Reducing Post-Hospital Mortality in HIV-infected Adults in Tanzania: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Reducing Post-Hospital Mortality in HIV-infected Adults in Tanzania (NCT03858998), a randomized na study. The primary outcome is Number of Participant Deaths in the First 12 Months Post-hospitalization.

2. Clinical Question

The trial addresses mortality following hospitalization among adults living with HIV in Tanzania. Its focus is not only whether participants are linked with local HIV clinics, but whether a case management approach is associated with a different count of deaths during the stated follow-up period. The registry also lists secondary outcomes concerning clinic attendance, ART adherence, viral suppression, and psychosocial or belief-related measures. Those outcomes can help describe other aspects of care and experience, but they are not the primary mortality endpoint.

The primary comparison is between the two randomized groups. The posted results record 43 deaths in the Linkage Case Management Intervention group and 42 in the Enhanced Standard of Care group, with 250 participants listed in each group. These observed counts are the descriptive basis for the reported comparison. They do not by themselves settle whether the groups differ beyond the uncertainty compatible with the data; interpretation also depends on the effect estimate and its confidence interval.

3. Trial Design

The registry describes a randomized, double-masked, parallel-group trial, with phase listed as NA. The enrollment field is 500, and the record lists 250 participants in each arm. The study status is completed. The sponsor is Weill Medical College of Cornell University. The registry lists a start date of 2019-03-05 and a primary completion date of 2023-05-31.

Participants in the intervention arm were assigned to a 90-day case management intervention intended to link hospitalized adults with HIV to local HIV clinics. The comparator is described in the arm description as current routine HIV care in Tanzania. In the posted outcome data, the comparator arm is named Enhanced Standard of Care. The registry does not explain whether these phrases refer to the same care condition or give further detail about the care delivered in either arm.

Random allocation creates a design intended to make the groups comparable at the point of assignment, while masking is intended to limit some forms of influence on trial conduct or assessment. Those are general features of randomized and masked trials; the registry record does not identify who was masked, how allocation was generated or concealed, or how masking was maintained. The parallel design means the record presents separate groups rather than a crossover comparison.

The primary outcome is the number of participants who died in the first 12 months after hospitalization. Death was to be determined through phone calls to relatives and confirmed by verbal autopsies, obituaries, hospital records, or death certificates. The record lists 13 outcome measures and 12 secondary outcomes as posted. It names outcomes including HIV clinic attendance, ART adherence, viral suppression, traditional health beliefs, self-efficacy, stigma, social support, and perceived need for HIV services. Further protocol details, including eligibility criteria, recruitment procedures, and the original statistical design assumptions, 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.

Overall Study

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 Number of Participant Deaths in the First 12 Months Post-hospitalization, measured over 12 months.

The primary endpoint counts participants who died during the first 12 months after hospitalization. The registry describes death ascertainment through phone calls to relatives, with confirmation by verbal autopsies, obituaries, hospital records, or death certificates. The posted endpoint is a count of participants with the event, rather than a time-to-event estimate or a measure of how long participants survived.

Mortality is the outcome of direct interest in a study whose stated aim is to assess whether case management can improve the one-year mortality rate among hospitalized adults with HIV in Tanzania. A count over a defined follow-up period makes the result interpretable as the occurrence of death during that period. The registry also reports group sizes and a risk-difference analysis, which compares the observed event risks between the two arms.

As a general statistical point, an endpoint needs a clear definition and ascertainment process so that events are counted consistently. The listed confirmation sources provide the registry's description of how deaths were to be determined, but the record does not report how often each source was used, whether ascertainment was complete for all participants, or how uncertain death reports were handled. Those details matter when assessing how fully the endpoint captures deaths during follow-up.

7. Results Posted in the Registry

• Linkage Case Management Intervention: 17.2% (43/250) versus 16.8% (42/250) on control; risk difference 0.4 percentage points (95% CI -6.2 to 7.0), risk ratio 1.02 (0.69–1.51); unadjusted two-proportion z-test p = 0.905.

8. Statistical Interpretation

9. Statistical Methods Explained

The primary analysis compares a binary event: whether a participant died during the first 12 months after hospitalization. The registry reports the event count in each randomized group and identifies the sponsor's method as Chi-squared, with the parameter reported as a risk difference. A risk difference is an absolute contrast between the event risks in two groups. It is expressed on the risk scale, so zero represents equal observed risks, positive values indicate a higher risk in the intervention group under the stated comparison orientation, and negative values indicate a lower risk.

What does the risk difference describe?

