Physician Coaching to Reduce Opioid-related Harms: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Physician Coaching to Reduce Opioid-related Harms (NCT02433496), a registered phase not stated study. The primary outcome is Overall Rate of Opioid Prescribing.

2. Clinical Question

The clinical question is about a prescribing outcome, not a direct measure of pain relief or other patient health outcomes. The intervention is aimed at clinical practice: the project brief describes a pilot test of guideline translation and physician coaching to promote adherence to evidence-based opioid-prescribing guidelines in primary care. The registry also lists secondary outcomes related to prescribing practices, screening, pain-management agreements, provider dropout, and characteristics of patients and clinics.

The registry record does not report a numerical comparison for the primary endpoint. It therefore does not establish whether the observed proportion differed between groups, how large any difference was, or how precisely it was estimated. The record also does not state how clinics were allocated, so the comparison should not be interpreted as a randomized treatment contrast.

3. Trial Design

The study, titled “Physician Coaching to Reduce Opioid-related Harms,” is registered as completed. Its recorded start is 2014-07 and its primary completion date is 2017-06-30. The sponsor is the University of Wisconsin, Madison. The registry lists enrollment of 53132; that figure is the recorded enrollment, not a statement of how many participants contributed to the primary endpoint analysis.

The intervention is physician coaching, described as an organizational intervention delivered to primary care clinics. The arm description says that the physician-coaching group includes 4 intervention clinics within the University of Wisconsin's Department of Family Medicine, with site visits and phone/email communication. Some of the intervention description is incomplete in the registry record. The control group is described as 4 primary care clinics that would receive no intervention. A de-identified dataset was to be created to examine differences in outcome variables between the intervention and control clinics.

The registry explicitly characterizes this as not a randomized comparison; allocation is not stated. The arms are therefore not a like-for-like control comparison according to the record. This matters because clinic characteristics, patient mix, or other differences could affect outcomes independently of coaching. The record does not report whether such differences were measured or adjusted for in an analysis.

The primary endpoint is the overall rate of opioid prescribing, with a timeframe of up to 3 years. It is defined as the proportion of patients with a chronic pain diagnosis receiving daily opioids. The record lists 14 outcome measures and names secondary outcomes including opioid/benzodiazepine co-prescribing, urine drug screening, mental health screening, pain-management agreements, high-dose patients, provider dropout, participating patient demographics, and participating clinic characteristics. It lists 0 analyses. Statistical methods, endpoint results, and analysis-population 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.

Before assignment: Clinics were paired and randomly assigned to be grouped in the intervention or control group. The intervention group was invited. If clinics in this group refused to participate, the other clinic of the pair was invited to be a part of the intervention group. Once clinics agreed to take part in the study they were immediately assigned to a group.

Overall Study

Analysis population for the primary outcome, as defined in the registry: This is the number of patients in each group.

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: Participants include patients and providers. Patient counts for Physician coaching, Control Group, and Refused group were 29956, 13296, and 9813, respectively. Provider counts for Physician coaching, Control Group, and Refused group were 34, 13, and 20, respectively.

6. Primary Endpoint

The registry lists the primary outcome as Overall Rate of Opioid Prescribing, measured over Up to 3 years.

The primary endpoint is titled “Overall Rate of Opioid Prescribing.” It is defined as the proportion of patients with a chronic pain diagnosis who receive daily opioids, with a timeframe of up to 3 years. A proportion is intended to describe how common an outcome is within a defined group. In this case, it focuses on daily opioid prescribing among patients with chronic pain rather than on an individual patient's change over time.

This kind of endpoint can be used to assess whether prescribing practice differs between clinics exposed to a practice-change intervention and clinics without that intervention. It is a measure of clinical practice, consistent with the study brief's focus on promoting adherence to prescribing guidelines. It does not by itself measure whether patients experienced changes in pain, function, adverse effects, or other health outcomes. Those outcomes are not identified as the primary endpoint in the registry record.

