Paired Promotion of Colorectal Cancer and Social Determinants of Health Screening: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Paired Promotion of Colorectal Cancer and Social Determinants of Health Screening (NCT04585919), a randomized na study. The primary outcome is Conditional Odds Ratio of Patients Screened for Colorectal Cancer During Intervention Compared to Patients Screened During Usual Care.

2. Clinical Question

The question is about implementing paired outreach in community health centers, rather than comparing two separately assigned groups followed throughout the same period. The registry describes sites randomized to a sequence of implementation starts. As sites moved from usual care to the intervention, they contributed observations under both conditions and served as their own controls.

The primary endpoint concerns completion of colorectal cancer screening by any guideline-based method among patients unscreened at the beginning of the relevant step. The intervention offered FIT screening and social determinants of health screening at the same contact point. The study rationale was that addressing health-related social needs alongside cancer screening might support efficiency and patient-centeredness, but the posted primary outcome is colorectal cancer screening completion. The record also describes social determinants of health screening and process outcomes as study targets; it does not report those results in detail.

The primary comparison is described as a conditional odds ratio, but the registry does not report its numerical estimate or confidence interval. It reports a mixed-model analysis and a p-value. Consequently, the direction and magnitude of an adjusted intervention effect cannot be read from the p-value alone.

3. Trial Design

This was a randomized implementation study with a sequential, stepped-wedge model and no masking. Four community health centers were randomized to a sequence of implementation starts. Each site contributed usual-care and intervention periods, with sites serving as their own controls. The record describes an average of 2.5 control periods and 5.5 intervention periods per site. The primary outcome timeframe is described as 13 pre-implementation months and 14 post-implementation months.

The intervention was clinic-staff outreach offering colorectal cancer screening with FIT and screening for social determinants of health. The usual-care description says sites first had a period of usual care and then provided the paired intervention. This design evaluates the intervention as it was introduced across clinical settings; it is not a conventional parallel-group study in which separate sets of sites remain assigned to intervention or control throughout.

The registry records enrollment as 26. It also lists arm-level records with n values of 33380.0 for the paired screening intervention and 27230.0 for usual care. The record does not explain how those n values relate to the enrollment figure, nor does it clearly define their unit or analytical scope. They should not be treated as a reconciliation of the enrollment figure without further information.

The study is marked completed, and the phase is listed as NA. The recorded start date is 2020-10-01 and the primary completion date is 2022-01-25. Harvard School of Public Health (HSPH) is named as sponsor. The record characterizes the work as implementation science: it examines implementation of an evidence-based intervention in routine practice and organizational factors that may influence implementation. It does not report details of the randomization procedure, allocation concealment, or the statistical plan beyond the outcome and analysis descriptions.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Before assignment: Community health center staff are only considered enrolled.

Usual Care

Paired Outreach

Analysis population for the primary outcome, as defined in the registry: 4 Community health centers were randomized to a sequence of implementation start and each observed for an average of 2.5 control periods (stepped wedge steps) and 5.5 intervention periods (stepped wedge steps).

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: Demographic data were not collected from participants (i.e., community health center staff) described in Participant Flow. Patients received the paired screening intervention as part of their routine care and were not individually recruited for the study because the intervention is being offered as part of routine care. Baseline characteristics and outcomes reported are based on the patient-level data.

6. Primary Endpoint

The registry lists the primary outcome as Conditional Odds Ratio of Patients Screened for Colorectal Cancer During Intervention Compared to Patients Screened During Usual Care, measured over 13 pre implementation months, 14 post-implementation months.

The primary endpoint is colorectal cancer screening completion by any guideline-based method, documented in the electronic health record for patients who were unscreened at the beginning of a stepped-wedge step. The registry frames the main comparison as a conditional odds ratio for patients screened during intervention periods compared with patients screened during usual-care periods. Its stated timeframe is 13 pre-implementation months and 14 post-implementation months.

This endpoint measures whether screening was completed, rather than only whether outreach was attempted or a FIT kit was mailed. That distinction matters for implementation: an offer or mailing is an activity, while documented screening completion is a clinical process outcome. The record also lists secondary outcomes concerning FIT kits mailed and reminders, but those do not replace the primary completion measure.

The analysis description says the comparison adjusted for patient socio-demographics, comorbidities, number of healthcare visits, and secular trend. Adjustment is intended to account for specified differences and time patterns in the model, but the registry does not give the covariate definitions, coding, or model specification. It also does not report the conditional odds ratio or its confidence interval. The posted arm-level values are 0.068 and 0.028; the record does not establish that either value is the adjusted conditional odds ratio. They should not be interpreted as that effect estimate.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What does a conditional odds ratio describe?

In general, an odds ratio compares the odds of an outcome between conditions. A conditional odds ratio is an estimate tied to the covariates and structure of the fitted model, rather than a simple comparison of unadjusted proportions. In this record, the outcome is colorectal cancer screening completion, and the conditions are intervention and usual-care periods. The intended adjusted comparison accounts for patient socio-demographics, comorbidities, number of healthcare visits, and secular trend.

