Two Approaches to Routine HIV Testing in a Hospital Emergency Department: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Two Approaches to Routine HIV Testing in a Hospital Emergency Department (NCT01258582), a randomized na study. The primary outcome is Acceptability of the HIV Test.

2. Clinical Question

The registry describes the study’s second phase as examining differences in acceptability according to the testing method offered. Its summary also refers to an initial phase that compared approaches to routine HIV counseling, testing, and referral services, and says that phase was closed in July 2008. The reported primary outcome for this phase is acceptance among randomized participants in each arm. The available record does not give further detail about the emergency department’s workflow, participant eligibility, or how an offer of testing was presented.

3. Trial Design

This was a randomized, parallel-group trial with no masking. The two reported arms were fingerstick HIV testing and oral HIV testing. The registry lists enrollment of 1651 and reports arm sizes of 830 for fingerstick and 821 for oral testing. The listed phase is NA. The study is marked completed, and Massachusetts General Hospital is listed as sponsor.

The registry gives a start date of 2009-05 and a primary completion date of 2010-01. It describes this as the second phase of the Optimizing Strategies for Universal HIV Testing (The USHER Trial), and identifies the second phase’s focus as acceptability by testing method. The summary also says the initial phase was closed in July 2008. More detail about the relationship between the phases, the recruitment process, and the transition between them is not reported in the registry record.

Randomization means that participants were assigned to one of the two testing approaches according to an allocation process intended to support a fair comparison. The allocation ratio and details of sequence generation or concealment are not reported in the registry record. With no masking, participants and study staff were not masked to the assigned testing approach, according to the listed design. The record does not explain whether outcome assessment involved any additional masking.

The primary outcome was the proportion of participants who accepted HIV testing within their assigned arm. The endpoint description specifies an intent-to-treat analysis and assessment on the day of enrollment. The registry reports one outcome measure and one analysis. Further details about the statistical analysis plan, the original design assumptions, or planned interim analyses 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

Analysis population for the primary outcome, as defined in the registry: Intent to treat analysis

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 Acceptability of the HIV Test, measured over Assess on day subject enrolled into the study.

The primary endpoint, Acceptability of the HIV Test, measures the proportion of participants in each randomized arm who accepted HIV testing. The registry defines the outcome as acceptance among participants randomized to fingerstick or oral-fluid testing and gives the timeframe as the day the participant enrolled. This is a binary outcome at the participant level: a participant either accepted the offered HIV test or did not. The posted arm results are proportions, with 0.67 reported for fingerstick testing and 0.69 for oral testing.

Acceptance is used to assess whether the testing approach offered is acceptable to participants in this emergency department setting. It is distinct from whether a test identified an infection, how accurate a test was, or whether participants completed referral or follow-up. Those outcomes are not the primary endpoint described in the registry record. The summary’s second-phase objective specifically concerns differences in acceptability based on the method offered.

The stated intent-to-treat population links the endpoint comparison to randomized assignment. The registry does not provide further operational details about what counted as acceptance, how declined or incomplete offers were handled, or how missing outcome information was treated. Those details matter when interpreting a proportion because they determine which participant records contribute to its numerator and denominator.

7. Results Posted in the Registry

• Oral HIV Testing: 0.7% (6/821) versus 0.7% (6/830) on control; risk difference 0.0 percentage points (95% CI -0.8 to 0.8), risk ratio 1.01 (0.33–3.12); unadjusted two-proportion z-test p = 0.985.

8. Statistical Interpretation

9. Statistical Methods Explained

The registry describes a randomized, parallel-group comparison. Each participant was assigned to fingerstick or oral HIV testing, and the primary outcome was the proportion who accepted testing in the assigned arm. In a parallel-group design, each participant contributes to one assigned group rather than crossing over between approaches. Randomization is intended to make groups comparable on average, although the registry record does not report the allocation ratio, the sequence-generation procedure, or allocation concealment.

What does the endpoint measure?

