Optimizing Care for Patients With Dementia: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Optimizing Care for Patients With Dementia (NCT03442322), a randomized na study. The primary outcome is Dispensing of Psychotropic Medications.

2. Clinical Question

The clinical question is about comparing two ways of organizing non-pharmacological dementia care, not about comparing a care approach with no intervention. The approaches differ in how staff coordinate: one emphasizes an integrated team and shared language, while the other emphasizes individual assessments and discipline-specific expertise. The registry describes antipsychotic medication dispensing as the primary outcome. It does not establish whether a prescription was appropriate, whether a resident took the medication, or why it was dispensed.

The brief description explains the interest in approaches that seek to address circumstances associated with disruptive behaviors. It also notes that the conditions under which each facility-based method is most effective were unclear. The comparative question is therefore whether the recorded primary outcome differed between the two assigned approaches during the stated intervention period. The registry record does not report a more detailed hypothesis or a prespecified direction for the difference.

3. Trial Design

The registry identifies a randomized, double-masked, parallel-group study comparing the transdisciplinary and multidisciplinary approaches. It lists the study as completed, with a start date of 2018-03-15 and a primary completion date of 2022-07-26. Phase is listed as NA. The sponsor is the University of Pittsburgh.

The brief description says the project would prospectively randomize 80 nursing homes to the two approaches. The registry lists enrollment as 53. It does not explain the relationship between the planned number of nursing homes and the listed enrollment, or specify in the enrollment field whether that figure counts facilities, residents, or another unit. The primary outcome counts residents with valid Minimum Data Set assessments, and the arm results report resident counts. The record's allocation field says randomized, but does not provide the randomization procedure or allocation ratio.

The transdisciplinary arm is described as integrated across disciplines, with core training and a common language intended to support continuity and sustainability. The multidisciplinary arm is described as problem-based, drawing on the expertise of individual providers to address care concerns. Both approaches are non-pharmacological treatment approaches. The registry does not provide detail about intervention fidelity, the amount of training, implementation at individual facilities, or whether all assigned facilities delivered their approach as planned.

The parallel-group label means the record describes the two approaches as concurrent comparison arms rather than as a sequence of treatments for each participant. The record does not report the masking procedures, who was masked, or how masking was maintained. It also does not report the original design assumptions, the planned analysis model, or a sample-size justification. These details should not be inferred from the posted resident counts or from retrospective planning illustrations.

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: This mixed methods study used quantitative methods to examine nursing home resident outcomes of the two non-pharmacological approaches to dementia care (i.e., the primary study outcomes). Qualitative methods included one-time interviews with family caregivers and nursing home staff intended to understand their experiences with the two non-pharmacological approaches. Flow chart below reflects the participating nursing homes (i.e., units) and their residents with dementia (i.e. participants).

Overall Study

Analysis population for the primary outcome, as defined in the registry: Population includes all nursing home residents with dementia in participating nursing homes with valid data in the Minimum Data Set (MDS) assessments.

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: This was a nursing home-level analysis, therefore not all participants were in the sample during the baseline period. Participants (i.e., nursing home residents) were included if they met the study criteria based on age, dementia status, and long-term nursing home residence, but could be included in the sample at any point of the study. Thus, there were participants who were in the study sample (and reflected in the flow overview), but were not in the study sample during the baseline period.

6. Primary Endpoint

The registry lists the primary outcome as Dispensing of Psychotropic Medications, measured over 6-month period (months 1-6).

The primary outcome is dispensing of psychotropic medications. More specifically, it counts nursing home residents with dementia who received one or more antipsychotic medications over the 6-month period (Months 1-6) of intervention. The outcome population consists of residents with dementia in participating nursing homes who had valid data in Minimum Data Set assessments. The arm results present the number of residents with the outcome and the number counted in each arm.

This outcome is used to describe medication dispensing during the intervention period, an observable event that can be summarized for each arm and compared between approaches. It is not a direct measure of behavior, quality of life, medication appropriateness, or medication use beyond the stated period. The registry's wording refers to residents who received one or more medications; it does not specify dose, duration, indication, or whether a medication was newly started or continued.