The registry also describes the primary outcome population as “the number of patients in each group.” This wording does not fully clarify how the proportion is constructed, including the denominator or how chronic pain diagnoses and daily opioid use are identified. Those operational details are not reported in the registry record. Nor does the registry report endpoint counts, group-specific proportions, an estimated contrast, or uncertainty around that contrast. As a result, the endpoint's intended meaning is identifiable, but its observed result cannot be interpreted from the registry record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What type of outcome is the primary endpoint?

The registered primary endpoint is a proportion: the proportion of patients with a chronic pain diagnosis who receive daily opioids. As a general statistical explanation, a proportion represents the count of people meeting a defined condition relative to a specified eligible group. Interpreting it requires clear definitions of both the numerator and denominator. Here the endpoint description identifies the clinical condition and outcome, but the record's population wording refers to the number of patients in each group and does not fully specify the denominator. The operational definition is not reported in the registry record.

When a proportion is compared between groups, the analysis typically reports the group-specific estimates and a contrast, such as a difference in proportions or another prespecified measure. The choice of contrast affects its interpretation. The registry record reports no estimate or contrast for this endpoint, so no particular effect measure can be attributed to the study.

What does precision mean?

As a general statistical explanation, a confidence interval describes a range of values compatible with the data under the assumptions of the analysis. A narrow interval indicates greater statistical precision than a wide interval, while an interval that includes values representing little or no difference may leave substantial uncertainty about the direction or size of an effect. A confidence interval is not a guarantee that the true effect lies within the interval. No confidence interval for the primary endpoint is reported in the registry record, so its precision cannot be assessed.

Why does the clinic structure matter?

The intervention is described as being delivered to clinics, with patients belonging to those clinic settings. As a general statistical principle, outcomes from people treated in the same clinic may be more alike than outcomes from people in different clinics. This is called clustering. If clustering is present, an analysis that treats every patient as wholly independent can misstate uncertainty. Methods that account for clustering may be needed, depending on the design and data structure. The registry record does not report the analysis method, whether clustering was considered, or how many clinics contributed outcome data. It is therefore not possible to determine whether the reported uncertainty, if any, reflects the clinic structure.

The record also says this was not a randomized comparison and that allocation was not stated. As a general principle, random allocation can help balance measured and unmeasured characteristics between groups. In a nonrandomized comparison, differences between groups may be associated with the outcome even without an intervention effect. Adjustment can address some measured differences but cannot automatically remove bias from unmeasured factors. The registry record does not report whether adjustment was performed or which variables were considered.

What would a test of statistical significance tell us?

As a general explanation, a significance test evaluates how compatible observed data are with a specified null hypothesis under the assumptions of the test. A p-value does not measure the probability that the null hypothesis is true, the size of an effect, or its practical importance. Interpreting a test requires the estimate, analysis plan, and assumptions, as well as attention to the study design. The registry record lists 0 analyses and reports no p-value for the primary outcome. Statistical significance cannot be assessed from the information recorded.

The study record includes a sample-size-planning description that says its planning illustration is computed from the observed effect, uses a two-sided significance level of 0.05, equal allocation, and a normal approximation, with the effect to detect stated as a fraction of the observed effect. These are assumptions for a planning illustration, not evidence that the study used those assumptions in its design or analysis. The record does not report the numerical planning results, and no sample-size illustration should be treated as observed power. A planning calculation cannot replace an estimate and uncertainty interval from the actual endpoint analysis.

For outcomes such as proportions, the analysis also depends on how observations are classified and on how missing data are handled. As a general matter, missing outcomes can alter both the numerator and denominator or the population represented by the estimate. The registry record does not report a missing-data method. It also does not specify the model, covariate adjustment, inferential procedure, or whether analysis choices were prespecified. These details are needed to evaluate the assumptions behind any numerical result, and they are not reported in the registry record.

