The Impact of Implementing a “Pain, Agitation, and Delirium Bundle” in a Pediatric Intensive Care Unit: Improved Delirium Diagnosis

Abstract

Delirium is associated with significant negative outcomes, yet it remains underdiagnosed in children. We describe the impact of implementing a pain, agitation, and delirium (PAD) bundle on the rate of delirium detection in a pediatric intensive care unit (PICU). This represents a single-center, pre-/post-intervention retrospective and prospective cohort study. The study was conducted at a PICU in a quaternary university-affiliated pediatric hospital. All patients consecutively admitted to the PICU in October and November 2017 and 2018. Purpose of the study was describe the impact of the implementation of a PAD bundle. The rate of delirium detection and the utilization of sedative and analgesics in the pre- and post-implementation phases were measured. A total of 176 and 138 patients were admitted during the pre- and post-implementation phases, respectively. Of them, 7 (4%) and 44 (31.9%) were diagnosed with delirium ( p  < 0.001). Delirium was diagnosed in the first 48 hours of PICU admission and lasted for a median of 2 days (interquartile range [IQR]: 2–4). Delirium diagnosis was higher in patients receiving invasive ventilation ( p  < 0.001). Compliance with the PAD bundle scoring was 79% for the delirium scale. Score results were discussed during medical rounds for 68% of the patients in the post-implementation period. The number of patients who received opioids and benzodiazepines and the cumulative doses were not statistically different between the two cohorts. More patients received dexmedetomidine and the cumulative daily dose was higher in the post-implementation period ( p  < 0.001). The implementation of a PAD bundle in a PICU was associated with an increased recognition of delirium diagnosis. Further studies are needed to evaluate the impact of this increased diagnostic rate on short- and long-term outcomes.

Introduction

Delirium in the pediatric intensive care unit (PICU) is a serious problem that has recently attracted much attention. ¹ Delirium is defined as an acute neurological dysfunction characterized by fluctuations in awareness, attention, and cognition. ² Pediatric delirium is associated with poor outcomes, including longer duration of invasive ventilation and length of hospital stay, higher medical costs, and increased mortality. ³ The prevalence of delirium in critically ill children hospitalized in a PICU is estimated at 20 to 38% ^{4 5} and it may affect up to 53% of invasively ventilated children. ⁵ Unfortunately, delirium in critically ill patients, especially children, is challenging to diagnose and delirium is still under diagnosed in PICUs, especially the hypoactive form of delirium. ^{3 6} Recently, diagnostic tools dedicated to screening delirium in PICU patients have been developed. The Cornell Assessment of Pediatric Delirium (CAPD) scale has the advantages of being validated in children from birth to 21 years, including children with developmental delay and of not requiring patient participation. ⁶

Critically ill children are often excessively sedated during their PICU stay. ⁷ Undersedation, oversedation, or insufficient analgesia have consequences, such as inadequate pain control, increased anxiety, and may lengthen duration of invasive ventilation. ⁷ The implementation of a specific bundle for pain relief and comfort is a management strategy aiming to reduce sedation exposure while maximizing comfort. ⁸ In 2013, the Society of Critical Care Medicine (SCCM) published “Clinical Practice Guidelines for the Management of Pain, Agitation, and Delirium (PAD) in Adult Patients in the Intensive Care Unit,” encouraging intensive care units (ICUs) to develop patient-centered protocols to prevent and treat PAD in critically ill patients ⁹ and to implement a PAD bundle in every ICU. In the adult ICU population, a systematic review demonstrated that implementation of a PAD bundle improves clinical outcomes. ¹⁰ A recent study of more than 6,000 patients reported that higher PAD bundle equivalent compliance is associated with decreased mortality and increased delirium-free days. ¹¹ Another recent study has shown reduction in brain dysfunction after implementation of an ICU bundle. ¹² In 2017, Simone et al ¹³ have demonstrated a decrease in pediatric delirium prevalence after ICU bundle implementation, including early mobilization protocols. In 2016, the European Society of Pediatric and Neonatal Intensive Care (ESPNIC) recommended the use of different scales to better assess PAD in PICU. ¹⁴

The primary objective of this study was to evaluate the impact of a PAD bundle implementation on the rate of delirium diagnosis in our PICU. The secondary objectives were to evaluate compliance with the PAD bundle and impact of PAD bundle implementation on the use of analgesic and sedative medication.

