We conducted a retrospective cohort study of all patients admitted to the PICU of CHI Crumlin, Dublin, Ireland between 11th March 2019 and 12th March 2021 who met our eligibility criteria.

The PICU in CHI Crumlin is Ireland’s largest PICU with 23 Beds and approximately 1000 annual admissions [17].

The Delirium QI team prepared a bundle of resources consisting of: instructions and videos publicly available from Erasmus MC-Sophia Children’s Hospital [18]; presentations (tailored to the time available and format of activity) on theory; paper and electronic ‘PD-SOS Process Map’ (Supplementary Fig. 1); a ‘Delirium Algorithm’ (Supplementary Fig. 2) and a paper copy of the SOS-PD score was placed in the bed-side folder of all beds; supervised scoring using real-world patient care scenarios or video resources [11, 12, 18]. The resource material was saved on a shared folder for access and reference by staff during their working shift. The QI team delivered a mixture of individual training including mock scoring at bedspaces and in small groups with nurses, doctors and healthcare assistants, longer format presentations and mock screening sessions with mostly nursing staff, tailored training to doctors as part of their medical education sessions were provided, awareness posters, screensavers, culminating in an ‘official launch’ on World Delirium Awareness Day (WDAD) with posters, quizzes and ‘drop-in’ event for staff. Local training practice for nursing is that 80% of staff are educated prior to commencement of a policy.

Ability to document SOS-PD scoring was added to the PICU clinical information management system in CHI Crumlin in Oct 2019. Staff training was completed and scoring formally commenced as PICU policy on WDAD 11th March 2020. Local policy is to screen all admissions, regardless of age after 24 h. This is despite the SOS-PD tool validation only in patients > 3 months old and admissions > 48 h. This decision was taken to help embed screening in daily practice [11, 12].

To facilitate ongoing training recording of scores was included in the flowsheet for all patients on the clinical information management system (CIMS) (IntelliSpace Critical Care & Anesthesia Information System (ICCA), Philips, United Kingdom) prior to this threshold being achieved. Screening was launched as formal policy on WDAD 11th March 2020.

Patients who met the criteria of the validated population for SOS-PD screening were included. This was all admissions ≥ 48 h, ≥ 3 months of age, who required invasive MV between 11 March 2019 and 12th March 2021. Encounters were divided into two cohorts; Cohort 1 before the official launch of delirium screening as standard of care: 11th March 2019-11th March 2020, and Cohort 2 after the official launch of delirium screening as standard of care: 12th March 2020–12th March 2021. Data were collected from day of admission until day of discharge or day 28 of PICU admission, whichever was sooner. Patients who never underwent invasive MV, were admitted < 48 h or with a history of head injury, neurological complications, comatosed state, deeply sedated or receiving continuous neuromuscular blocking agent infusions, not responding to stimuli and < 3 months of age were excluded.

A formal sample size calculation was not carried out. The study site has approximately 1000 annual admissions to be screened for eligibility. All eligible patients during the study period were to be included to minimise risk of selection bias when sampling from the study population. The study period was limited to one year before and after implementation of screening to limit the effect of temporal changes in practices which may occur as time progresses.

Primary outcomes were pharmacotherapy usage as measured by, percentage of patients receiving, dose per weight per day of use (dose/kg/day) and duration of use of: opioids, benzodiazepines, alpha agonists, sedating antihistamines and chloral hydrate.

Secondary outcomes included: PICU length of stay (LOS), duration of invasive MV, pain (NRS ≥ 5) and withdrawal (SOS ≥ 4) [13], time spent with COMFORT-B score between 11 and 16 [14], delirium screening compliance (% predicted scores), percentage with delirium detected, co-morbid conditions (as defined by Paediatric Complex Chronic Conditions) [19].

Cardiopulmonary bypass time was extracted from the National Institute for Cardiovascular Outcomes Research (NICOR) database [20]. Other data, unless specified otherwise, were extracted from ICCA with assistance from the PICU data manager.

Data were summarized using frequencies and percentages for categorical variables. Continuous data are described by mean and standard deviation, with those displaying some evidence of skew described by median and interquartile range (IQR). Chi-squared tests for categorical variables and Wilcoxon-rank sum for continuous variables were used to explore potential differences between cohorts. Dose per kg per day of use was calculated by dividing the total dose per kg by the number of days of use. Two multivariable logistic regression models for the potential differences between cohorts were explored; model 1 included variables that were clinically relevant and showed evidence of a statistical difference between cohorts and model 2 adjusted for these and risk factors identified in published research [4, 10]. Statistical analysis was undertaken with Stata v18 (StataCorp. 2019. Stata Statistical Software: Release 16. College Station, TX: StataCorp LLC.) and p-values < 0.05 deemed as statistically significant.