The posted risk estimates are 0.172 in the intervention group and 0.168 in the comparator group, and the reported difference is 0.004. This expresses the contrast between the observed group risks without adjusting for other variables. It is not a relative risk, odds ratio, or time-to-event estimate. The comparison also lists a relative risk of 1.024 with a confidence interval from 0.695 to 1.508, and an odds ratio of 1.029 with a confidence interval from 0.645 to 1.641. These are different ways to express a group contrast, and their numerical scales and interpretations differ.

As a general statistical explanation, risk differences are often useful when the question concerns the absolute separation between groups. Relative measures instead compare risks multiplicatively or compare odds. Different measures can convey the same underlying event data from different perspectives; none should be interpreted without noting which measure was selected. The registry's reported primary comparison is the risk difference.

What does the Chi-squared test assess?

The sponsor identifies a Chi-squared test for the comparison. As a general concept, this test assesses whether the pattern of event and non-event counts across groups is inconsistent with an assumption of no association between group and outcome. A p-value summarizes the compatibility of the observed data with that reference model and its assumptions. The reported p-value is 0.91. The unadjusted re-calculation from posted summary statistics is 0.905. A p-value does not quantify the probability that the null hypothesis is true, and it does not tell whether an estimated effect is clinically important.

The trial result also includes a confidence interval for the risk difference. As a general explanation, a confidence interval gives a range of effect estimates compatible with the data and the method used, at the stated confidence level. It conveys uncertainty in the estimate, whereas the point estimate alone gives only one value. The sponsor reports a 95% interval from -0.06 to 0.07; the unadjusted re-calculation is from -0.062 to 0.070. The intervals cross zero, so the data are compatible with no risk difference as well as with effects in either direction.

Why consider over-dispersion?

Over-dispersion is a general concern when count variation is greater than a model expects. It can make standard errors too small if the model does not account for that extra variation. It is especially relevant for count outcomes analyzed with models that impose a particular variance structure. The registry's primary endpoint is a count of participants with an event, and its reported analysis is a Chi-squared comparison of the two groups. Dispersion is listed as null, and the record does not report an over-dispersion assessment or a model-based adjustment. It would therefore be inappropriate to infer that a particular dispersion model was used.

What does the re-calculation add?

The re-calculated estimates are simple, unadjusted calculations from the posted summary statistics. They provide a check on the reported group contrast using the displayed event counts and group sizes, rather than a substitute for the sponsor's complete analysis. The recalculated risk difference is 0.004 and its interval is close to the sponsor's rounded interval. The p-values are also close at the precision displayed. The registry does not flag a discrepancy between the posted analysis and the re-calculation.

General statistical interpretation still depends on design and conduct. For example, adjustment for baseline variables, treatment of missing outcomes, and the population included in the analysis can affect estimates and their uncertainty. The registry record does not report these details for the primary analysis. It also does not provide enough information to determine whether the analysis followed an intention-to-treat approach or whether any additional analyses were conducted outside the reported comparison.

10. Confidence Intervals and Precision

The sponsor reports a 95% confidence interval for the risk difference from -0.06 to 0.07. The unadjusted re-calculation from the posted summary statistics gives an interval from -0.062 to 0.070. Both intervals include zero, the value corresponding to no observed difference on the risk-difference scale. The endpoints also extend in both directions, so the data are compatible with a lower event risk in the intervention group or a higher event risk.

As a general explanation, a confidence interval represents uncertainty around an estimate under the statistical method used. It should not be read as a range containing a fixed proportion of individual participant outcomes, nor as a probability statement about the particular interval after it has been calculated. The width of an interval is informative about precision: a narrower interval indicates a more tightly estimated effect, while a wider interval leaves more possible values consistent with the data.

Here, the point estimate is 0.004, but the interval spans values on either side of zero. Thus the point estimate alone gives an incomplete picture of the result. The confidence interval does not support a precise conclusion about the direction or magnitude of the group difference. The sponsor's and re-calculated intervals differ slightly in their displayed limits because they are reported at different precision; the registry does not identify a discrepancy.

Precision is also shaped by outcome frequency, group size, event ascertainment, and analysis choices as a general matter. The posted groups each list 250 participants and the event counts are 43 and 42. The registry record does not report a separate precision target for the original design, loss-to-follow-up details, or a planned confidence-interval method beyond the reported analysis. These absences limit assessment of how the observed precision compares with the trial's original aims.

11. Planning a Similar Trial: Sample Size

This is a planning illustration, not a re-evaluation of the original trial. It asks: if a new study wanted to detect a difference like the one observed, or a smaller one, how large would it need to be? It is deliberately not “observed power” of the finished trial, which is not a meaningful quantity.