The numerical conditional odds ratio is not reported in the registry record. The listed arm values of 0.068 and 0.028 are not identified as that adjusted estimate, so they cannot be used to recover it. Without the effect estimate, readers cannot determine the adjusted magnitude from the record. A p-value alone does not fill that gap.

Why use a mixed model in a stepped-wedge study?

In general, mixed models can represent clustered or repeated observations, such as data arising within sites and over time, while estimating a comparison that accounts for the model's specified structure. A stepped-wedge design has periods under usual care and intervention at each site, with transition timing varying across sites. The registry states that four community health centers were randomized to a sequence of implementation starts and that sites served as their own controls.

The record names a mixed-model analysis and says the primary comparison adjusted for patient characteristics, healthcare visits, and secular trend. It does not report the exact model, the random or fixed effects included, the link function, the handling of site and period, or the assumptions used to estimate uncertainty. Those details are needed to independently assess the model or reproduce the sponsor's analysis.

Why account for secular trend?

In general, secular trend means change over calendar time that could occur independently of the intervention. Screening practices, clinic workflows, or other conditions might shift during a study. In a stepped-wedge design, implementation status changes over time, so separating intervention-associated changes from time-associated changes is an important analytic task. The registry says the primary analysis adjusted for secular trend, but it does not specify how that trend was modeled or whether other time-varying factors were considered.

What does the reported p-value say?

In general, a p-value describes how incompatible the observed data are with a specified null hypothesis under the analysis model. The registry reports p = 0.005 for the mixed-model superiority analysis. This is a test result, not an effect size and not a confidence interval. It does not say how large the adjusted difference is, how precise the estimated effect is, or what probability the intervention is beneficial. Because the adjusted conditional odds ratio and its interval are not reported, the p-value cannot be used to assess the range of effect sizes consistent with the data.

A confidence interval, in general, expresses the uncertainty of an estimate under the statistical procedure used. It helps readers judge both precision and the range of values compatible with the analysis. Neither the interval nor its confidence level is reported for the primary comparison in this record. The exact statistical threshold or testing procedure beyond the reported p-value is also not described.

Why can over-dispersion matter for count outcomes?

In general, count outcomes may vary more than a simple statistical model expects. This is called over-dispersion; if ignored, uncertainty can be misestimated. The registry lists secondary outcomes involving numbers of FIT kits and reminders, which are count-like measures. However, the primary endpoint is described as a conditional odds ratio for screening completion, and the record does not give the distributional model used for the secondary outcomes. Over-dispersion is therefore a general consideration, not a reported feature or documented problem in this trial.

What can be recalculated from the posted summaries?

A recalculation from summary statistics is useful only when the reported values and their units support the particular comparison. In this case, the record indicates that values are missing or not estimable for a two-arm comparison for the count-other endpoint. It does not provide a recalculated effect or interval for the primary adjusted conditional odds ratio. The arm-level values and n entries do not resolve those missing components, and no new effect estimate should be inferred from them.

The record describes a planning illustration based on a fraction of the observed effect, with a two-sided significance level of 0.05, equal allocation, and a normal approximation. Such calculations, when available, are planning illustrations rather than observed power. No specific planning result is reported in the record. The statistical plan, sample-size justification, interim analyses, missing-data methods, and analysis-population definitions are not reported in the registry record.

10. Analyses Reported by the Sponsor

The sponsor's reported analysis is a mixed-model superiority comparison, with p = 0.005. The registry does not report the corresponding conditional odds ratio or a confidence interval, so the effect magnitude and precision cannot be compared with an independently recalculated adjusted estimate.

The record indicates that values are missing or not estimable for a two-arm recalculation of the count-other endpoint. The arm values of 0.068 and 0.028 do not supply the missing adjusted conditional odds ratio, and the record does not identify them as the model-based effect estimate. They should not be used to create a recalculated result that the posted information does not support.

There is no stated discrepancy between a recalculation and the sponsor's analysis. More broadly, a calculation from arm-level summaries would not reproduce a mixed-model result that adjusts for patient characteristics, healthcare visits, and secular trend. The sponsor's p-value therefore remains the reported inferential result; its interpretation is limited by the missing estimate, interval, and model details.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists five outcome measures and names three secondary outcomes: number of FIT kits mailed, number of reminders per FIT mailing, and number of reminders per FIT return. The names of any other outcome measures beyond the primary endpoint and those listed secondary outcomes are not reported in the registry record. The number of analyses is listed as two, but the record does not explain which analyses these were or how they relate to the outcomes.

Multiplicity concerns arise when several outcomes or comparisons are tested. In general, conducting multiple tests can increase the chance of observing at least one apparently unusual result by chance. Whether and how that issue was addressed in this study is not reported in the registry record. The record does not state whether the primary outcome was prespecified in a detailed analysis plan, whether secondary tests were adjusted, or whether outcomes were ranked hierarchically. The p-value should therefore be understood as the reported result for the identified mixed-model comparison, without assuming a multiplicity procedure.

Missing-data handling is not reported in the registry record. The primary outcome description refers to screening documented in the electronic health record, but it does not explain how absent or incomplete records were classified, whether records could be linked across settings, or how missing covariates were handled. The analysis population and any exclusions are also not described. These details matter because the endpoint depends on documentation and the fitted model includes patient-level covariates.