The outcome is binary: acceptance or non-acceptance of the offered HIV test. The arm-level results are proportions, reported as 0.67 for fingerstick and 0.69 for oral testing. A proportion describes the share of participants counted as having the outcome within an arm. Its interpretation depends on the rules used to classify acceptance and on which participants have usable outcome data. The registry says the primary analysis is intent to treat, but it does not explain the handling of missing outcomes or deviations from assigned testing.

General: For a binary endpoint, comparing proportions is one way to describe how often an outcome occurred in two groups. A difference on the absolute scale describes how far apart the proportions are. Other measures, such as a relative risk or odds ratio, answer different questions and can appear more or less pronounced depending on how common the outcome is. The sponsor’s listed parameter is the difference in acceptance rates.

How should the sponsor analysis be read?

The registry reports a chi-squared analysis of the difference in acceptance rates. The sponsor-reported estimate is -0.022, its 95% confidence interval runs from -0.067 to 0.023, and its p-value is 0.34. General: A chi-squared test evaluates whether observed group outcome frequencies are compatible with a null hypothesis of no association between group and outcome, subject to the test’s assumptions. The registry does not give additional detail about the implementation of this test, such as any continuity correction or adjustment for other variables.

A p-value is calculated under a specified null hypothesis and statistical model. In general, a larger p-value means the data are less surprising under that null model than a smaller p-value would be; it is not a measure of the probability that the null is true. The sponsor result does not cross a conventional threshold for statistical significance, but that does not prove there is no difference. The confidence interval is important because it shows a range of values compatible with the data and method, including both zero and nonzero differences.

What is the separate recalculation?

A separate unadjusted recalculation reports a risk difference of < 0.001, a 95% confidence interval from -0.008 to 0.008, and p = 0.985. These figures are not aligned with the sponsor’s reported estimate and confidence interval. The registry record does not reconcile the difference. The recalculation is not adjusted for other variables, and it should not be treated as if it were the sponsor’s analysis. The record does not describe covariate adjustment in the sponsor analysis either.

General: An unadjusted comparison uses the reported group outcome information without accounting for additional covariates. Adjusted analyses can address measured factors, but they depend on the variables and model specified. No adjusted estimate is reported in the registry record. When two reported analyses do not agree numerically, understanding the endpoint coding, analysis population, and data used is necessary; those details are not available in the registry record.

What does a confidence interval add?

General: A confidence interval conveys uncertainty around an estimate under a statistical procedure. It is not a guarantee that the true value lies in the particular interval, and it is not a range containing a fixed proportion of individual outcomes. Wider intervals generally indicate less precision, while narrower intervals indicate greater precision for the parameter and model at issue. The sponsor’s interval for the acceptance-rate difference includes zero and also includes values on either side of zero. The arm-specific intervals describe uncertainty in each arm’s proportion, not uncertainty in the between-arm difference.

The registry identifies the primary analysis population as intent to treat. General: In an intent-to-treat analysis, participants are generally analyzed according to randomized assignment, even if the treatment or procedure received differs from assignment. The exact implementation can vary, particularly when outcomes are missing. The record does not explain whether every randomized participant contributed an observed primary outcome or how missingness was addressed, so the label alone does not resolve those questions.

The record lists one primary outcome measure and one analysis, and no secondary outcomes are posted. It does not report whether any multiplicity strategy was planned or whether there were interim analyses. The statistical plan, including the original assumptions used to design the trial, is not reported in the registry record. Those omissions limit assessment of how closely the posted analysis followed a prespecified plan.

10. Confidence Intervals and Precision

Precision concerns how narrowly the data estimate an underlying outcome or difference. The registry reports a 95% confidence interval of 0.64 to 0.7 for fingerstick acceptance and 0.66 to 0.72 for oral acceptance. These intervals describe uncertainty in the arm-specific proportions. They should not be interpreted as ranges of individual acceptance outcomes, and they do not directly answer how precisely the difference between arms has been estimated.

For the between-arm comparison, the sponsor reports a difference of -0.022 and a 95% confidence interval from -0.067 to 0.023. Because the interval includes zero, the data are compatible with no difference under this analysis. It also includes values on either side of zero, so the result does not rule out a difference in either direction. Whether the interval rules out differences that would matter in practice cannot be judged from the registry record because no threshold for an important difference is reported.

