Investigating the effects of cuing medication availability on patient-controlled analgesia pump usage in pediatric patients: Results of a randomized controlled trial

Abstract

Objectives:

The study of Patient-Controlled Analgesia (PCA) behaviors has led to greater understanding of factors that affect the pain experience. Although PCA behaviors can be influenced by cues to medication availability, no studies have examined the effects of such cues in pediatric populations.

Methods:

This randomized controlled trial examined patient satisfaction with pain management, PCA behaviors, opioid consumption and state-anxiety in a sample of 125 post-surgical children and adolescents (7 – 17 years). Patients were randomized to a “light” group (lockout period status cued by light on PCA button) or control group (no cues to medication availability).

Results:

Although cuing did not affect patient satisfaction with pain management (P = 0.11), patients in the light group consumed significantly more opioid than those in the control group (adj. P = 0.016). This effect was primarily due to children in the light group (Median 0.019, Interquartile Range 0.012 – 0.036 mg/kg/hr) consuming more opioid than children (≤12 years) in control group (Median 0.015, Interquartile Range 0.006 – 0.025 mg/kg/hr) (P = 0.007). In contrast to the control group, for patients in the light group opioid consumption was unrelated to pain and the proportion of patients with a 1:1 injections:attempts ratio was higher (P < .001) across the study period.

Discussion:

The overall pattern of results suggests that patients in the light group used their PCA in response to the light, more than in response to their pain, which likely reflects operant influences on PCA behavior by pediatric patients.

Introduction

In 1968, Sechzer published the first study on Patient-Controlled Analgesia (PCA) in adults.¹ Pediatric studies followed 20 years later.² In comparison with intramuscular injections, the systematic titration of small amounts of analgesic medication achieves better pain control with less opioid use and with increased patient satisfaction.^3–6 In addition to these achievements, the early scientists are credited with the insight that pressing a PCA button is a behavioral phenomenon, and as such, is subject to the laws of operant and classical conditioning.^3,6 Indeed, PCA “demand behavior” has come to be understood as a manifestation of the patient’s pain experience.⁶ Despite the progress over the past half-century, however, there is still much to be learned about PCA use and PCA-related behaviors after surgery.⁶

Studies of PCA-related behavior typically examine factors that influence injections or “met demands” (i.e., button presses that result in medication delivery) and attempts or “unmet demands” (i.e., button presses during the lockout period, which do not result in medication delivery). The injections:attempts ratio and the frequency of attempts have been particularly informative, as they have been correlated with various forms of psychological distress and clinical outcomes. For example, perioperative anxiety, worry, emotional distress and avoidance have been shown to be associated with increased attempt frequency, as well as with increased opioid consumption and pain intensity.^7–9

PCA behavior has also been shown to be affected by cues to PCA pump lockout status. Hecker and Albert compared 2 PCA pumps that differed in the cues available to patients.¹⁰ The pump that cued lockout status reduced anxiety and improved perceptions of pain intensity/pain relief, as well as patient and nurse satisfaction with the PCA pump. These findings are congruent with more recent research on adult patients’ frustrations with PCA pumps. Patak et al. found that 49% of the patients in their study did not know if their button press would result in medication, and of those, 22% believed that the “not knowing” made their pain worse.¹¹ When presented with a list of potential improvements, the majority indicated that they would prefer a button with a light that signaled onset/offset of the lockout period. This capability is now a feature of the commercially available Carefusion Alaris^™ PCA pump (CareFusion Corporation, San Diego, CA). While the existing literature would suggest that these new features may be beneficial, they have not been examined in adolescents or children.