The sample-size figures are planning illustrations based on the observed effect, not a report of the trial's original design calculation. For this binary endpoint, the displayed values are participants per arm, not total events. The planning assumptions are a two-sided significance level of 0.05, equal allocation between arms, and a normal approximation; the effect to detect is the stated fraction of the observed effect. The trial's own original design assumptions are not reported in the registry record. These figures are not observed power.

The 80% power column gives the illustrative number of participants per arm for a design that would have 80% power under the listed assumptions and effect size. The 90% power column gives the corresponding number per arm for 90% power. Power is a planning concept: it describes the chance that a specified test would detect an assumed effect if that effect were present and the model assumptions held. It is not a measure of whether the observed result is true.

The rows consider planning for 100% of the observed effect, 75% of the observed effect, and 50% of the observed effect. For 100% of the observed effect, the figures are 138,434 participants per arm for 80% power and 185,323 participants per arm for 90% power. For 75% of the observed effect, the figures are 245,529 and 328,693 participants per arm, respectively. For 50% of the observed effect, the figures are 551,144 and 737,824 participants per arm, respectively.

The curve is a visual way to follow how the planning requirement changes across effect assumptions. Smaller effects need more participants to detect under the same assumptions, because a smaller separation between groups is harder to distinguish from random variation. These illustrations use the observed effect as their reference point, so they should not be treated as a recommended sample size or as evidence that the completed trial had a particular probability of success.

12. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

n is the number of participants in each arm, p₀ and p₁ are the control and treatment proportions, and p̄ is their average.

Worked calculation for this trial

1. Control proportion p₀ = 0.168; treatment proportion p₁ = 0.172; average p̄ = 0.170.
2. First term: z_1−α/2 × √[2p̄(1−p̄)] = 1.96 × 0.5312 = 1.0412.
3. Second term: z_1−β × √[p₀(1−p₀) + p₁(1−p₁)] = 0.842 × 0.5312 = 0.4471.
4. n = (1.0412 + 0.4471)² / (0.0040)² = 138,433.4, rounded up to 138,434 participants per arm for 80% power.

The smaller-effect rows in the table above repeat the same calculation with the effect multiplied by 0.75 and 0.5 (for ratios, applied to the logarithm of the ratio).

R code

alpha <- 0.05
z_a   <- qnorm(1 - alpha / 2)          # two-sided
p0 <- 0.168                          # control proportion
p1 <- 0.172                          # treatment proportion
n_per_arm <- function(power, frac = 1) {
  q1 <- p0 + (p1 - p0) * frac               # smaller effect = fraction of the observed difference
  pbar <- (p0 + q1) / 2
  z_b <- qnorm(power)
  ceiling(round((z_a * sqrt(2 * pbar * (1 - pbar)) + z_b * sqrt(p0 * (1 - p0) + q1 * (1 - q1)))^2 / (q1 - p0)^2, 6))
}
for (f in c(1, 0.75, 0.5))
  cat(sprintf("%3.0f%% of observed effect: n per arm = %s (80%% power), %s (90%% power)\n", 100 * f,
              format(n_per_arm(0.80, f), big.mark = ","), format(n_per_arm(0.90, f), big.mark = ",")))

# Check against R's built-in function (unrounded n per arm)
cat("power.prop.test:", round(power.prop.test(p1 = p0, p2 = p1, power = 0.80)$n, 1), "\n")

# Simulation: how often does the z-test reject at the 80%-power sample size?
set.seed(3858998)
n    <- n_per_arm(0.80)
reps <- 20000
x0 <- rbinom(reps, n, p0); x1 <- rbinom(reps, n, p1)
ph <- (x0 + x1) / (2 * n)
z  <- ((x1 - x0) / n) / sqrt(ph * (1 - ph) * 2 / n)
cat(sprintf("Simulated power at n = %s per arm: %.3f (%d replicates)\n", format(n, big.mark = ","), mean(abs(z) > z_a, na.rm = TRUE), reps))

Output from running the code

100% of observed effect: n per arm = 138,434 (80% power), 185,323 (90% power)
 75% of observed effect: n per arm = 245,529 (80% power), 328,693 (90% power)
 50% of observed effect: n per arm = 551,144 (80% power), 737,824 (90% power)
power.prop.test: 138433.4 
Simulated power at n = 138,434 per arm: 0.798 (20000 replicates)

The simulation draws many pairs of event counts at the planned sample size and counts how often the two-proportion z-test rejects the null. It should land close to the target power; at very small required sample sizes it can fall below it, because the normal approximation is rough there.