The unadjusted recalculation has a risk difference of < 0.001 and a 95% confidence interval from -0.008 to 0.008. This interval is not aligned with the sponsor’s interval. The registry record does not explain why the estimates differ, which limits a unified interpretation of precision. The two confidence intervals refer to separate analyses and should not be combined or treated as interchangeable.

General: A confidence interval depends on the data, the quantity being estimated, and the statistical method. It summarizes sampling uncertainty under that method, not every source of uncertainty. It does not account automatically for unclear outcome definitions, missing data, selection effects, or differences between the posted analyses. Here, the record’s limited explanation of the divergence between the sponsor analysis and recalculation is an important qualification when assessing precision.

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, not observed trial outcomes. Because the endpoint is a binary proportion and the planning description specifies that all other outcome types use participants per arm, these figures are participants per arm, not total events. The planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The effect to detect is set to the stated fraction of the observed effect. These figures do not describe the trial’s own original design assumptions, which are not reported in the registry record.

The 80% power column gives the number of participants per arm in a hypothetical study designed to have an 80% chance of detecting the specified effect under the planning assumptions, if that effect is present and the statistical model applies. The 90% power column uses a higher target chance of detection and therefore gives a different planning requirement. Neither column reports a probability calculated after the trial about whether its result was detected.

The rows for 100%, 75%, and 50% of the observed effect illustrate how planning changes when the target effect is reduced. Smaller effects are harder to distinguish from random variation, so they require more participants under otherwise fixed assumptions. The curve should be read in the same planning context: it illustrates the relationship between the assumed effect and the participant requirement, rather than describing the observed acceptance results. These figures are not observed power.

Very large planning requirements can result when the assumed effect is extremely small. They should be interpreted as consequences of the stated inputs and approximation, not as a claim that such a study was planned or would necessarily be appropriate. The registry does not report the trial’s original sample-size justification or the effect size the original investigators intended to detect.

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 = 0.007; treatment proportion p₁ = 0.007; average p̄ = 0.007.
2. First term: z_1−α/2 × √[2p̄(1−p̄)] = 1.96 × 0.1201 = 0.2355.
3. Second term: z_1−β × √[p₀(1−p₀) + p₁(1−p₁)] = 0.842 × 0.1201 = 0.1011.
4. n = (0.2355 + 0.1011)² / (0.0001)² = 18,037,298.2, rounded up to 18,037,299 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 <- 0.0072289156626506                          # control proportion
p1 <- 0.00730816077953715                          # 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(1258582)
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 = 18,037,299 (80% power), 24,146,825 (90% power)
 75% of observed effect: n per arm = 32,022,929 (80% power), 42,869,615 (90% power)
 50% of observed effect: n per arm = 71,953,981 (80% power), 96,325,966 (90% power)
power.prop.test: 18037298 
Simulated power at n = 18,037,299 per arm: 0.799 (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 analysis reports a chi-squared comparison of acceptance rates, with an estimated difference of -0.022, a 95% confidence interval from -0.067 to 0.023, and p = 0.34. This is the registry’s reported analysis of the primary acceptance outcome. Its interval includes zero, and its p-value does not provide clear statistical evidence of a difference under the reported test. The sponsor analysis does not establish equivalence.

A separate unadjusted recalculation gives a risk difference of < 0.001, with a 95% confidence interval from -0.008 to 0.008 and p = 0.985. The estimates and intervals differ from the sponsor’s reported result, and the registry record does not explain why. The recalculation is a distinct analysis of posted summary information; it should not be presented as the sponsor’s result or as a correction to it. The difference between the two analyses makes it especially important to identify which result is being discussed.

The registry does not provide enough detail to reconcile the analyses, including a fuller account of the data and coding used for each. The sponsor’s reported analysis is the direct posted analysis of the difference in acceptance rates. The recalculated result can be reported transparently alongside it, with the unresolved difference noted, rather than used to imply greater certainty about the treatment comparison.