10. Exact Estimates for Each Arm

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

R code and output

arm    <- c("Physician Coaching", "Control Group")
n      <- c(29956, 13296)
events <- c(4, 2)
for (i in seq_along(n)) {
  ci <- binom.test(events[i], n[i])$conf.int          # Clopper-Pearson exact interval
  cr <- qbeta(c(0.025, 0.975), 1 + events[i], 1 + n[i] - events[i])   # Bayesian, uniform Beta(1,1) prior
  wi <- (function(x, n) { p <- x / n; z <- qnorm(0.975); d <- 1 + z^2 / n
    c((p + z^2 / (2 * n) - z * sqrt(p * (1 - p) / n + z^2 / (4 * n^2))) / d, (p + z^2 / (2 * n) + z * sqrt(p * (1 - p) / n + z^2 / (4 * n^2))) / d) })(events[i], n[i])
  cat(sprintf("ARM|%s|%d|%d|%.1f|%.1f|%.1f|%.1f|%.1f|%.1f|%.1f\n", arm[i], n[i], events[i], 100 * events[i] / n[i],
              100 * ci[1], 100 * ci[2], 100 * wi[1], 100 * wi[2], 100 * cr[1], 100 * cr[2]))
}

ARM  Physician Coaching  29956  4  0.0  0.0  0.0  0.0  0.0  0.0  0.0
ARM  Control Group  13296  2  0.0  0.0  0.1  0.0  0.1  0.0  0.1

11. Multiplicity, Missing Data and Analysis Populations

The registry lists 14 outcome measures and names secondary outcomes concerning co-prescribing, screening, pain-management agreements, high-dose patients, provider dropout, and patient and clinic characteristics. The record lists 0 analyses, and it does not provide results for the primary endpoint or the secondary outcomes. It also does not report which outcomes were tested, whether a hierarchy of outcomes was used, or whether any adjustment for multiple comparisons was planned.

As a general statistical principle, examining multiple outcomes or multiple comparisons can increase the chance of finding an apparently unusual result even when there is no underlying effect. Methods to manage this issue can include prespecifying a primary outcome, defining an order for testing, or adjusting the interpretation of tests. These approaches answer different questions and depend on the analysis plan. No such plan is reported in the registry record, so multiplicity cannot be evaluated for this study from the registered analysis information.

The record does not report how missing endpoint data were defined or handled, whether patients were excluded from analysis, or how the primary analysis population was determined. It also does not state whether all eligible patients, only patients with complete data, or another group was analyzed. These omissions matter because an estimate based on a subset may not represent the intended population if inclusion depends on outcome or patient characteristics.

The registry's enrollment figure does not resolve these questions. Enrollment is not interchangeable with an endpoint denominator or an analysis population. The denominator, missing-data approach, and analysis-population rules are not reported in the registry record.

12. Safety Results

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

Deaths

Other (non-serious) adverse events

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

The registry lists serious adverse event figures for the physician-coaching and control groups. For the physician-coaching group, it records 34 at risk, 0 serious adverse events, 0 other adverse events, and 0 deaths. For the control group, it records 13 at risk, 0 serious adverse events, 0 other adverse events, and 0 deaths. No leading serious-event terms are listed.

These entries describe the counts recorded in the safety information; they do not establish that no adverse events occurred among all people enrolled or that the groups had identical safety experience. The registry record does not explain the definition of “at risk,” the safety observation period, or how these figures relate to the enrollment and endpoint populations. It also does not report event ascertainment procedures or detail the scope of safety collection. Those details are necessary to interpret the counts as rates or to compare safety between groups. No incidence rates should be inferred from the entries alone.

The zero counts are not a test of safety equivalence and do not show that an intervention has no risk. They should be read narrowly as the registry's recorded counts, with the limitations in population and observation details kept in view.

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

The registry record makes clear what the study set out to examine: whether physician coaching in primary care clinics was associated with the overall rate of daily opioid prescribing among patients with chronic pain. It describes coaching as an organizational intervention and identifies a control group of clinics not receiving an intervention. It also gives a broad endpoint timeframe of up to 3 years.

Those design and endpoint descriptions do not establish what happened to prescribing. The record does not report group-specific endpoint results, an estimated comparison, or uncertainty around a comparison. As a result, it cannot support a conclusion about whether prescribing rates differed, the direction of any difference, or how large or precise it might have been. Completion status is not a substitute for an outcome analysis.

The study is also not described as randomized, and allocation is not stated. This limits causal interpretation because the groups may differ for reasons other than coaching. The record does not say whether those differences were assessed or addressed statistically. The intervention's intended aim—to promote adherence to opioid-prescribing guidelines—is not itself evidence that prescribing changed.