We hypothesized that the implementation of a PAD bundle would increase the rate of delirium detection, decrease the use of analgesics and sedatives and increase the non-pharmacological treatment of delirium.

Materials and Methods

Study Design and Patient Population

This is a single-center pre- and post-intervention prospective cohort study. This study was conducted at the Centre Hospitalier Universitaire (CHU) Sainte-Justine PICU, a 32-bed unit within a quaternary university-affiliated pediatric hospital. Baseline (pre-implementation) data were collected in October and November 2017 and post-implementation data in October and November 2018. For both cohorts, all patients consecutively admitted to the PICU for at least 24 hours were included. Pregnant or postpartum women and patients >21 years were excluded. The same calendar period was used for the pre- and post-implementation period to avoid significant differences in population characteristics. The local institutional review board approved this study (no.: 2018–1898) and consent was waived.

Study Procedures: PAD Bundle Development and Implementation

A PAD bundle consists of validated scales to assess pain, sedation, delirium, and withdrawal in children. The PAD bundle used at CHU Sainte-Justine was developed in 2018 by an interdisciplinary team (PICU physicians, nurses, and pharmacists). The scales used are consistent with ESPNIC recommendations (grade A). ¹⁴ Assessments of pain and agitation were completed every 4 hours. COMFORT-B scale was used for invasively ventilated patients. ^{15 16} For noninvasively ventilated patients, the Face, Legs, Activity, Cry, and Consolability (FLACC) scale was used for patients under 6 years old, ¹⁷ visual numeric scale (VNS) for patients 6 years or older, ¹⁸ and revised FLACC scale for children with cognitive impairment. ^{15 19} Level of arousal, agitation, and sedation was measured by the Richmond Agitation and Sedation Scale (RASS). ²⁰ Delirium screening was completed at 8 to 12 hours using the CAPD ¹⁴ , which we had previously translated and validated in French language. ²¹ The CAPD is a widely used screening tool which has been validated in children from birth to 21 years, including children with developmental delay. It is a rapid delirium screening scale that does not require patient participation. ⁶ Finally, the withdrawal assessment tool (WAT-1), a validated tool in pediatric critical care for screening of opioid and benzodiazepine withdrawal, was used every 12 hours in eligible patients.

The PAD bundle implementation started 8 weeks before the prospective phase of the study. PICU nurses and physicians received training on clinical aspects of the bundle. Physicians and nurses participated in a 1-hour educational session that communicated the evidence supporting implementation of a PAD bundle. ^{9 10 13 14} Additional clinical education was provided to nurses on the proper administration of the PAD bundle. After this teaching period, PAD bundle was considered to be routine standard of care for all patients admitted to the PICU. PICU staff was supported throughout this process by the interprofessional PAD team.

Data Sources, Demographic, and Clinical Data

Data were collected through chart review of eligible patients. The following data were collected: demographic data (age and sex), characteristics of PICU stay (main diagnosis, invasive and noninvasive ventilation duration, PICU length of stay, pediatric logistic organ dysfunction score [PELOD 2], diagnosis of delirium and characteristics of the delirium if present, doses of analgesics, and sedatives given during PICU stay). In the implementation phase, results of the various PAD scales were also collected. Data were collected from time of PICU admission until end of PICU stay or day 28, whichever came first.

Outcomes

The primary outcome was the impact of PAD bundle implementation on the rate of delirium detection. For the pre-implementation cohort, the diagnosis of delirium was evaluated by chart review, based on the use of the word “delirium” in any medical notes or by prescription (de novo) of any of the following medications: haloperidol, olanzapine, quetiapine, or risperidone. For the post-implementation cohort, diagnosis of delirium was based on CAPD scores equal or greater than 9 which had to be confirmed by the medical team either by writing the word “delirium” in the medical notes or by discussing the delirium diagnosis and management during rounds, as reported by pharmacists.