The primary results list 161 of 742 residents in the transdisciplinary arm and 315 of 1,787 in the multidisciplinary arm as having the outcome. The registry also reports outcome proportions of 0.217 and 0.176, respectively, for the comparison. Those figures describe the recorded results in the defined data population. They do not by themselves establish why the observed counts differ, and they should be read alongside the sponsor's reported analysis and the limitations of the available information.

7. Results Posted in the Registry

• Multidisciplinary Approach: 17.6% (315/1787) versus 21.7% (161/742) on control; risk difference -4.1 percentage points (95% CI -7.5 to -0.6), risk ratio 0.81 (0.69–0.96); unadjusted two-proportion z-test p = 0.017.

8. Secondary Efficacy Endpoints

The registry also posts results for secondary outcomes. For each outcome below, the endpoint type, the test applied to the posted summary statistics, the result and its interpretation are given in turn. Secondary endpoints are usually not adjusted for multiple testing, so they support rather than confirm the primary finding.

Unintended Weight Loss

• Endpoint type: Binary: the number of participants with the event, compared as proportions.
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • Multidisciplinary Approach: 16.0% (302/1884) versus 16.0% (125/781) on control; risk difference 0.0 percentage points (95% CI -3.0 to 3.1), risk ratio 1.00 (0.83–1.21); unadjusted two-proportion z-test p = 0.987.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Multidisciplinary Approach and the control. The interval gives the range of effects that remain compatible with the data.

Falls

• Endpoint type: Binary: the number of participants with the event, compared as proportions.
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • Multidisciplinary Approach: 39.7% (750/1887) versus 41.5% (326/785) on control; risk difference -1.8 percentage points (95% CI -5.9 to 2.3), risk ratio 0.96 (0.87–1.06); unadjusted two-proportion z-test p = 0.392.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Multidisciplinary Approach and the control. The interval gives the range of effects that remain compatible with the data.

Depressive Symptoms

• Endpoint type: Binary: the number of participants with the event, compared as proportions.
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • Multidisciplinary Approach: 5.4% (90/1676) versus 5.4% (37/680) on control; risk difference -0.1 percentage points (95% CI -2.1 to 1.9), risk ratio 0.99 (0.68–1.43); unadjusted two-proportion z-test p = 0.945.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Multidisciplinary Approach and the control. The interval gives the range of effects that remain compatible with the data.

The posted secondary endpoints cover unintended weight loss, falls, and depressive symptoms. For the resident-level comparisons, the unadjusted recalculated confidence intervals include zero, and the reported unadjusted p-values do not indicate clear separation between arms. The corresponding sponsor analyses also report p-values that do not indicate a clear difference under those analyses. This overall pattern is less decisive than a claim of a consistent secondary-outcome effect would require.

These outcomes provide context for the primary medication-dispensing comparison, but they measure different events. They should not be combined into a single summary of benefit or harm, and they do not establish whether a difference in medication dispensing was accompanied by a change in these other outcomes. The record's secondary outcome list contains 12 posted secondary results, while the registry lists 24 outcome measures and 24 analyses overall. The relationship between all listed outcomes and the repeated outcome names is not fully explained in the record.

Unadjusted secondary findings carry less weight when many outcomes are examined. In general, testing multiple outcomes without adjustment increases the chance of finding one or more apparently unusual results by chance. The registry does not report whether a multiplicity correction, hierarchy, or other strategy was prespecified. Thus, individual unadjusted p-values should be treated as exploratory context rather than as independent confirmations of the primary result. This qualification applies regardless of whether a secondary p-value appears small or large.

The secondary results also share limitations of the primary comparison: the registry does not give complete information about the sponsor's models, missing-data procedures, or how the nursing-home assignment structure was handled. The secondary outcomes can describe the posted data, but the record does not support stronger claims about how the approaches compare across a broad range of resident outcomes.

