Evaluating the Impact of Patient Photographs for Preventing Wrong-Patient Errors: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Evaluating the Impact of Patient Photographs for Preventing Wrong-Patient Errors (NCT03626766), a randomized na study. The primary outcome is Number of Order Sessions With at Least One Retract-and-Reorder (RAR) Event as Identified by the Wrong-Patient Retract-and-Reorder (RAR) Measure..

2. Clinical Question

3. Trial Design

The registry describes a randomized, parallel-group trial with no masking. It lists Columbia University as the sponsor, a completed status, enrollment of 10426, a start date of 2021-11-16, and a primary completion date of 2023-10-01. The study concerns medical errors and electronic medical records.

The registered description calls the study a multi-site, four-arm trial across academic medical centers using two electronic health record systems. Because the systems support photographs differently, the record describes separate designs. In Allscripts, providers were randomized to see a verification alert with or without a patient photograph. In Epic, providers were randomized in a 2x2 factorial design to no photograph, a banner photograph, a verification-alert photograph, or photographs in both locations.

The posted primary outcome lists four arms: Photo in Verification Alert, Photo in Banner, Photo in Banner and Verification Alert, and No Photo. The recorded arm sizes are 2171, 2202, 2196, and 2199, respectively. These are arm-level counts recorded with the outcome; the registry does not describe in this entry how they relate to provider participation, order-session exposure, or the two electronic health record systems.

The registry lists a primary outcome timeframe of 2 years. It does not report the randomization procedure in more detail, allocation concealment, site-specific enrollment, or whether the analysis accounted for clustering by provider or site. Those details 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 Order Sessions With at Least One Retract-and-Reorder (RAR) Event as Identified by the Wrong-Patient Retract-and-Reorder (RAR) Measure., measured over 2 years.

The primary endpoint is the number of order sessions with at least one event detected by the Wrong-Patient Retract-and-Reorder measure. An order session is defined as a series of orders placed by one clinician for one patient that includes at least one wrong-patient order. The measure identifies an order placed for a patient, retracted within 10 minutes, and then placed by the same provider for a different patient within the next 10 minutes.

The registry describes this as an automated, validated, and reliable measure endorsed by the National Quality Forum as NQF #2723. It is intended to capture near-miss events that the provider catches before an order reaches the patient and causes harm. This makes it a process measure of a specific type of wrong-patient ordering, rather than a count of all wrong-patient orders or of patient injuries.

The posted values are counts of qualifying order sessions for each arm. They are not accompanied by the total number of order sessions at risk in each arm, an exposure-adjusted rate, or an effect estimate. Therefore, the counts alone do not establish the relative frequency of events across arms. The registry does not report further endpoint details, such as handling of repeated events within a session.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

Randomization assigns participants to study groups by chance, which is intended to reduce systematic differences between groups at baseline. In this trial, the registry identifies the allocation as randomized, the model as parallel, and masking as none. A parallel design means the arms are compared as separate groups. The record does not describe the random sequence generation or concealment procedures, so those implementation details are not reported in the registry record.

What does the outcome count?

The primary outcome is a count of order sessions meeting a specific event definition. The recorded event is identified through the Wrong-Patient Retract-and-Reorder measure: an order is placed for one patient, retracted within 10 minutes, and then placed by the same provider for a different patient within the next 10 minutes. The outcome counts sessions with at least one such event. It does not count every individual order or every type of wrong-patient error.

Counts are often analyzed using statistical models designed for count outcomes. A model may compare event counts while accounting for the amount of time or activity at risk, such as the number of order sessions. The choice of model depends on how the events arise and on the data structure. The registry does not report an analysis model, an exposure denominator, or a model-based estimate for this endpoint.

Why does the denominator matter?

A count describes how many events were recorded, but it does not describe how often events occurred unless the amount of opportunity for events is also known. If one arm contributes more order sessions, its event count could be larger even if the event frequency per session were similar or lower. This is a general statistical point, not a claim about the unreported exposure in this trial. The registry lists arm sizes but does not identify them as numbers of order sessions; it does not post arm-specific exposure totals or rates.

For clustered data, observations from the same provider or site may be related rather than independent. In general, such dependence can affect uncertainty estimates, and methods may need to account for clustering. The trial involves providers at multiple academic medical centers, but the analysis does not state whether clustering was addressed. That method is not reported in the registry record.

What can be learned from the arm counts?