Diagnosis rate was defined as the number of occurrences of delirium divided by the total number of admissions for each period of observation (data as reported in the current literature). To account for different lengths of stay between patients, incidence density rates are also reported. Person-time of the “at risk population” was defined as time between admission and occurrence of delirium or end of follow-up (discharge or end of the observation period).

During the post-implementation phase, compliance with scores assessment was evaluated by chart review. To ensure feasibility and consistency of the measurement for each patient, compliance with scores was defined as the presence of assessment of pain and agitation at least every 4 hours and assessment of delirium at least every 12 hours on days 1, 3, and 5 of PICU stay. Another compliance marker was discussion of the PAD bundle scores during daily rounds. In our PICU, pharmacists participate in daily rounds for all patients. Therefore, pharmacists were asked to record if PAD bundle scores were discussed at rounds during the 2-month enrollment period. To conclude that the PAD bundle had been discussed, the results of at least one of the PAD bundle scales needed to be discussed.

The last outcome studied was the cumulative doses of analgesics and sedatives per kilogram and per day. The cumulative dose was calculated for entire PICU stay and divided by the number of days of PICU admission. To allow comparison, opioids were converted to morphine equivalents, benzodiazepines to midazolam equivalents, and clonidine to dexmedetomidine equivalents (the list of dose equivalences used is provided in the Annex). Infusions, boluses, and patient/nurse-controlled analgesia were included. Patients who received patch doses of fentanyl were excluded from this outcome analysis because correlation between fentanyl patch and intravenous doses of opioids are not reliable according to the literature.

Statistical Analysis

Continuous variables are presented as median with interquartile range (IQR). The rate of delirium diagnosis during each study period was compared using Fisher's exact test. The cumulative doses of analgesics and sedatives were compared using Mann–Whitney. Statistical analysis was performed using SPSS 24.0 (SPSS Inc., Chicago, Illinois, United States). Level of significance was set to p  < 0.05.

The sample size was calculated on the basis of the prevalence of delirium, with a margin of error of 5 and a 95% confidence. We expected a difference in delirium diagnosis of 20% (5–25%). In an unpublished study conducted in our PICU before using any delirium screening tool, we found a prevalence of delirium of less than 5%; in the literature, the prevalence ranges from 20 to 38%. ^{4 5} Taking into account, a bilateral type-1 error of 5%, a minimum of 130 patients in the pre- and post-implementation periods was required to achieve a statistical power of 80%. Considering the admission rate in our unit, we decided to conduct the study over two periods of 2 months to reach the expected number of participants.

Results

Population Characteristics

A total of 206 patients were admitted to PICU in the pre-implementation period and 180 patients in the post-implementation period. Thirty (pre-implementation) and 42 (post-implementation) patients were admitted for less than 24 hours and were excluded. A total of 176 and 138 patients were included for analysis in the pre- and post-implementation cohorts, respectively. Demographic data are summarized in Table 1 . The two populations were very similar and comparable. Half of the patients were under age 2 years. Most patients were admitted for respiratory insufficiency or cardiac disease. Approximately 40% of patients required invasive ventilation, and one-third received exclusively noninvasive ventilation.

Table 1. Demographic and clinical description of patients.

All patients	2017 n  = 176 a	2018 n  = 138 a
Age (mo)	28 (5–116)	27 (5–101)
0–23	84 (48)	65 (47)
24–59	21 (12)	19 (14)
60–156	40 (23)	31 (22)
> 156	31 (18)	23 (17)
Male	106 (60)	87 (63)
PELOD score	4 (2–6)	4 (2–7)
Reason for PICU admission
Respiratory insufficiency/failure	70 (40)	55 (40)
Cardiac disease (nonsurgical)	6 (3)	11 (8)
Postcardiac surgery	29 (16)	22 (16)
Postsurgery (noncardiac)	24 (14)	15 (11)
Neurologic disorder	18 (10)	13 (9)
Infectious/inflammatory disease	12 (7)	8 (6)
Trauma	3 (2)	5 (4)
Other	14 (8)	9 (7)
Duration PICU stay (d)	4 (3–7)	4 (3–7)
Invasive mechanical ventilation (h)	66 (37.5)	60 (43)
Noninvasive ventilation (h) b	61 (34.7)	46 (33.3)

Abbreviations: PICU, pediatric intensive care unit; PELOD, pediatric logistic organ dysfunction.