Secondary outcomes registered for this trial, for reference (only those above were recalculated):

• Unintended Weight Loss
• Unintended Weight Loss
• Unintended Weight Loss
• Falls
• Falls
• Falls
• Depressive Symptoms
• Depressive Symptoms

9. Statistical Interpretation

10. Statistical Methods Explained

The registry reports two kinds of statistical information for the primary outcome: arm-level counts and a sponsor analysis. A separate unadjusted recalculation compares the arm-level event proportions. It is useful to distinguish these because they answer related but not necessarily identical statistical questions.

What does a risk difference express?

In general, a risk difference is the event proportion in one group minus the event proportion in the comparison group. It is expressed on the proportion scale, so zero corresponds to equal observed proportions. The primary recalculation compares multidisciplinary with transdisciplinary care and reports -0.041. The negative sign indicates that the observed proportion was lower in the first group in that subtraction. The estimate is based on the posted counts of residents with the outcome and the corresponding arm totals.

This is an unadjusted comparison. In general, an unadjusted comparison summarizes the observed group results without accounting for other variables in a model. It does not adjust for differences in resident or facility characteristics, and it does not itself handle any clustering arising because residents are grouped within nursing homes. The registry record does not report whether the sponsor's model accounted for facility clustering or which covariates it included.

What does the confidence interval add?

In general, a confidence interval describes a range of values compatible with the data and the method used to calculate the interval. It helps readers judge the precision of an estimate: a narrower interval indicates more precision than a wider interval on the same scale. The recalculated 95% confidence interval is -0.075 to -0.006. It does not include zero, but the interval still spans multiple possible magnitudes of the difference. A confidence interval is not a guarantee that the true value lies in the interval, nor is it a probability statement about this particular estimate.

How should the p-value be read?

In general, a p-value describes how unusual data at least as incompatible with a stated null hypothesis would be under the assumptions of the statistical procedure, if that null hypothesis were true. It is not the probability that the null hypothesis is true, and it does not measure clinical importance or the probability that the result will replicate. The recalculation's unadjusted p-value is 0.017. The sponsor's reported p-value is 0.492. Those values come from different analyses, and the registry does not provide enough detail to reconcile them.

The registry names the sponsor method as linear regression with a difference-in-difference estimator. In general, difference-in-differences methods compare changes across groups over time, using the relative change between groups as the estimand. The record does not specify the time points, regression terms, covariates, standard-error method, or the scale of the estimate. It reports a value of 1.25 and p = 0.492 without a confidence interval. The model's precise interpretation therefore cannot be reconstructed from this entry alone.

Why can the analyses differ?

In general, an adjusted regression estimate can differ from a direct comparison of the observed arm proportions because the methods can target different quantities and account for data structure or covariates differently. A difference-in-difference estimator may also use information about changes over time that is not represented in a comparison of arm-level outcome counts. These are general reasons analyses can produce different estimates; the registry does not state which explains the discrepancy in this trial. Since the discrepancy flag is true, the sponsor's own analysis takes precedence in reporting the trial's result.

The posted results also include secondary endpoints and sponsor analyses. A p-value from an unadjusted recalculation for a single endpoint does not account for testing many endpoints. In general, repeated testing increases the chance that at least one result appears unusual by chance if no effects exist. Whether the sponsor adjusted for multiplicity, and what analysis populations or missing-data procedures were used, is not reported in the registry record. Readers should therefore distinguish the descriptive arm results, the unadjusted recalculations, and the sponsor's model-based analyses rather than treating them as one unified analysis.

11. Confidence Intervals and Precision

The primary unadjusted recalculation estimates a risk difference of -0.041, with a 95% confidence interval from -0.075 to -0.006. The interval provides a range of values compatible with the observed counts under the recalculation's assumptions. Because it does not include zero, the recalculation is consistent with a difference in the observed proportions. Its width also indicates that the exact magnitude is not pinned down to a single value. The interval does not remove uncertainty about confounding, facility clustering, missing observations, or the applicability of the calculation to the sponsor's analysis.

