listed start date is 2016-09-26, and its primary completion date is 2020-11-02. These dates describe the registry record and do not provide the timing of individual site transitions or visits.

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.

Baseline

Wave 1

Wave 2

Wave 3

Analysis population for the primary outcome, as defined in the registry: Medical record review. Final analysis accounts for clustering by facility, time (stepped wedge design), and newly enrolled in care status.

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: Clinics were assigned to the training intervention. Medical records were evaluated to assess study outcomes

6. Primary Endpoint

The registry lists the primary outcome as Retention in HIV Care, measured over Up to four years after baseline.

The primary endpoint is retention in HIV care, expressed as a count of participants meeting the registry’s retention definition. It covers two routes into the outcome: a return for the first follow-up visit within 3 months among newly enrolled adolescent clients, or a follow-up visit after a re-engagement visit. Re-engagement is described for people lost to follow-up for >90 days since their last visit, when there is no record of death or transfer.

This endpoint is used to assess whether engagement with care differs between the period before clinician training and the period after training at participating sites. It focuses on a documented follow-up event rather than a provider’s assessment of communication skills. Medical record review is the stated measurement approach.

In general, an operational definition helps make an outcome more consistent across participants and sites. Here, the definition distinguishes newly enrolled clients from those returning after loss to follow-up, and the final analysis is described as accounting for newly enrolled status. However, the registry record does not report the outcome counts by period, the estimated change, or its precision. It also does not provide further detail about how incomplete or ambiguous medical records were handled.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The trial is described as randomized and stepped-wedge, with facilities serving as clusters. In general, cluster randomization assigns groups, such as facilities, rather than assigning each individual independently. People treated at the same facility may share staff, procedures, and local context, so their outcomes can be more alike than outcomes from people at different facilities. An analysis that treats every person as statistically independent can therefore understate uncertainty or otherwise misrepresent the comparison. The registry explicitly cautions that a simple individual-level re-analysis would be misleading.

A stepped-wedge design introduces an intervention at participating clusters over time, so sites contribute observations during control and intervention periods. The registry defines these periods as before and after clinician training. In general, time is important in this design because outcomes may change for reasons other than the intervention. The registry states that the final analysis accounts for time and facility clustering. The specific model, assumptions, and method for estimating uncertainty are not reported in the registry record. Although the registry also labels the design as crossover, the available description does not provide enough detail to characterize the sequence or timing of transitions beyond the stepped-wedge description.

The primary outcome is a participant count based on a binary event definition: a person either meets the stated retention criterion or does not. In general, a binary endpoint can be summarized as a count or as a proportion, but those summaries are not interchangeable with an adjusted treatment effect. The registry says the primary outcome is change in retention between intervention and control periods. It also says the final analysis accounts for newly enrolled status, which matters because the endpoint definition includes a distinct route for newly enrolled clients and another for people returning after re-engagement. The analysis’s exact treatment of those categories is not reported.

The study’s stated outcome-measurement method is medical record review. In general, record-based endpoints depend on whether visits are documented and whether records can be linked and classified consistently. The registry does not report completeness of records, adjudication procedures, handling of missing or conflicting information, or whether follow-up availability differed between periods. Those details should not be assumed.

The registry lists secondary outcomes in several domains: pass/fail scores for health care workers, numeric scores from standardized patient actors, health care worker competency, health care workers’ satisfaction with skills, adolescent patients’ satisfaction with care, and secondary retention in HIV care. It lists 12 outcome measures, but the registry record does not report the results for those outcomes in the information. A list of outcomes establishes what was planned or recorded as measured; it does not establish an observed effect.

For clustered trials, general statistical practice is to use an analysis that reflects the grouping and trial timing, rather than to treat all observations as independent. Depending on the design and data, methods may use cluster-level summaries or models that account for within-cluster correlation and time. This is general methodological context, not a description of the method used in this trial. The registry does not report the model form, link function, covariance structure, degrees-of-freedom method, covariate set, or approach to missing observations. A numerical recalculation from participant-level counts would not recover the sponsor’s cluster- and time-adjusted result.

The registry’s planning description characterizes sample-size calculations as illustrations based on an observed effect, using 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. For time-to-event calculations, the description refers to a total number of events, while other calculations refer to participants per arm. The observed effect, effect fractions, and resulting planning values are not reported in the registry record. Such calculations are planning illustrations, not observed power, and they cannot be interpreted as evidence about the trial’s result.

In general, a confidence interval communicates uncertainty around an estimated effect, while a p-value summarizes how unusual the data would be under a null model and its assumptions. Neither can be reconstructed responsibly from the design description alone. The registry reports no primary estimate, confidence interval, or p-value. Since the design is cluster-randomized, the sponsor’s own analysis takes precedence over a simple re-analysis; the sponsor’s numerical analysis is not reported in the registry record.