The posted counts are 5587 in Photo in Verification Alert, 5836 in Photo in Banner, 5598 in Photo in Banner and Verification Alert, and 6279 in No Photo. Descriptive counts can show the recorded number of outcome events in each arm. They cannot establish the size or precision of a group effect without a defined comparison and appropriate denominators. The registry posts no confidence intervals, p-values, adjusted estimates, or analysis results.

A confidence interval, in general, gives a range of values reasonably compatible with the data and the assumptions of an analysis. Its width reflects precision: a wider interval indicates greater uncertainty about the effect estimate. Since no interval or estimate is posted, there is no reported precision measure for the comparisons. A lack of a posted interval does not mean that uncertainty is absent; it means the record does not quantify it.

How does the design shape the comparisons?

The trial description includes a two-arm comparison in Allscripts and a 2x2 factorial design in Epic. A factorial design can be used to study more than one intervention component, and may allow investigators to assess component effects and combinations. Interpreting those effects depends on the prespecified analysis and how interaction between components is handled. The registry does not report an analysis plan or results for such effects. Although four arms are listed for the outcome, there is no posted estimate comparing all arms or isolating the effect of each photograph location.

The record identifies the primary outcome timeframe as 2 years. A timeframe states the period associated with outcome collection, but it does not itself establish how follow-up was defined for each provider or session. Those details, along with any handling of missing or incomplete outcome data, are not reported in the registry record.

The registry’s sample-size assumptions describe a planning illustration based on the observed effect, with a two-sided significance level of 0.05, equal allocation, and a normal approximation. Such illustrations estimate the participants per arm that might be needed to detect an assumed effect under stated conditions. They are planning scenarios, not observed power and not evidence that a comparison reached a particular precision. The record does not report the trial’s own sample-size justification, and the posted outcome lacks an effect estimate from which a reader could evaluate the illustration in context.

10. Multiplicity, Missing Data and Analysis Populations

The registered study includes several photograph placements and describes distinct designs across two electronic health record systems. This creates the possibility of multiple comparisons, including comparisons of individual placements, combinations of placements, and system-specific effects. When many hypotheses are tested, the chance of finding at least one apparently unusual result by chance can increase. General approaches to multiplicity include prespecifying a primary contrast or adjusting inference across a family of tests.

The registry record does not report a multiplicity strategy, a prespecified contrast hierarchy, or an analysis plan identifying which comparisons were primary. It also posts no inferential analyses for the outcome. Therefore, the effect of any multiplicity adjustment cannot be assessed from this record.

The endpoint is based on electronic records, but the record does not state whether all randomized providers contributed outcome data, how incomplete records were handled, or whether any participants were excluded from the analysis. Missing-data methods, analysis populations, and deviations from randomized assignment are not reported in the registry record. The listed arm sizes and endpoint counts do not by themselves identify an intention-to-treat or per-protocol population.

These absences limit interpretation of precision and comparability, but they do not establish that missing data or multiplicity problems occurred. They mean the registry does not document how these issues were addressed.

11. 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 lists serious adverse events, other adverse events, and deaths for each of the four groups. For each group, it records zero affected for serious adverse events, zero affected for other adverse events, and zero deaths. The listed numbers at risk correspond to the arm-level figures recorded in the safety entry: 2171, 2202, 2196, and 2199.

These figures should be read as registry-reported counts, not as proof that no adverse events could occur or that the interventions have been established as safe. The record does not provide event descriptions because no serious-event terms are listed, and it does not explain the safety surveillance definitions or reporting procedures. It also does not report a comparative statistical analysis of safety across groups.

The study interventions involve where patient photographs appear in an electronic health record. The adverse-event entry reports no affected participants in the listed categories, but the registry does not describe whether all events were actively collected or how long safety follow-up lasted. The primary outcome timeframe is 2 years; the safety entry does not separately specify a safety follow-up period.

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

The record documents a randomized evaluation of patient photographs displayed in an electronic health record and a primary outcome intended to capture a defined near-miss wrong-patient ordering pattern. It supplies four arm-level counts of qualifying order sessions and describes how the measure identifies those events. These details help readers understand the outcome and the study’s intended comparison.

The posted counts do not establish that one placement produced fewer events than another. They are not accompanied by order-session denominators, rates, effect estimates, confidence intervals, or p-values. No analysis is posted, and the record identifies a two-arm comparison as not estimable from the available values. As a result, the comparative effect and its precision are not reported in the registry record.