^aContinuous data are presented as median (interquartile range), categorical data are presented as number (%).

^bPatients who have been invasively ventilated are excluded from this category.

Delirium

During the post-implementation period, 64 patients had positive screens for delirium. Of those, 20 (14%) were discarded by the medical team, mostly due to a severe developmental delay or the presence of another underlying diagnosis responsible for the agitation. Detection of delirium significantly increased from 4.0% in the pre-PAD bundle period to 31.9% in the post-implementation period ( p  < 0.001) and the incidence density rates increased significantly from 0.6 to 7.3 per 100 person-days of at-risk population ( p  < 0.001). The rate of delirium diagnosis significantly increased in the post-implementation period in invasively ventilated patients (9.1 vs. 53.3%, p  < 0.001), noninvasively ventilated patients (0 vs. 21.7%, p  < 0.001) and nonventilated patients (0.9 vs. 2.9%, p  < 0.001). The comparison of delirium diagnosis rates between the retrospective and the prospective cohorts is shown in Table 2 .

Table 2. Delirium data.

All patients	2017 n  = 176 a	2018 n  = 138 a	p -Value
Rate of delirium diagnosis	4%	31.9%	<0.001
Incidence density rates (person-days)	0.6/100	7.3/100	<0.001
Rate of delirium diagnosis according to respiratory status
Invasive ventilation	9.1%	53.3%	<0.001
Noninvasive ventilation b	0%	21.7%	<0.001
No ventilation	0.9%	2.9%	<0.001

^aDetection methods of delirium were different between the two periods.

^bPatients who have been invasively ventilated are excluded from this category.

Characteristics of delirium patients in the post-implementation cohort are presented in Table 3 . Delirium was diagnosed early in the PICU stay (day 1, IQR: 0–2) and lasted for 2 days (IQR: 2–4). The diagnosis of delirium was higher in the group with invasive ventilation ( p  < 0.001). Cumulative daily doses of analgesics and sedatives were also higher in patients with delirium ( p  < 0.001). As the pre-implementation cohort only contained seven cases of diagnosed delirium, the characteristics of delirium are not provided.

Table 3. Characteristics of delirium (post-implementation cohort).

All subjects; n  = 138 a	Delirium 44 (31.9%)	No delirium 94 (68.1%)	p -Value
Time from PICU admission to onset of delirium (d)	1 (0–2)	NA	NA
Duration of delirium (d)	2 (2–4)	NA	NA
Cumulative daily dose of opioids–morphine equivalent (mg/kg/day) b	0.45 (0.06–1.07)	0.00 (0.00–0.23)	<0.001
Cumulative daily dose of benzodiazepines–midazolam equivalent (mg/kg/day)	0.00 (0.00–0.09)	0.00 (0.00–0.00)	<0.001
Cumulative daily dose of dexmedetomidine equivalent (mcg/kg/day)	3.14 (0.16–6.87)	0.34 (0.00–2.61)	<0.01
Patients with invasive ventilation ( n  = 60)	32 (53.3)	28 (46.7)	<0.001
Patients with noninvasive ventilation ( n  = 46)	10 (21.7)	36 (78)	<0.05
Patients with no mechanical ventilation ( n  = 32)	2 (6.3)	30 (93.8)	<0.001

Abbreviations: NA, not applicable; PICU, pediatric intensive care unit.

^aContinuous data are presented as median (interquartile range); categorical data are presented as number (%).

^bOne patient excluded due to the use of a fentanyl patch.

Delirium was mostly managed with pharmacological measures when diagnosed in the pre-implementation cohort, with six patients (86%) who received olanzapine and two patients (29%) who received melatonin. For one patient (14%), non-pharmacological measures were documented. In the post-implementation period, 10 patients (23%) received olanzapine and 6 patients (14%) received melatonin. For five patients (11%), non-pharmacological measures were prescribed by the medical team; for 26 patients (59%), non-pharmacological measures were discussed during the daily round and applied but not formally prescribed by the medical team.