Although PCAs have been used to manage pain in pediatric patients for more than 30 years, concerns about safety still exist.¹² Therefore the opportunity to respond to a cue indicating medication availability necessitates an evaluation prior to use in a children’s hospital. Developmental differences in cognition and behavior would suggest that children and adolescents’ use of a PCA may be influenced by a cue to medication availability in different ways, and in ways that are different than adults. For example, limitations in their cognitive abilities may result in children pressing the PCA button whenever the light is green, whereas adolescents’ greater cognitive capacity would allow for more self-reflection prior to pushing the PCA button, regardless of whether the light cue is available or not.^13,14 From a behavioral perspective, it is also plausible that the greater impulsivity in childhood as compared with adolescence,¹⁵ may influence a child’s use of a PCA when the PCA interface (i.e. the button) now has an attention-getting stimulus.

The overall aim of this randomized controlled trial (RCT) was to evaluate clinical outcomes associated with PCA use when the availability and delivery of pain medication were cued. The primary aim was to examine the effects of PCA pump cueing on pediatric patients’ satisfaction with their pain management after surgery. We hypothesized that the cues would increase satisfaction with pain management over the control condition (no cues). Secondary aims focused on whether cueing would affect opioid consumption, PCA behaviors, and anxiety. Based on operant explanations of PCA use behavior, we hypothesized that responding to the light could override responding to pain, resulting in increased opioid consumption, increased opioid-related side effects, and opioid consumption that was unrelated to pain. We hypothesized that cuing would also be associated with fewer attempts and lower anxiety in the light group, as compared with those in the control group. Finally, we explored the hypothesis that the light cue would differentially affect the way children (≤12 years) and adolescents (≥13 years) use the PCA pump, such that in comparison with adolescents, children would consume more opioid, have a greater number of side effects, and make a greater number of attempts..

Materials and Methods

This single-center, prospective, RCT was conducted at a large children’s hospital in the Midwest. The study was approved by the hospital Institutional Review Board, and was registered at www.clinicaltrials.gov (ID: NCT02456909) by the principal investigators prior to study enrollment. Enrollment was April 2015 – August 2017.

Participants

Potential participants between the ages of 7 – 17 years, undergoing an orthopedic, abdominal or thoracic surgery for which PCA is routinely prescribed for post-operative pain management, were identified from the hospital surgery schedule. Eligible patients were opioid naïve, and patients and at least one parent were able to read and speak English. Exclusion criteria included 1) prior experience with PCA, 2) cognitive delay precluding independent use of the PCA button, or 3) pre-operative use of a daily anxiolytic or antidepressant medication.

Procedures

Recruitment and Randomization

Patients were approached immediately prior to surgery. After informed consent and assent were obtained, participants were randomized into either the control group or the experimental “light” group and were stratified by gender and surgical procedure (orthopedic, abdominal, or thoracic) in blocks of 4. The randomization scheme was created by a statistician on the research team using Windows version 6.0 of “rand.exe” (http://block-stratified-randomization.software.informer.com). Participants were enrolled by the clinical research coordinator, who used the scheme to assign participants to groups after consent. Recruitment continued until each group included 60 completed participants (defined as having completed at least 24 hours of PCA use and completed assessment of the primary outcome). Carefusion Alaris^™ PCA pumps were used for all patients. The PCA profile (i.e. configuration of the light cue operation) for both groups was set up by a nurse in the post-anesthesia care unit (PACU). The starting opioid dose is standardized, and was set up by the PACU nurse. For the control group, the PCA pump was used according to the hospital’s standard operating procedures: The PCA emitted a “beep” with every button press, and the status of the green light on the button never changed (i.e., it remained continually lit). For the light group, the PCA pump was configured so that the PCA button only emitted a “beep” when a button press resulted in successful opioid delivery. The green light on the button flashed during opioid delivery, was darkened immediately after delivery and was illuminated at the end of the lockout period, indicating a dose was ready to be delivered. Treatment of side effects was determined by the floor nurse who was not a member of the study team. To ensure consistent messaging, patients and parents in both groups were given scripted instructions regarding PCA use. The patient and guardian were also given a printed copy of the instructions prior to surgery.