13. Analyses Reported by the Sponsor

The sponsor's own posted analysis identifies a Chi-squared method and reports a risk difference of 0.004, a 95% confidence interval from -0.06 to 0.07, and p = 0.91. A simple unadjusted re-calculation from the posted event counts and group sizes gives the same risk difference at the displayed precision, a 95% interval from -0.062 to 0.070, and p = 0.905.

The small differences in the interval limits and p-value reflect the precision at which results are displayed. The registry does not flag a discrepancy between the sponsor analysis and the re-calculation. The re-calculation is a check based on posted summary statistics; it cannot reproduce unreported adjustments, analysis-population choices, or other details of the sponsor's full analytic process. The registry record does not report such additional details for this comparison.

Accordingly, the sponsor's reported result remains the primary account of the trial's analysis, while the re-calculation offers a basic consistency check. Both indicate a small positive point estimate with an interval extending across zero. Neither should be interpreted without the uncertainty conveyed by the confidence interval.

14. Multiplicity, Missing Data and Analysis Populations

The registry lists 13 outcome measures and 12 secondary outcomes as posted, in addition to the primary mortality endpoint. Named secondary outcomes include HIV clinic attendance, ART adherence, viral suppression, traditional health beliefs, self-efficacy, stigma, social support, and perceived need for HIV services. The record lists 1 analysis. It does not describe whether the secondary outcomes were tested under a prespecified hierarchy, whether multiplicity was controlled, or how secondary findings should be treated relative to the primary outcome.

As a general statistical point, testing multiple outcomes can increase the chance of observing at least one apparently unusual result even when there is no underlying effect on any outcome. Methods such as adjustment or a prespecified testing hierarchy can address that issue, but the registry record does not report whether such methods were used. The primary mortality result should therefore be distinguished from any interpretation of the broader set of outcomes.

The record does not report how missing outcome data were handled, the amount of missingness, or whether all participants were followed through the stated 12-month period. It describes several sources for confirming deaths, but it does not state whether the status of every participant was known. It also does not report a sensitivity analysis for uncertain or missing death status.

The analysis population is not described in the registry record. In particular, the record does not state whether the primary comparison included all randomized participants, only participants with an observed outcome, or another defined population. It does not describe exclusions, protocol deviations, adherence-based analyses, or imputation. These are material details because the analysis population and missing-data approach can influence the interpretation of a randomized comparison.

Other features of trial conduct and analysis are likewise not reported, including allocation concealment, the identities of people masked, and the handling of any interim analyses. Their absence from the registry record should not be taken as evidence that they did or did not occur. The record supports interpretation of the posted comparison, but not a full reconstruction of the statistical analysis plan or trial conduct.

15. 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 posted adverse-event fields list 250 participants at risk in each group, with 0 serious adverse events and 0 other adverse events recorded for the Linkage Case Management Intervention group and the Enhanced Standard of Care group. The same fields list 43 deaths for the intervention group and 42 for the comparator group. The death counts are also the primary outcome counts.

These figures should be read as registry entries for the specified adverse-event fields, not as proof that no adverse experiences occurred. The registry does not provide serious adverse-event terms, and the serious-event term list is empty. It also does not describe the event definitions, reporting period, collection procedures, relatedness assessment, severity grading, or completeness of adverse-event reporting. Without those details, the meaning of a zero entry is limited to what was recorded in the posted fields.

Death is reported separately as an outcome count, so the presence of deaths in the primary endpoint is not inconsistent with zero entries in the serious and other adverse-event fields. The record does not explain the relationship between outcome ascertainment and adverse-event reporting. The posted safety fields therefore provide a narrow account and should not be used alone to characterize the full safety experience.

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

The result describes mortality recorded during the first 12 months after hospitalization in two randomized groups. The intervention group had 43 recorded deaths among 250 participants, and the enhanced standard-of-care group had 42 among 250. The risk difference was 0.004, with the sponsor's 95% confidence interval extending from -0.06 to 0.07. The point estimate is close to zero, and the interval leaves room for differences in either direction.

Statistically, the posted estimate does not precisely distinguish a benefit, harm, or no difference in mortality. That is a statement about the information and uncertainty in this comparison, not a claim that the intervention does or does not work. The p-value of 0.91 is not the probability that the intervention has no effect. It is a test statistic interpreted under the test's assumptions and the specified no-difference reference.

The finding applies to the trial's comparison as recorded: a 90-day linkage case management intervention versus a comparator described as current routine HIV care and named Enhanced Standard of Care. The registry does not clarify the comparator wording or provide enough detail to determine how closely the study conditions match other care settings. Results should not automatically be generalized to different populations, services, follow-up arrangements, or definitions of care.