14. Multiplicity, Missing Data and Analysis Populations

The registry lists one primary outcome measure and one reported analysis, and it reports no secondary outcomes. A single listed outcome does not establish that no other analyses were considered or conducted; the record simply does not provide additional outcomes. Multiplicity procedures, such as adjustments for testing several outcomes or comparisons, are not reported in the registry record. Nor are interim analyses or stopping rules reported. It is therefore not possible to assess whether the posted p-value was part of a broader set of planned tests.

The primary analysis population is described as intent to treat. This indicates that the stated analysis was intended to follow randomized assignment, but the registry does not explain how that principle was implemented for participants with missing outcome data or other departures from assignment. The record does not report the number of participants with missing acceptance outcomes, the reasons for missingness, or an imputation or sensitivity-analysis approach. It also does not give a participant-flow account that would clarify exclusions from the primary analysis.

The arm results include reported arm sizes of 830 for fingerstick and 821 for oral testing. The registry’s endpoint description says the outcome is the proportion accepting testing among those randomized within each arm, but it does not explain in detail how the displayed denominators relate to outcome availability. The posted results should therefore be read with the stated intent-to-treat label while recognizing that missing-data handling is not documented.

The record does not report a prespecified statistical analysis plan, rules for choosing among analyses, covariate-adjustment plans, or the handling of protocol deviations. These omissions do not show that any particular issue occurred; they limit what can be verified from the registry record. The divergence between the sponsor-reported result and the separate unadjusted recalculation is an additional reason to avoid treating either result as a complete account of every analytical choice.

15. Safety Results

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

Other (non-serious) adverse events

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

Most frequent non-serious adverse event terms

The registry lists serious adverse events for the fingerstick and oral-fluid groups. For fingerstick, 0 serious events are reported among 553 participants at risk; for oral fluid, 0 serious events are reported among 565 participants at risk. These figures describe the serious-event entries recorded in the adverse-event data, rather than the primary acceptance outcome.

Other adverse events are listed as 2 among 553 at risk in the fingerstick group and 1 among 565 at risk in the oral-fluid group. Deaths are not reported in the registry record. No top serious adverse-event terms are listed. The registry does not provide further detail about event definitions, severity grading beyond the serious-event classification, relatedness to the testing procedure, or how the safety population was defined.

These figures should be interpreted as the adverse-event information posted for the listed groups, not as evidence that no other events occurred outside the reporting scope. The at-risk counts differ from the arm sizes reported for the primary outcome, and the record does not explain the relationship between those populations. Small counts also provide limited information for comparing uncommon events. The safety entries do not change the interpretation of the primary acceptance analysis, which addresses a different outcome.

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

The result describes how often participants in each randomized arm accepted the HIV test offered in this trial. The posted proportions are 0.67 for fingerstick and 0.69 for oral testing. The sponsor’s analysis reports a difference of -0.022, with a 95% confidence interval from -0.067 to 0.023 and p = 0.34. On that analysis, the data do not provide clear statistical evidence of a difference between the approaches.

A result that does not provide clear evidence of a difference is not proof that the approaches have identical acceptance. The confidence interval includes zero but also includes nonzero differences. Whether those possible differences would be important depends on a threshold that the registry does not report. The trial is not described as an equivalence or non-inferiority study, so absence of statistical significance cannot be interpreted as evidence that the methods are equivalent.

The separate unadjusted recalculation gives a different estimate and interval, and the registry does not explain the divergence. This means the posted information does not support a single, fully reconciled numerical account of the between-arm comparison. The sponsor result and recalculation should be identified as different analyses, with the sponsor’s result recognized as the registry’s reported acceptance-rate analysis.

The outcome is acceptance on the day of enrollment. It does not tell whether participants had HIV, whether either test was more accurate, whether results were received or understood, or whether participants completed follow-up or referral. The record does not report those outcomes. The findings are specific to the trial population and emergency department context described in the registry, and do not by themselves establish how acceptance would compare in other settings.

