A proposed dose-aware study of incident delirium

Introduction

Earlier analyses relied on medication information without complete concurrent dose, route, and timing. A stronger test would measure actual administration before an independently assessed outcome. The Step 20B protocol proposes such a test for incident delirium in adults aged 65 years or older. ^{1, 2}

Its primary question is whether a prespecified, frozen, dose- and route-aware UnifiedACB exposure relates to a new delirium episode. A second question is whether it adds useful information beyond medication count and established anticholinergic scales. The protocol is a research plan, not an enrolled cohort or completed validation result. ¹

Methods

The preferred design is a multicenter prospective cohort; a high-quality retrospective electronic health record cohort may first assess feasibility. Eligible settings include acute medical, geriatric, and surgical inpatient care, with intensive care analyzed separately. Prevalent delirium at baseline would be excluded. ¹

Medication exposure would be identified by ingredient, formulation, dose, route, frequency, and administration time from medication administration records. The prespecified, frozen eligible model components would be measured before the first qualifying delirium assessment. Incident delirium would be assessed with Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria, the three-minute Confusion Assessment Method interview (3D-CAM), or a documented validated workflow. The analysis would account for baseline cognition, illness severity, sedatives and opioids, frailty, kidney and liver status, site, and missingness. The CAM and 3D-CAM are established assessment approaches; the exact outcome procedure must be chosen and fixed before the study. ^{1, 3, 4}

The proposed baseline exposure uses the admission or reference U calculated before the first eligible outcome assessment. A secondary time-varying analysis would sum eligible raw R contributions from a daily or prior-24-hour window and transform once. Post-outcome administrations must not be counted as pre-outcome exposure. A home medication list would be reconciled with actual administrations rather than treated as proof that every medicine was given. ¹

For each component, the protocol requires ingredient and combination-product mapping, formulary code, dose unit, route, formulation, frequency, and timestamp. Unknown exposure or noneligible model evidence would be listed separately from an eligible subtotal; unknowns would never be imputed as zero without an explicitly justified sensitivity analysis. Secondary outcomes proposed in the protocol include delirium duration, delirium-free days, falls, urinary retention or constipation, length of stay, discharge disposition, and mortality. ¹

The adjusted primary association would be estimated per U point with logistic or time-to-event methods as appropriate to the outcome timing and site structure. Planned checks include nonlinearity, missing data, age and baseline-cognition strata, route completeness, medication-group influence, and leave-one-site-out analysis. Any discrimination or calibration claim would require a defined prediction model and a common-patient comparison with prespecified clinical and legacy-scale baselines. ¹

Results

No participants have been enrolled and no delirium outcome results are available. This section records protocol status only. It does not imply that the model has been prospectively validated. ¹

The protocol-ready elements include a proposed population, exposure windows, outcome definition, covariate groups, and sensitivity analyses. They remain design specifications. Feasibility, expected precision, sample size, ethics approval, data quality, and independent adjudication still have to be established at a participating site before any analysis can begin. ¹

Discussion

Before execution, an appropriate site and governed data access, delirium ascertainment, expected precision, sample size, ethics review, and independent adjudication are required. The score version and analysis code should be locked before outcome linkage. Null findings would be reported. Even a single positive study would require independent replication and evaluation against established measures. ¹

Delirium may itself lead to medication changes, which makes the order of administration and assessment essential. Illness severity, surgery, infection, sedation, baseline cognitive impairment, and frailty can influence both medication exposure and outcome. A time-aware design can reduce some ambiguities but cannot remove all observational confounding. Separating ICU care is planned because exposure patterns and assessment workflows may differ from general wards. ¹

The results, if the study is conducted, should report the proportion of drugs and people with eligible model inputs alongside effect estimates and uncertainty. A score could correlate with delirium yet add little to an existing clinical model or be too incomplete for routine use. Conversely, a null result should be preserved without tuning the score to the same outcomes and presenting the retuned value as independent validation. ^{1, 2}

References (AMA style)

1. Chaar M. Step 20B MAR/delirium validation protocol: prespecified future dose-aware study. Unified ACB Tool; 2026. Published September 28, 2026. Source ↗
2. Chaar M. Validation report: NHANES and MEPS evidence, frozen cohorts and limits. Unified ACB Tool; 2026. Published September 28, 2026. Source ↗
3. Inouye SK, van Dyck CH, Alessi CA, Balkin S, Siegal AP, Horwitz RI. Clarifying confusion: the Confusion Assessment Method. Ann Intern Med. 1990;113(12):941-948. doi:10.7326/0003-4819-113-12-941 Source ↗
4. Marcantonio ER, Ngo LH, O'Connor M, et al. 3D-CAM: derivation and validation of a 3-minute diagnostic interview for CAM-defined delirium. Ann Intern Med. 2014;161(8):554-561. doi:10.7326/M14-0865 Source ↗
5. von Elm E, Altman DG, Egger M, et al. The STROBE statement: guidelines for reporting observational studies. Ann Intern Med. 2007;147(8):573-577. doi:10.7326/0003-4819-147-8-200710160-00010 Source ↗