10. Exact Estimates for Each Arm

No two-arm comparison is shown for this trial, but each arm's result can still be summarized with exact interval estimates computed from the posted counts. The Clopper-Pearson interval is exact and conservative; the Wilson interval is a common closer-to-nominal alternative; the Bayesian credible interval uses a uniform Beta(1, 1) prior. For small groups the three can differ noticeably.

R code and output

arm    <- c("Adolescent Records - Intervention Exposed", "Adolescent Records - Unexposed")
n      <- c(2123, 2472)
events <- c(1570, 1855)
for (i in seq_along(n)) {
  ci <- binom.test(events[i], n[i])$conf.int          # Clopper-Pearson exact interval
  cr <- qbeta(c(0.025, 0.975), 1 + events[i], 1 + n[i] - events[i])   # Bayesian, uniform Beta(1,1) prior
  wi <- (function(x, n) { p <- x / n; z <- qnorm(0.975); d <- 1 + z^2 / n
    c((p + z^2 / (2 * n) - z * sqrt(p * (1 - p) / n + z^2 / (4 * n^2))) / d, (p + z^2 / (2 * n) + z * sqrt(p * (1 - p) / n + z^2 / (4 * n^2))) / d) })(events[i], n[i])
  cat(sprintf("ARM|%s|%d|%d|%.1f|%.1f|%.1f|%.1f|%.1f|%.1f|%.1f\n", arm[i], n[i], events[i], 100 * events[i] / n[i],
              100 * ci[1], 100 * ci[2], 100 * wi[1], 100 * wi[2], 100 * cr[1], 100 * cr[2]))
}

ARM  Adolescent Records - Intervention Exposed  2123  1570  74.0  72.0  75.8  72.0  75.8  72.0  75.8
ARM  Adolescent Records - Unexposed  2472  1855  75.0  73.3  76.7  73.3  76.7  73.3  76.7

11. Multiplicity, Missing Data and Analysis Populations

The registry lists 12 outcome measures, including the primary retention outcome and secondary measures of provider performance, competency, satisfaction, adolescent satisfaction, and retention. When a study examines multiple outcomes, there is a general possibility of multiplicity: considering many tests can increase the chance that at least one appears noteworthy by chance. Whether and how to address multiplicity depends on the prespecified hierarchy and analysis plan. The registry record does not report a multiplicity adjustment, an outcome hierarchy beyond identifying the primary endpoint, or a statistical analysis plan that would clarify these choices.

The record lists 0 analyses. It does not report primary or secondary effect estimates, confidence intervals, p-values, or the analysis populations used. The summary says the final primary analysis accounts for facility clustering, time, and newly enrolled status, but does not describe the exact model or analysis set. The number of participants with available outcome data by period, and the extent of missing outcome data, are not reported in the registry record.

In general, missing data can affect an estimate when outcome availability is related to participants’ characteristics, care, or study period. Interpreting a result requires knowing how missingness was assessed and handled, and whether any sensitivity analyses were conducted. The registry does not report those details. It also does not report whether participants were analyzed according to an intention-to-treat principle or another population definition. These omissions prevent assessment of how incomplete records, departures from the intended intervention period, and the set of included participants may have affected the primary comparison.

The registry’s design description notes that a simple individual-level re-analysis would be misleading for this cluster-randomized study. Accordingly, the sponsor’s own analysis takes precedence. The record available for this analysis does not include that analysis’s numerical results or enough methods detail to reproduce it.

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 registry’s safety figures list 2123 people at risk in the SP Training Intervention group and 2472 at risk in the No Training Intervention group. For each group, the recorded count with a serious event is 0, the count with another reported event is 0, and deaths are 0. These are the adverse-event figures recorded for the two periods or groups; they are distinct from the registry’s enrollment figure and should not be interpreted as the number of trial facilities.

In general, a recorded count of zero means that no events of that category are listed in the registry results for the stated group. It does not by itself establish that no event occurred outside the recorded surveillance, or explain how event collection and attribution were conducted. The registry does not provide further detail about safety monitoring, event definitions, reporting procedures, or completeness. The figures also do not establish a comparative safety effect: no adjusted comparison or uncertainty measure is reported, and the trial’s main focus is clinician training and retention in care. These counts should therefore be read as the posted event counts, not as proof of absence of risk.

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

The registry describes a trial designed to evaluate whether clinician training using standardized patient actors is associated with a change in retention in HIV care for adolescents and youth. It gives a specific retention definition and describes a randomized stepped-wedge comparison across facilities. It also names provider and patient-related secondary outcomes. These details explain the research question and the intended measurement approach.