Compliance to the PAD Bundle

During post-implementation period, the pain scale was completed at least every 4 hours for 74, 93, and 89% of the patients on days 1, 3, and 5 of PICU stay, respectively, the agitation scale was completed at least every 4 hours for 69, 90, and 89% of the patients on days 1, 3, and 5 of PICU stay, respectively, and the delirium scale was completed at least every 12 hours for 79, 95, and 87% of the patients on days 1, 3, and 5 of PICU stay, respectively. Additionally, results of the PAD scores were discussed during daily rounds in 68% of all patients and in 100% of patients whose scores were positives.

Analgesics and Sedatives

The use of analgesics and sedatives during pre- and post-implementation periods is presented in Table 4 . Four patients from the pre-implementation phase and one patient from the post-implementation phase were excluded from opioid dose comparison due to use of a fentanyl patch. The percentage of patients who received opioids and benzodiazepines was not statistically different between cohorts. The cumulative daily doses of opioids and benzodiazepines was not statistically different either. More patients received dexmedetomidine in the post-implementation period ( p  < 0.001), and its cumulative daily dose was higher ( p  < 0.001).

Table 4. Analgesics and sedatives data.

All patients	2017 n  = 176 a	2018 n  = 138 a	p -Value
Number of patients who have received opioids b	90 (52.3)	82 (59.8)	0.33
Cumulative daily dose of opioids (mg/kg/day)	0.01 (0.00–0.25)	0.08 (0.00–0.41)	0.13
Number of patients who have received benzodiazepines	35 (19.9)	37 (26.8)	0.15
Cumulative daily dose of benzodiazepines (mg/kg/day)	0.00 (0.00–0.00)	0.00 (0.00–0.00)	0.19
Number of patients who have received dexmedetomidine	68 (38.6)	83 (60.1)	<0.001
Cumulative daily dose of dexmedetomidine (mcg/kg/day)	0.00 (0.00–1.99)	1.02 (0.00–4.45)	<0.001

^aContinuous data are presented as median (interquartile range), categorical data are presented as number (%).

^bFour patients in 2017 and one patient in 2018 were excluded due to the use of fentanyl patches.

Discussion

The principal finding of this study is an increase in the rate of delirium diagnosis in PICU patients after implementation of a PAD bundle. In our post-implementation cohort delirium prevalence of 31.9% is consistent with the findings in the literature, based on studies performed in Australia, and the United States. ^{4 5} The medical team discarded 14% of the patients screened positive with the CAPD which is very similar to the original description of the CAPD tool. ⁶ The increased diagnosis of delirium is the result of the implementation of an effective screening tool including the CAPD. The improved screening and diagnosis of delirium observed in our study (4–32%) has the potential to benefit all PICU patients. Delirium is known to be associated with negative short- and long-term outcomes. ^{3 9 12 22 23} In the pediatric population, delirium remains under diagnosed; however, with the implementation of the PAD bundle, we demonstrated the potential to improve diagnosis of delirium. While the short and long-term impacts of the PAD bundle still has to be demonstrated, we hypothesize that patients care could be improved and the negative outcomes due to delirium could be minimized such as long-term brain dysfunction and mortality. ^{11 12 24}

Patients who did not receive invasive ventilation benefitted most from implementation of the PAD bundle. In the pre-implementation period, delirium was diagnosed nearly exclusively in patients receiving invasive ventilation. This is likely due to diagnosis of only hyperactive subtypes of delirium. In invasively ventilated patients, hyperactive delirium is easier to diagnose through agitation and sedation difficulties, even without routine screening. In the post-implementation period, delirium was diagnosed in all patients, whether they were receiving invasive, noninvasive, or no ventilator support. PAD bundle implementation has significantly improved the diagnosis rate of delirium in our unit, probably by improving the diagnosis of hypoactive delirium which the most difficult to diagnose. The hypoactive form is the most frequent form and is associated with worse outcome. ³

In this study, compliance with the different scores of the PAD bundle was between 69 and 95%, which is higher that reported in the literature. ^{25 26} Day three compliance was the highest with the delirium scale reported at 95%. Throughout the PAD implementation process and the post-implementation period, nurses, physicians, and pharmacists were increasingly aware of the importance of delirium due to the numerous reminders made by the research team and this interest in delirium probably explains the high compliance with the delirium screening scale. PAD bundle scores were discussed during rounds for 68% of the patients. Moreover, all the patients for whom the scores were outside of the expected range and required intervention were discussed during the medical round. The high compliance with the PAD had a strong impact on the diagnosis of delirium and should therefore be maintained. The impact of the PAD implementation and its compliance on the long-term remain unknown. Since turnover of health professionals in the PICU is high, delirium awareness will need to be maintained regularly to ensure bundle compliance.