The endpoint is also narrower than the general concept of wrong-patient error. It identifies a retracted order followed by an order for another patient by the same provider within the specified time windows. The registry describes these as near misses caught before they reach the patient and cause harm. The endpoint therefore does not measure all wrong-patient orders, errors that were not retracted in this pattern, or patient harm directly.

In general, a randomized design can support causal comparisons when the analysis follows the design and relevant data are available. Here, however, the registry does not post the analyses needed to assess the randomized comparison. The absence of a reported result is not evidence that the intervention had no effect, nor is it evidence that it reduced errors. It indicates that the registry record does not provide an inferential answer to the primary question.

The safety entry records zero serious adverse events, zero other adverse events, and zero deaths in every listed group. Those are the reported counts for the registry entry; they do not establish that risks are absent or resolve unreported details of safety monitoring.

13. Why This Trial Matters Statistically

This trial addresses a statistical challenge common in health care quality research: a safety intervention is evaluated using events that are uncommon in relation to the volume of routine activity, while the relevant opportunity for an event may vary across providers. The primary endpoint is not simply a count of patients with an outcome; it counts order sessions meeting a specific near-miss definition. Interpreting such counts requires clarity about the exposure denominator and the dependence of observations within providers or sites.

The study also illustrates the importance of distinguishing a registered design from a reported analysis. The record describes randomization, four photograph conditions, and different designs across Allscripts and Epic. But it does not post a model, a comparison estimate, or uncertainty measures. Without these, arm counts cannot answer whether the groups differ statistically or how precisely any difference is estimated.

Its factorial component raises another broadly relevant issue: the interpretation of individual components and combinations depends on the planned contrasts and on how interactions are handled. The registry gives no analysis plan or results for those questions. Clear reporting of the estimand, denominator, model, and inferential approach would be necessary to understand what conclusions the design supports.

Finally, the outcome is a process measure of a specific type of self-caught near miss. Statistical interpretation must remain aligned with that definition. Counts of these events cannot automatically be generalized to all wrong-patient errors or clinical harm. The record therefore offers useful information about the study design and measurement strategy, while leaving the comparative statistical question unresolved.

14. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

The principal limitation of the posted results is that they contain arm-level event counts without exposure totals or a statistical comparison. The arm sizes are listed, but the record does not identify these as order-session denominators. Consequently, event rates, relative effects, and their precision are not reported. The two-arm comparison is also marked as not estimable from the posted values.

The registry provides no analysis plan, model specification, confidence intervals, p-values, or adjusted estimates. It does not state whether provider or site clustering was addressed, despite the multi-site provider setting. Details of the randomization process, outcome-data completeness, missing-data handling, and analysis population are not reported in the registry record.

The outcome definition is specific: it captures order sessions with a qualifying retract-and-reorder event. This excludes wrong-patient errors that do not follow that event sequence and does not directly measure patient harm. The record does not report the total number of order sessions or how repeated events within a session were handled beyond defining the session outcome as including at least one event.

The trial description combines a two-arm Allscripts design and a 2x2 Epic design, while the posted outcome lists four arms. The registry does not provide system-specific outcome counts or explain how the two designs were combined for the posted results. It also does not describe the planned handling of multiple comparisons or the possible interaction between photograph locations.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: Evaluating the Impact of Patient Photographs for Preventing Wrong-Patient Errors — NCT03626766.
• Registry record last updated 2025-03-17; status last verified 2025-02.
• Results first posted 2025-03-17.
• 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.

18. Record Summary

NCT03626766 is a completed randomized, parallel-group trial sponsored by Columbia University, with enrollment of 10426. It evaluated patient photographs displayed in electronic health records to prevent wrong-patient electronic orders. The registered designs differed by system: a two-arm verification-alert comparison in Allscripts and a 2x2 factorial design in Epic.

The primary outcome was the number of order sessions with at least one event identified by the Wrong-Patient Retract-and-Reorder measure over a timeframe of 2 years. The registry lists four arms and their event counts, but no analysis results or comparative effect estimates. Exposure denominators, analysis methods, missing-data procedures, and multiplicity handling are not reported in the registry record.

The safety entry records zero serious adverse events, zero other adverse events, and zero deaths in each group. The registry does not provide a statistical conclusion about whether any photograph placement changed the primary outcome.