What the registry record does not provide is the primary result. There is no reported retention count by period, effect estimate, confidence interval, or test of the intervention-period comparison. The posted information therefore does not establish the direction, magnitude, or precision of any change in retention. Completion status and the presence of a detailed outcome definition are not substitutes for the missing results.

In general, even a reported difference would need interpretation in light of the design, including clustering by facility and changes over time. The registry says its final analysis accounts for these features and newly enrolled status, but the model and numerical findings are not reported. The sponsor’s cluster-adjusted analysis takes precedence over a simple individual-level recalculation, and that numerical analysis is not reported in the registry record.

The safety figures list zero serious events, zero other events, and zero deaths in each of the two listed groups, with the respective at-risk counts recorded. These are descriptive registry entries; they do not answer the primary retention question or establish the absence of unrecorded events. The listed secondary outcomes likewise describe outcome domains, not findings.

Accordingly, the defensible interpretation is limited: the registry records a completed study with a randomized stepped-wedge design, a clinician-training intervention, and a defined retention endpoint, but it does not report the outcome analysis needed to judge what the trial observed. No claim that the intervention improved, worsened, or had no effect on retention follows from the available results.

14. Why This Trial Matters Statistically

This trial raises statistical issues that are common in health-service research. The intervention is delivered to health care workers at facilities, while the primary outcome is measured among adolescent clients. That structure makes the facility an important unit of the design and analysis. In general, outcomes within a facility may be correlated, so a method that accounts for clustering is important when estimating the intervention-period comparison.

The stepped-wedge structure also makes time part of the comparison. Facilities contribute a control period before training and an intervention period after training, so changes over time must be distinguished from changes associated with the training period. The registry states that the final analysis accounts for time and facility clustering, along with newly enrolled status. This is useful information about the intended analysis, even though the actual model and result are not reported.

The endpoint links a service-delivery intervention to a documented care-engagement event. It combines follow-up after initial enrollment with follow-up after re-engagement, making a clear operational definition important. The registry provides that definition, but does not report how record completeness or missing outcomes were handled.

The case also illustrates why an outcome description and completed-study status cannot stand in for a treatment-effect estimate. Without counts, an estimate, and a measure of uncertainty, readers cannot assess the size or precision of the primary comparison. The registry cautions against individual-level re-analysis in this clustered design, so the sponsor’s own analysis is the appropriate basis for inference; its numerical result is not reported in the registry record. Statistical interpretation is consequently constrained not by the absence of a stated research question, but by the absence of the analysis results needed to answer it.

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 main limitation for interpreting this trial is that the registry record does not report the primary retention results. It gives no counts by intervention and control periods, estimated change, confidence interval, or statistical test. As a result, the direction, magnitude, and precision of the primary comparison cannot be assessed from the posted results.

Important design and analysis details are also absent. Although the record identifies a randomized stepped-wedge design, labels the model as crossover, and says the final analysis accounts for facility clustering, time, and newly enrolled status, it does not specify the model, sequence of site transitions, measurement schedule, or method for estimating uncertainty. The registry also cautions that a simple individual-level re-analysis would be misleading; without the sponsor’s numerical analysis, that analysis cannot be evaluated.

The enrollment figure is recorded as 24, while the study summary describes 24 facilities. The unit represented by the enrollment figure is not clarified in the available record. Details about participant flow and the number contributing primary outcome data are not reported. The registry does not explain how missing, incomplete, or conflicting medical records were handled, or define the analysis population.

Although 12 outcome measures are listed, results and multiplicity procedures are not reported. The safety entries show zero recorded events in the listed categories, but the record does not describe event ascertainment or reporting completeness. These limitations concern what can be inferred from the registry information; they do not establish whether additional analyses or documentation exist elsewhere.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Patient Actor Training to Improve HIV Services for Adolescents in Kenya — NCT02928900.
• Registry record last updated 2024-12-09; status last verified 2024-10.
• Results first posted 2024-12-09.
• 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 describes NCT02928900, titled “Patient Actor Training to Improve HIV Services for Adolescents in Kenya,” with an official title referring to simulated patient encounters and engagement in HIV care. The sponsor is the University of Washington. The study is marked completed, with a start date of 2016-09-26 and primary completion date of 2020-11-02.

The design is described as randomized and stepped-wedge at 24 HIV care and treatment facilities. The behavioral intervention is clinician training using standardized patient actors; the control period is standard of care before training. The primary outcome is retention in HIV care, assessed through medical record review and defined by follow-up after enrollment or re-engagement.

The registry lists secondary outcomes across health care worker performance, competency and satisfaction, adolescent satisfaction, and retention. It does not report the primary analysis results, effect estimate, uncertainty measure, or detailed statistical model. Thus the research question and design are documented, but the posted results do not permit an assessment of the primary comparison.