PCA monitoring and Pain assessments

Either morphine or hydromorphone (converted to morphine equivalents) was administered using PCA for both groups as prescribed by the anesthesia service. Using morphine equivalents, our standardized PCA orders are as follows: 20 mcg/kg for the PCA dose and basal (if used), 8 minute lockout, and a 120 mcg/kg hourly maximum.

Patients had continuous pulse oximetry. Pain and sedation scores were recorded every 2 hours for the first 24 hours, with a minimum of every 4 hours thereafter. PCA use, including injections (i.e. button presses resulting in delivered doses) and attempts (i.e. button presses during the lockout period), were documented hourly by a nurse who was not a member of the study team. If for some reason the nurse was unable to document the hourly number of injections and attempts (e.g. busy with another patient), the following hour’s data would reflect the previous 2 hours’ worth of attempts and injections.

Data collection terminated when the PCA was discontinued, or up to 60 hours post-surgery, whichever came first. The study period was defined as beginning when the patient arrived in the Post-Anesthesia Care Unit (PACU) through Post-Operative Day (POD) 2.

Measures

All outcomes were collected on post-operative days (PODs) 0–2, with the exception of state anxiety on POD 0, which was collected prior to surgery in the pre-op holding area. The research coordinator administered the STAI once on each POD, and also extracted all other data from the participant’s electronic health record (EHR).

Demographics, anthropometrics, and treatment

Demographics (age, sex, race and ethnicity) and anthropometrics (height, weight, and BMI percentile) were extracted from the EHR. Benzodiazepine (lorazepam, diazepam), acetaminophen, ibuprofen, and ketorolac use and intraoperative methadone administration were also extracted from the EHR. These data were collapsed across PODs, and reported as presence/absence. Pre-existing chronic pain was also extracted from the EHR and reported as presence/absence.

Primary Outcome: Patient satisfaction with pain management

The primary outcome was the patient’s response to a single question, focused specifically on satisfaction with pain management: “How satisfied were you with your pain management (was your pain well controlled)?,” with 0 = “very dissatisfied” and 10 = “very satisfied”. Participants answered additional questions to rule out factors that could potentially explain between-group similarities or differences in opioid consumption (e.g. Did you understand how to use the pain pump?). These questions were completed independently by the patient within 24 hours of the end of the study period. Questions were developed by experts in pediatric pain management and were based on the study by Patak et al.¹¹

Secondary Outcomes

Opioid consumption and opioid-related side-effects

Data on opioid consumption and treatment of related side effects were extracted from the EHR. Total opioid consumed via PCA was calculated from all opioid delivered via the basal rate and via injections, and is reported in mg/kg/hr. Hydromorphone was converted to morphine equivalents by multiplying the total amount of hydromorphone consumed (mg) by 5.¹⁶

Data on side effects requiring treatment included nausea/vomiting, pruritus, over sedation (using the Children’s Hospital of Wisconsin Sedation Scale,¹⁷ and respiratory depression. Treatment was coded as yes/no for each POD.

Pain intensity

Pain scores: Pain scores were recorded from the time patients left the PACU until discontinuation of the PCA or through POD 2 (whichever came first) and were extracted from the EHR. Pain was assessed using the age appropriate, hospital approved pain scale. Children either used a 0–10 (0 = no pain, 10 = worst pain possible) numeric rating scale, or the 0–10 Faces Pain Scale-Revised [10].

PCA usage

The total number of hourly injections and attempts were extracted from the EHR.

State Anxiety

The Spielberger State-Trait Anxiety Inventory for Children (STAI-C), state version, is a well-validated 20 item questionnaire used to assess anxiety about a current situation or impending event¹⁸. Each question begins with the stem “I feel…” and ends with 3 choices (e.g., “very calm,” “calm,” “not calm”). Respondents are instructed to consider “how you feel right now.” A total score is computed from a sum of the items, with scores ranging from 0 – 60. The STAI-C was administered by the research coordinator once before surgery (POD 0) and once per day on POD 1 and POD 2. Elevated state anxiety was based on T-scores and was defined as ≥ 1 standard deviation above the mean.¹⁹