The endpoint concerns prescribing practice, not patient health benefit or harm. Even a well-estimated change in prescribing would not, by itself, establish a change in pain, function, or adverse outcomes. The registry's safety entries list zero adverse event and death counts in both groups, but do not provide enough information about the safety population or observation process to interpret those entries as a complete assessment of risk.

Overall, the registry record supports a description of the research question and broad design, but not an empirical conclusion about the primary endpoint. Results, analysis details, and limitations of the observed comparison are not reported in the registry record.

14. Why This Trial Matters Statistically

This study raises a statistically important challenge: the intervention is delivered at the clinic level, while the primary endpoint is defined for patients. When an intervention and outcome operate at different levels, analysis needs to reflect how patients are grouped within clinics. As a general principle, failing to account for clinic-level clustering can make estimates appear more precise than the data justify. The registry record does not report whether or how this structure was handled.

The comparison also illustrates why design information matters alongside outcome measures. The record says the study was not a randomized comparison and that allocation was not stated. In that setting, a difference between intervention and control clinics could reflect characteristics of clinics or patients as well as coaching. Statistical adjustment may address measured differences, but the registry record does not describe an adjustment strategy or the factors considered.

There is also a useful distinction between a registered question and a reported result. The endpoint is defined in broad terms, the study is marked completed, and enrollment is recorded. Yet the record lists 0 analyses and supplies no endpoint estimate or uncertainty interval. Without those elements, readers cannot judge the size, precision, or statistical compatibility of any observed difference.

Finally, the study's focus on prescribing practice should be kept separate from claims about patient outcomes. Measuring a practice pattern can be valuable for evaluating guideline implementation, but it does not directly establish effects on health. Clear outcome definitions, transparent denominators, an analysis plan suited to clinic-level delivery, and reported estimates with uncertainty are central to interpreting such research. The registry record leaves these analysis details and results unreported.

15. Limitations

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

The registry's intervention description is incomplete: it identifies site visits and phone/email communication but does not provide the full description of the coaching program. The available record therefore does not establish the intervention's full content, intensity, duration, or consistency across clinics. These details are relevant to interpreting how the intervention might relate to prescribing outcomes, but they are not reported in the registry record.

The comparison is explicitly not randomized, and allocation is not stated. The record describes intervention and control clinics but does not report how those clinics came to be in their respective groups. Potential baseline differences among clinics and their patient populations may therefore affect the comparison. Whether baseline characteristics were assessed, balanced, or included in an adjusted analysis is not reported in the registry record.

The primary endpoint is defined as a proportion, while the population description refers to the number of patients in each group. The record does not clarify the denominator, the operational definitions of chronic pain or daily opioid receipt, or how the up-to-3-year timeframe was applied. It also does not report whether the outcome was assessed at the patient or clinic level in the analysis, how repeated observations were handled, or whether clustering was incorporated.

Although enrollment is listed as 53132, the record does not provide an endpoint analysis population, missing-data counts, exclusions, or group-specific results. It lists 0 analyses. The absence of an analysis and results prevents assessment of effect size, precision, and robustness. Safety figures are also limited by the absence of definitions for the at-risk population and observation period. These record-level gaps make the study's outcome and safety findings difficult to evaluate.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Physician Coaching to Reduce Opioid-related Harms — NCT02433496.
• Registry record last updated 2018-07-06; status last verified 2018-07.
• Results first posted 2018-07-06.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

19. Record Summary

NCT02433496, “Physician Coaching to Reduce Opioid-related Harms,” is recorded as completed and sponsored by the University of Wisconsin, Madison. The study start is listed as 2014-07, the primary completion date as 2017-06-30, and enrollment as 53132.

The registry describes physician coaching for intervention primary care clinics and a control group of clinics receiving no intervention. The comparison is not randomized, and allocation is not stated. The primary endpoint is the proportion of patients with chronic pain receiving daily opioids over up to 3 years. The record lists 14 outcome measures and 0 analyses. It does not report primary endpoint results, statistical estimates, confidence intervals, or analysis-population details. Consequently, the registered question and design can be described, but the endpoint's observed result cannot be assessed from the registry record.