In general, confidence intervals describe statistical uncertainty under a specified method; they do not capture every source of uncertainty in a study. The registry does not report confidence intervals for the sponsor's difference-in-difference estimate. Its precision therefore cannot be judged from the sponsor result as posted. The sponsor reports an estimate of 1.25 with p = 0.492, but the estimate's scale and model details are not reported, further limiting direct interpretation.

For the secondary outcomes, recalculated intervals are also posted. Their general pattern is that the intervals include zero, indicating that the summary-statistic results do not sharply distinguish the arm proportions for those outcomes. That does not establish equivalence or show that the approaches have identical outcomes: an interval including zero can be compatible with a range of positive and negative differences. The registry does not report equivalence margins or a non-inferiority framework.

Precision should also be considered in light of the structure of the study. The brief describes nursing homes as the units to be randomized, whereas primary outcome counts are reported for residents. The registry does not state how that grouping was reflected in the sponsor's uncertainty estimates. The posted recalculated interval should be interpreted as a summary-statistic estimate, not as a substitute for a fully specified analysis that accounts for the trial design.

12. Planning a Similar Trial: Sample Size

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

The sample-size figures are planning illustrations based on the observed effect, not the trial's original design calculation. The primary and secondary outcome illustrations are expressed as participants per arm. The sample-size 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 the stated fraction of the observed effect. The registry's original design assumptions are not reported in the registry record.

The 80% power column gives the illustrative number of participants per arm for a design with an 80% chance of detecting the specified effect under those assumptions, if that effect and the other assumptions were correct. The 90% power column gives the corresponding number for a 90% chance. These are design-planning concepts, not probabilities that the observed result is true. This is not observed power.

Rows for planning for 100%, 75%, and 50% of the observed effect show how the assumed effect size changes the illustration. A smaller effect is harder to distinguish from random variation, so it requires more participants to reach the same planned power under the stated assumptions. For the primary outcome, the illustrations are for participants per arm; they are not totals across both arms or facility counts. The figures should not be read as the number enrolled in the trial or as a recommendation for a future study. They are sensitive to their assumptions, and the registry does not report whether the original trial's design used these assumptions.

The sample-size curve is another way of displaying the planning relationship between effect size and required sample size. It should be read as a planning aid associated with the stated assumptions, not as additional evidence about the trial's treatment comparison. The registry does not report sample-size illustrations as total events for these outcomes; the event-total convention applies to time-to-event calculations, while these illustrations are participant counts per arm.

13. 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 = 0.217; treatment proportion p₁ = 0.176; average p̄ = 0.197.
2. First term: z_1−α/2 × √[2p̄(1−p̄)] = 1.96 × 0.5621 = 1.1016.
3. Second term: z_1−β × √[p₀(1−p₀) + p₁(1−p₁)] = 0.842 × 0.5613 = 0.4724.
4. n = (1.1016 + 0.4724)² / (-0.0407)² = 1,495.2, rounded up to 1,496 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 <- 0.216981132075472                          # control proportion
p1 <- 0.176273083379966                          # 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(3442322)
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 = 1,496 (80% power), 2,002 (90% power)
 75% of observed effect: n per arm = 2,711 (80% power), 3,629 (90% power)
 50% of observed effect: n per arm = 6,215 (80% power), 8,319 (90% power)
power.prop.test: 1495.2 
Simulated power at n = 1,496 per arm: 0.802 (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.

14. Analyses Reported by the Sponsor

The registry reports the sponsor's primary analysis as linear regression using a difference-in-difference estimator. For the two groups, the reported estimate is 1.25 and the reported p-value is 0.492. No confidence interval is reported for this estimate, and the record does not give enough model detail to determine its scale or reconstruct the analysis.

The unadjusted recalculation from the posted primary arm counts instead reports a risk difference of -0.041, a 95% confidence interval from -0.075 to -0.006, and an unadjusted p-value of 0.017. This calculation compares the arm-level event proportions; it is not the sponsor's regression analysis. The two estimates should not be treated as estimates on the same scale or combined into a single result.