Power Analysis:

An a priori sample size calculation was based on the primary outcome. Rahman et al. examined patient satisfaction with pain relief for those randomized to pain management with PCA vs. intravenous boluses of analgesia titrated to patient comfort levels.²⁰ Based on a 1–5 scale (1 = excellent, 5 = disappointing), the average pain relief rating for patients in the PCA group was significantly better at 2.0 (± 0.37) than the rating for patients in the control group at 3.0 (± 0.24). With a power of 80%, a type I error rate (alpha) of 0.05, and a two-tailed t test, 60 patients would have been needed in each group to detect a difference of at least 0.16 SD between the light and control groups.

Statistical Analysis

Modified intent-to-treat analyses included all randomized participants, with the exception of those who did not receive the allocated intervention, as shown in Figure 1. Means and standard deviations or median and interquartile range (IQR) when data were skewed, were used for continuous data, and categorical items were described using frequency statistics. A Chi-square test or a Fisher’s exact test was used to explore the relations between categorical variables, and bivariate correlations (Pearson correlation test) were used to explore correlations between continuous variables. Comparisons between intervention and control groups and between age groups [children (≤12 years) and adolescents (13–17 years)] were conducted using a Mann-Whitney test. Missing data were treated using listwise deletion, and assumed to be missing completely at random. A two-sided p-value<0.05 was used to define what was considered significant.

Figure 1.

CONSORT diagram.

Multivariable analyses using a mixed model with random time were used to examine the effect of the intervention group and age group on two dependent variables (DV): 1) opioid consumption over time and, 2) injections:attempts ratio over time. The independent covariates (IV) considered were pain scores, methadone use and the interaction between methadone use and intervention group, age group, intervention group, and the interaction between age group and intervention group. A backward model selection procedure was used, with a p-value <0.05 as a criterion to keep the variable in the multivariable model. Based on this criterion, the interaction between intraoperative methadone and intervention group was dropped from the model. The DV opioid consumption was transformed using square root to satisfy model assumptions. The parameters were estimated using Maximum likelihood, and unstructured covariance structure was used for repeated measurements. The mixed model with random effects allows for missing data, under the assumption that data were missing at random. Statistical analyses were calculated using SPSS 24 (IBM SPSS Inc., Chicago, IL) and SAS 9.4 software (SAS Institute Inc., Cary, NC).

Results

Sample characteristics

A total of 647 patients were screened for eligibility, with the final sample consisting of 125 participants (7 – 17 years), 64 in the light group and 61 in control group (CONSORT flow diagram shown in Figure 1). Reasons for non-approach are detailed in Figure 1. Examples of reasons in the “logistics” category include surgery cancelled and no study personnel available to recruit. Examples of reasons in the “other” category include no guardian available to consent, history of drug abuse, continuous sedation prior to surgery, and current treatment for cancer. Early discontinuation from the study occurred primarily due to patients receiving something other than a PCA for post-surgical pain management. The final analyses included 60 in each group with at least 24h of PCA use and completion of the pain satisfaction question. The groups did not differ in any demographic or anthropometric variable examined (P > 0.05), with the exception of weight (Table 1). Patients in the control group were slightly heavier than patients in the light group (P = 0.036) by weight, but not by BMI percentile. No between-group differences were found for methadone administration (light group: 5, 7.8% vs control group: 11, 18%, P = 0.087), or benzodiazepine use (light group: 33, 51.6% vs control group: 36, 59.0% P = 0.40). No between-group differences were found for the presence of chronic pain (light group: 9, 14.1% vs. control group: 8, 13.1%, P = 0.54).

Table 1.

Participant characteristics by group and whole sample.