The management of patients diagnosed with delirium also changed between the pre- and post-implementation cohorts. In the pre-implementation period, nearly all delirious patients were treated pharmacologically, mainly with olanzapine, likely due to the diagnosis of hyperactive subtypes of delirium only. Only one patient was treated with non-pharmacological measures. In the post-implementation period, most patients diagnosed with delirium were treated with non-pharmacological measures only. The non-pharmacological approach to manage delirious patient includes the revision of the analgesics and sedatives administered toward an analgesia-first sedation, and the optimization of the patient's environment such as noise reduction and maintenance of appropriate circadian rhythm. This approach is the first-line treatment for patients with delirium. ^{9 27} As a direct consequence of the increased awareness of all the health care providers in our PICU, optimization of the patient's environment has become part of routine care. Moreover, the non-pharmacological measures described above are also part of the important elements to prevent delirium in PICU patients; hopefully, with the increased awareness and the attention paid to the patients' environment, delirium rate could be decreased in our PICU.

In the post-implementation cohort, when delirium was reliably detected, most patients were diagnosed within the first 48 hours of PICU admission and delirium lasted from 1 to 22 days, with a median of 2 days. This is consistent with existing pediatric literature. ²² Demographic characteristics of patients diagnosed with delirium are also consistent with previous publications. ⁵ Delirium was more prevalent in patients requiring invasive ventilation, and patients diagnosed with delirium received a greater amount of morphine cumulative daily dose during their PICU stay. While we cannot draw any causal relationships from the design of this study, our data confirm associated risk factors for delirium. While we did not expect to find significant difference between the doses of analgesics and sedatives used in both periods since the focus was on delirium diagnosis and not on its management; dexmedetomidine was used in more patients and in higher cumulative doses in the post-implementation period. We hypothesize that dexmedetomidine could have been used as a treatment for delirious patients. Indeed, the cumulative daily dose of dexmedetomidine was calculated based on the whole PICU stay and the available data do not allow to investigate if the administration of dexmedetomidine happened before or after the diagnoses of delirium. Another hypothesis is that PICU physicians were more aware of the delirium risk factors and therefore prescribed more dexmedetomidine for delirium prevention, as proposed in the pediatric anesthesia literature. ^{28 29}

Limitations

This study has limitations. As it included patients from a single PICU, findings may not be widely generalizable. Nevertheless, our study populations were similar to other PICU studies and our results are similar to other recent publications from other PICUs. ^{13 25} The pre‐ and post-implementation design of our study allows us to identify associations but not causal relationships. It may have influenced the diagnosis of delirium during the prospective phase since the health care professionals were aware of the ongoing study. Also, the definitions used to diagnose delirious patients were different between the two periods. Finally, the use of chart review to identify treatment of delirium may have prevented identification of non-pharmacological treatment. However, a potential underestimation of nonpharmacologic treatment would not have negatively impacted our results.

In the future, the impact of timely diagnosis and management of delirium on its duration, the use of pharmacological treatment and patient outcome should also be studied. Furthermore, future studies on pediatric delirium should focus on the long-term impact of delirium on patient outcomes in terms of functionality, quality of life, and neurocognitive function, as this has never been described in the pediatric population.

Conclusion

In conclusion, the implementation of a PAD bundle in the PICU improved the rate of delirium diagnosis. The use of a screening tool to detect delirium allowed clinicians to better diagnose delirium. The use of the PAD bundle was particularly significant in diagnosing delirium in patients without invasive respiratory support. Further studies are needed to evaluate the impact of increased delirium diagnosis on our ability to prevent, identify, and intervene to improve short- and long-term PICU patient outcomes.

Funding Statement

Funding This study was supported by the Centre Hospitalier Universitaire Sainte-Justine Research Center.