The endpoint counts deaths, but the registry does not report the completeness of follow-up, missing outcome handling, the analysis population, or the exact statistical analysis plan. Those limitations matter when judging how robustly the posted estimate represents the randomized comparison. Nor does this primary outcome alone describe clinic linkage, ART adherence, viral suppression, or other listed outcomes.

In short, the posted data show similar observed group counts and a risk-difference estimate near zero, with substantial uncertainty around the direction and size of any difference. The result does not establish equivalence, rule out meaningful effects, or provide a definitive conclusion about outcomes beyond this trial's reported endpoint and follow-up period.

17. Why This Trial Matters Statistically

This trial illustrates why a point estimate and a confidence interval need to be read together. The posted risk difference is 0.004, while its 95% confidence interval ranges from -0.06 to 0.07 in the sponsor analysis. A point estimate near zero may appear straightforward, but an interval extending in both directions shows that the data do not pin down the direction or size of the underlying difference precisely.

The comparison also demonstrates how different statistical measures answer related but non-identical questions. The risk difference describes the absolute separation in event risks, while the relative risk and odds ratio express the comparison on other scales. The registry reports all three, but the primary sponsor analysis uses the risk difference. Readers should keep the measure in view rather than treating all effect estimates as interchangeable.

Mortality after hospitalization is a clear, consequential outcome, yet its interpretation depends on reliable follow-up and event ascertainment. The registry describes multiple sources for confirming death, while leaving the completeness of ascertainment and handling of missing information unreported. Statistical precision cannot compensate for uncertainty about whether outcomes were captured consistently.

The planning illustrations further show a general design principle: detecting smaller effects requires more participants under otherwise unchanged assumptions. They are based on the observed effect and are not the original trial design calculation. This distinction matters because a post hoc planning illustration cannot establish what the study was designed to detect or how likely its result was to reach a threshold.

Finally, the record lists multiple secondary outcomes alongside the primary endpoint. That breadth can support a more complete description of care and participant experience, but it also raises questions about multiplicity and which analyses were prespecified. The registry does not report a multiplicity strategy. Careful interpretation therefore requires separating the primary mortality comparison from other outcomes and recognizing the limits of the posted analysis details.

18. 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 provides a concise account of the trial but does not report several details needed for a fuller assessment. Eligibility criteria beyond the description of hospitalized adults with HIV in Tanzania are not available. The record also does not report recruitment locations, baseline characteristics, the allocation-generation process, allocation concealment, or which people were masked. Although the design is labeled randomized, double-masked, and parallel, those labels do not reveal how the procedures were implemented.

The comparator is described as current routine HIV care in Tanzania, while the posted arm title is Enhanced Standard of Care. The registry does not reconcile those descriptions. This makes it difficult to characterize the care contrast beyond the wording attached to each record field.

Follow-up and outcome completeness are also unclear. The registry gives the endpoint definition and death-confirmation sources, but not the amount of missing follow-up, the completeness of ascertainment, or the rules for resolving uncertain cases. It does not state how missing outcomes were handled, which participants were included in the primary analysis, or whether the analysis followed an intention-to-treat principle.

The original sample-size rationale, statistical analysis plan, planned adjustment variables, interim analyses, and multiplicity approach are not reported in the registry record. The sample-size illustrations are based on the observed effect and cannot fill in those missing design details. The record lists secondary outcomes but does not provide their methods or results in the registry record for this analysis. These limits constrain interpretation of the primary result and prevent a comprehensive evaluation of the full trial report.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: Reducing Post-Hospital Mortality in HIV-infected Adults in Tanzania — NCT03858998.
• Registry record last updated 2025-04-23; status last verified 2025-04.
• Results first posted 2024-05-02.
• 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.

22. Record Summary

The registry record describes a completed, randomized, double-masked, parallel-group trial sponsored by Weill Medical College of Cornell University. Enrollment is listed as 500, with 250 participants in each arm. The trial began on 2019-03-05 and has a primary completion date of 2023-05-31. Phase is listed as NA.

The intervention was a 90-day case management approach intended to link hospitalized adults with HIV in Tanzania to local HIV clinics. The comparator is described as current routine HIV care and is named Enhanced Standard of Care in the posted results. The primary endpoint counts deaths during the first 12 months after hospitalization.

The posted counts are 43 deaths in the intervention group and 42 in the comparator group. The sponsor reports a risk difference of 0.004, a 95% confidence interval from -0.06 to 0.07, and p = 0.91. The interval spans zero, so the estimate is imprecise about the direction and magnitude of a group difference. Details about missing data, analysis populations, and the original statistical plan are not reported in the registry record.