The record does not report an interim-analysis plan, stopping rules, or a detailed statistical analysis plan. It also does not explain the relationship between the enrollment entry of 26 and the arm-level n values of 33380.0 and 27230.0. These are gaps in reporting; their absence does not establish that the corresponding procedures were not used. The registry lists zero adverse events in both arms, but does not describe event collection or ascertainment methods.

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

The posted results report a mixed-model comparison of colorectal cancer screening during paired-intervention and usual-care periods, with p = 0.005. The analysis description says the comparison adjusted for patient socio-demographics, comorbidities, healthcare visits, and secular trend. The result is evidence against the model's null comparison, conditional on the analysis assumptions, but the record does not show the size of the adjusted effect or its precision.

The result does not establish how much screening changed, because the conditional odds ratio and confidence interval are not reported. The listed arm values, 0.068 and 0.028, are not identified as the adjusted model estimate. The record also does not provide a full account of the secondary outcomes or their results. It is therefore not possible from the posted primary result alone to characterize the overall implementation experience or compare all clinical and process outcomes.

The findings concern implementation of paired outreach in four community health centers using a stepped-wedge design. They do not by themselves establish what would happen in other clinical settings or under different implementation conditions. The registry does not report site-specific estimates, subgroup analyses, or enough model detail to evaluate variation across centers.

The record's completion status and its report of no adverse events do not establish the absence of harms. The event collection methods are not described. Nor does a small p-value establish that the intervention is important to patients, feasible in other settings, or effective for every outcome considered. Those conclusions require evidence beyond the reported test result.

13. Why This Trial Matters Statistically

This study raises a methodological question common in implementation research: how to evaluate a change introduced across sites over time when each site contributes both usual-care and intervention observations. The stepped-wedge structure makes sites their own controls, while randomized implementation sequence and adjustment for secular trend are intended to support the intervention comparison. The model must account appropriately for the clustering and timing in the data; the registry identifies a mixed model but does not give enough detail to inspect its structure.

Pairing colorectal cancer screening outreach with social determinants of health screening also creates multiple kinds of outcomes. Screening completion is a clinical process outcome, whereas FIT mailing and reminders describe implementation activity. Distinguishing these measures helps clarify whether an intervention was offered, whether patients completed screening, and what operational processes accompanied delivery. The record lists these outcome areas but does not report all their results.

For statistical interpretation, this record illustrates why a p-value is not a substitute for an effect estimate and interval. The reported p = 0.005 conveys a test result, but without the conditional odds ratio and its confidence interval, readers cannot assess the adjusted magnitude or precision. The unclear relationship between the enrollment entry and arm-level n values is another reason to interpret each registry field within its stated scope rather than assume that all quantities refer to the same unit of analysis.

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 primary limitation for interpreting the result is incomplete reporting of the central adjusted estimate. The registry identifies a conditional odds ratio as the primary comparison and reports p = 0.005 from a mixed-model analysis, but does not report the odds ratio, confidence interval, or detailed model specification. This prevents assessment of the size and precision of the adjusted result from the registry record.

The study includes four community health centers and uses a sequence-based stepped-wedge implementation. The registry describes sites serving as their own controls, but does not provide site-specific results or enough detail about period and site effects to assess how much results varied across centers. It also does not describe the randomization procedure or allocation concealment.

The enrollment figure of 26 and arm-level n values of 33380.0 and 27230.0 are not reconciled in the record. The unit and analytical scope of the arm-level n values are not made clear, so they should not be interpreted as directly comparable participant totals. The posted arm values are also not identified as the adjusted conditional odds ratio.

Details on missing-data handling, analysis populations, covariate definitions, multiplicity, and the full statistical plan are not reported in the registry record. Although the primary outcome relies on electronic health record documentation, the record does not explain how missing or incomplete documentation was handled. The registry names three secondary outcomes but does not provide their results in the registry record described. These reporting gaps limit assessment of the full set of clinical and implementation findings.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: Paired Promotion of Colorectal Cancer and Social Determinants of Health Screening — NCT04585919.
• Registry record last updated 2023-08-01; status last verified 2023-07.
• Results first posted 2023-08-01.
• 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

NCT04585919 is a completed randomized implementation study of paired outreach for colorectal cancer and social determinants of health screening. Four community health centers were randomized to an implementation sequence in a stepped-wedge design; sites contributed usual-care and intervention periods and served as their own controls.

The primary endpoint was colorectal cancer screening completion by any guideline-based method, documented in the electronic health record for patients unscreened at the beginning of a step. The registry reports a mixed-model analysis with p = 0.005 and describes adjustment for patient socio-demographics, comorbidities, healthcare visits, and secular trend. The conditional odds ratio and confidence interval are not reported, so the magnitude and precision of the adjusted comparison remain unclear.

The record lists secondary outcomes concerning FIT kits and reminders. It does not report full results or detailed methods for those outcomes, nor does it provide the statistical plan, missing-data approach, or analysis-population definitions.