The discrepancy flag is true. Accordingly, the sponsor's own analysis takes precedence when describing the trial's reported result. That does not erase the observed counts or make the recalculation the sponsor's finding. It means the recalculation should be identified as a separate, unadjusted analysis and interpreted cautiously. The registry does not explain the difference between the analyses, so the reason for the contrasting estimates and p-values is not reported in the registry record.

15. Multiplicity, Missing Data and Analysis Populations

The registry lists 24 outcome measures and 24 analyses, and reports 12 secondary results. Multiple outcomes create a multiplicity issue: in general, the more statistical comparisons made, the greater the chance of at least one apparently notable result arising by chance. A prespecified testing hierarchy or adjustment can address that issue, but whether such a plan was used is not reported in the registry record. The record does not identify a primary family of hypotheses beyond naming the primary outcome, or state whether secondary endpoints were confirmatory or exploratory.

The sponsor's analysis plan, including the treatment of multiple comparisons, is not reported in the registry record. Nor are the handling of missing outcomes, the criteria for inclusion in each analysis, or whether any imputation was used. The primary outcome population is described as residents with dementia in participating nursing homes who had valid Minimum Data Set assessments. That definition clarifies the source population for the reported outcome, but it does not explain how many residents lacked valid assessments or how their missing data were handled.

The record does not describe the analysis population in terms such as intention-to-treat, per-protocol, or modified intention-to-treat. It also does not state whether residents were analyzed according to nursing-home assignment, whether residents moved between facilities, or how departures and incomplete follow-up were treated. These omissions make it difficult to evaluate possible selection effects or reproduce the reported analyses.

The brief description says nursing homes would be randomized, while outcomes are reported as resident counts. In general, outcomes of residents within the same facility may be correlated, so analyses may need to account for the facility as the assignment unit. The registry does not report whether the sponsor's model incorporated this structure, or whether the unadjusted recalculation's uncertainty accounts for it. These are important details for interpreting precision and the analysis population, not reasons to infer that the reported data are invalid.

16. 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 adverse-event record reports zero serious adverse events, zero other adverse events, and zero deaths in both approach groups. The listed at-risk counts are 1,291 for the transdisciplinary approach and 3,206 for the multidisciplinary approach for both serious and other adverse events. No serious adverse event terms are listed.

These entries should be read as what the registry record reports, not as proof that no adverse events occurred beyond the record's reporting scope. The at-risk counts differ from the resident counts in the primary outcome results. The registry does not explain the denominators, the event-collection period, the rules used to classify events, or the relationship between the safety population and the primary outcome population. It also does not provide event-level descriptions or a separate safety analysis.

The zero counts therefore give limited information about comparative safety. They do not establish that the approaches have identical safety profiles or that no events went unrecorded. The adverse-event entries are distinct from the primary outcome of medication dispensing, and they should not be used to infer anything about that endpoint.

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

The results describe a randomized comparison of two facility-based approaches and report resident counts for medication dispensing and selected secondary outcomes. For the primary endpoint, the arm-level counts and the unadjusted recalculation show a lower observed outcome proportion in the multidisciplinary arm. The sponsor's reported analysis, however, gives an estimate of 1.25 with p = 0.492. Because the record flags a discrepancy, the sponsor's own analysis takes precedence as the trial's reported analysis; the recalculation remains a separate descriptive comparison.

This does not establish that either approach works or fails, nor does it show that the approaches are equivalent. A non-significant p-value is not proof of no difference, and an unadjusted result with an interval excluding zero is not by itself proof of a causal effect. Interpretation depends on the analysis model, the assignment structure, missing data, and other design details that are not reported in the registry record.

The endpoint concerns recorded dispensing of one or more antipsychotic medications during a defined six-month period. It does not report why a medication was dispensed, how long it was used, or whether it changed resident behavior. The secondary outcomes concern different events and do not provide a single overall measure of care quality. The safety entries report zero events in the listed categories, but the reporting details and denominators are not fully explained.