These statistical results are not medical advice and do not determine which testing method should be used. They characterize the posted acceptance data and the uncertainty around the reported comparison. Further interpretation is limited by the lack of detail on missing outcomes and by the unresolved difference between the sponsor-reported analysis and the recalculation.

17. Why This Trial Matters Statistically

This trial illustrates why a clinical endpoint needs to be read in relation to the decision or behavior it measures. The primary outcome is acceptance of a test, not test performance or downstream health outcomes. A comparison of acceptance can inform questions about the acceptability of testing approaches in the studied setting, but it cannot answer questions about diagnostic accuracy or subsequent care.

It also illustrates the difference between a descriptive comparison and a formal between-group analysis. Arm-level proportions of 0.67 and 0.69 give a direct account of acceptance in each group. The sponsor’s confidence interval and p-value address uncertainty in the reported difference. Neither the arm values nor the p-value alone can establish that the methods are the same, and the registry does not report an equivalence margin that would support such a conclusion.

A further statistical lesson is that estimates are meaningful only when their underlying outcomes and analysis populations are clear. The sponsor-reported acceptance-rate difference and the separate unadjusted recalculation are not numerically aligned, and the registry does not reconcile them. Transparent reporting of outcome coding, denominators, missing-data handling, and analysis methods is needed to understand why analyses diverge and to judge which inference is most relevant.

Finally, planning calculations based on an observed effect are not a retrospective measure of whether a study had sufficient power. They are illustrations under stated assumptions, and smaller target effects require more participants. Keeping those planning illustrations separate from observed results helps prevent sample-size calculations from being misread as evidence that a treatment effect exists or does not exist.

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.
• Participant counts were reconstructed from posted percentages and arm sizes, so small rounding differences are possible.

The registry record leaves important details of the trial methods and analysis unreported. It does not describe eligibility criteria, recruitment procedures, the allocation ratio, sequence generation, or allocation concealment. The no-masking design is listed, but the record does not explain how that might have affected offers of testing, participant decisions, or outcome assessment. Further details about the testing process and the operational definition of acceptance are also absent.

The primary outcome is described as an intent-to-treat proportion assessed on the day of enrollment, but missing-outcome handling and participant flow are not reported. The record does not explain the relationship between the arm sizes for the primary result and the different at-risk counts in the adverse-event data. The safety entry also does not provide event definitions or additional context for the reported events.

The sponsor’s reported analysis and the separate unadjusted recalculation yield different estimates and confidence intervals. The registry does not explain the difference, including whether it reflects different outcome coding, denominators, or analysis procedures. This unresolved issue limits confidence in treating the recalculation as a direct confirmation of the sponsor result. The statistical analysis plan, original sample-size justification, and any interim analyses are not reported.

The record identifies a second-phase focus on acceptability but gives limited context about the earlier phase and how the phases relate. Its results address acceptance in the specified emergency department setting and at enrollment; generalization beyond that context is not established by the registry information. No reported secondary outcomes provide a basis to assess other aspects of testing or follow-up.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: Two Approaches to Routine HIV Testing in a Hospital Emergency Department — NCT01258582.
• Registry record last updated 2012-06-13; status last verified 2012-05.
• Results first posted 2012-06-05.
• 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 the completed USHER trial, Optimizing Strategies for Universal HIV Testing: Phase II, sponsored by Massachusetts General Hospital. It is a randomized, parallel-group trial with no masking, comparing fingerstick with oral HIV testing in a hospital emergency department. Enrollment is listed as 1651, with arm sizes of 830 for fingerstick and 821 for oral testing. The primary outcome is the proportion accepting testing on the day of enrollment, analyzed in an intent-to-treat population.

Acceptance is reported as 0.67 for fingerstick and 0.69 for oral testing. The sponsor-reported difference is -0.022, with a 95% confidence interval from -0.067 to 0.023 and p = 0.34. A separate unadjusted recalculation gives a different estimate and interval, and the registry does not explain the difference. Details about missing data, original sample-size assumptions, and multiplicity are not reported in the registry record.