The results apply to the recorded outcome populations and period. The registry does not provide enough information to determine how generalizable the findings are to facilities or residents beyond those represented in the data. It also does not report implementation fidelity or all features of the original statistical plan. The most defensible reading is therefore descriptive: the posted counts differ, the recalculation and sponsor analysis do not align, and the record lacks details needed to resolve that discrepancy or make broader claims.

18. Why This Trial Matters Statistically

This trial is statistically interesting because it compares two active ways of organizing non-pharmacological care rather than contrasting an intervention with no care approach. The clinical question concerns how integrated team practice compares with discipline-specific, problem-based assessment. That makes the definition and implementation of each approach important to interpreting any outcome contrast.

The registry describes nursing homes as the planned randomization units, while outcome counts are reported for residents. This distinction matters statistically. In general, residents within the same facility may share care practices and other characteristics, so resident-level observations may not be independent. Understanding how an analysis accounts for the assignment unit is important for estimating uncertainty. The sponsor's model details and handling of facility clustering are not reported in the registry record.

The record also illustrates why a simple comparison of posted arm summaries and the sponsor's model-based result should be kept distinct. The recalculation and sponsor analysis have different reported forms and contrasting p-values. A summary-statistic recalculation can make the observed arm contrast transparent, but it cannot substitute for a fully specified analysis that reflects the design. The discrepancy flag and absent model details are central interpretive details, not minor technicalities.

Finally, the number of listed outcomes makes multiplicity and prespecification relevant. Without a reported strategy for handling multiple comparisons, secondary findings should be interpreted cautiously. Clear reporting of outcome definitions, analysis populations, missing-data handling, and model assumptions would help readers evaluate both the primary estimate and its precision. These considerations are relevant to understanding the evidence, without determining what care approach should be used.

19. 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.

Important trial-specific limitations arise from incomplete statistical and design reporting. The registry says the project planned to randomize 80 nursing homes but lists enrollment as 53, without explaining the units represented by the enrollment figure or reconciling it with the plan. Resident counts appear in the outcome results, but the record does not state how many nursing homes contributed data to each outcome or how the assignment structure was represented in analyses.

The sponsor's primary analysis is described only as linear regression with a difference-in-difference estimator. The model specification, estimate scale, time points, covariates, standard-error method, and confidence interval are not reported. The unadjusted recalculation differs from that reported result, and the registry does not explain why. The discrepancy limits the interpretation of the summary-statistic result as a representation of the trial's intended analysis.

The primary endpoint is based on valid Minimum Data Set assessments, but the record does not state how missing or invalid assessments were handled, how many residents were excluded, or whether follow-up was complete. It does not describe the analysis population or whether the sponsor used an intention-to-treat approach. Intervention fidelity and variation in facility implementation are also not reported.

The registry lists many outcomes and analyses but does not report a multiplicity plan or distinguish confirmatory from exploratory secondary analyses. Safety reporting provides zero counts but leaves the event definitions, ascertainment period, and denominators unexplained. Finally, the registry does not give the original sample-size calculation or its assumptions. These gaps do not determine the direction of any bias, but they limit reproducibility and the strength of conclusions that can be drawn from the posted record.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: Optimizing Care for Patients With Dementia — NCT03442322.
• Registry record last updated 2024-08-19; status last verified 2024-03.
• Results first posted 2024-08-19.
• 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.

23. Record Summary

NCT03442322 is a completed randomized, double-masked, parallel-group comparison of transdisciplinary and multidisciplinary non-pharmacological approaches for nursing home residents with dementia. The primary outcome is dispensing of one or more antipsychotic medications during Months 1-6, among residents with valid Minimum Data Set assessments. The registry lists enrollment as 53 and reports resident-level primary outcome counts of 161 of 742 and 315 of 1,787.

The unadjusted recalculation reports a risk difference of -0.041 with a 95% confidence interval from -0.075 to -0.006 and p = 0.017. The sponsor's linear regression difference-in-difference analysis reports 1.25 and p = 0.492, without a confidence interval. The record flags a discrepancy, so the sponsor's analysis takes precedence. Secondary results and safety counts are also posted, but important details about multiplicity, missing data, analysis populations, and facility-level modeling are not reported in the registry record.