	Light Group (n=64)	Control Group (n=61)	Total Sample (N=125)
Gender, n (%)
Female	39 (60.9)	38 (62.3)	77 (61.6)
Race, n (%)
White	48 (75.0)	47 (77.0)	95 (76.0)
Black	8 (12.5)	8 (13.1)	16 (12.8)
NH	1 (1.6)	0 (0)	1 (.8)
Asian	2 (3.1)	0 (0)	2 (1.6)
Other	3 (4.7)	4 (6.6)	7 (5.6)
Unknown	2 (3.1)	2 (3.3)	4 (3.2)
Ethnicity, n (%)
Hispanic	2 (3.1)	6 (9.8)	8 (6.4)
Not Hispanic	58 (90.6)	52 (85.2)	110 (88.0)
Unknown	4 (6.3)	3 (4.9)	7 (5.6)
Age (years)
Child
Median	11.00	11.00	11.00
IQR	9.0 −11.5	9.0 −12.0	9.0 −12.0
Adolescent
Median	15.0	15.0	15.0
IQR	14.0 −17.0	14.0 −16.0	14.0 −17.0
Weight (kg)
Median	54.4	60.2	56.0*
IQR	43.0 – 64.1	51.2 – 74.8	45.9 – 69.0
Height
Median	160.4	162.6	162.0
IQR	153.5 −171.8	156.1–173.9	155.0 −173.1
BMI Percentile
Median	59.5	67.0	61.0
IQR	19.0 – 84.0	28.0 – 97.0	22.5 – 91.5
Surgical Type, n (%)
Orthopedic	37 (57.8)	40 (65.6)	77 (61.6)
Thoracic	5 (7.8)	6 (9.8)	11 (8.8)
Abdominal	22 (34.4)	15 (24.6)	37 (29.6)
Baseline State Anxiety
Median	46.0	47.0	43.0
IQR	43.3 – 48.0	41.3 – 49.0	46.0 – 49.0

^*Patients in the control group were slightly heavier by weight than patients in the light group (P = 0.036).

Descriptive data

Descriptive data on secondary outcomes are shown in Table 2.

Table 2.

Descriptive data (Median (IQR)) on secondary outcomes across PODs.

	Pain Intensity (0 – 10)		Opioid Consumption* (mg/kg/hr)		State Anxiety (T-score)
	Light Group	Control Group	Light Group	Control Group	Light Group	Control Group
POD 0	3.96 (2.48 – 5.35)	3.94 (2.76 – 5.50)	0.0252 (.015 – .038)	0.0248 (.015 – .034)	70.00 (69.00 – 72.00)	71.00 (67.00 – 73.00)
POD 1	3.43 (2.41 – 4.68)	3.91 (2.57 – 5.16)	0.0214 (.012 – .031)	0.0163 (.011 – .028)	55.00 (48.00 – 62.00)	55.00 (49.00 – 60.00)
POD 2	4.00 (2.89 – 5.00)	4.13 (2.43 – 5.21)	0.0162 (.0045 – .024)	0.0125 (.0057 – .023)	53.5 (46.00 – 61.00)	52.00 (46.00 – 58.50)

^*Patients in the light group consumed significantly more opioid than those in the control group (adj. P = 0.016). Median opioid consumption data presented to 3 significant figures.

Patient Satisfaction

Responses to the questions about satisfaction and use/understanding of the PCA pump are shown in Table 3. Overall, patients were very satisfied with their pain management (Q1), and groups did not differ in their satisfaction ratings (P = 0.11). For both groups, ratings were very high for the overall experience and ease of using the PCA pump (Q 2–3), and low for worry over getting enough pain medicine (Q5). Groups did not differ on any of these parameters. In comparison with the light group, it was significantly more difficult for patients in the control group to know whether the lockout period was over (Q4; P < 0.0001). The majority of patients indicated that they did not have difficulties using the PCA pump (Q6), that they were provided with adequate information on the use of the pump (Q7), and that they understood how to use the pump (Q8). Groups did not differ on these parameters.

Table 3.

Responses to questions about satisfaction and PCA use by group.

		Light Group (n=60)	Control Group (n=60)	P value
		Mdn (IQR)
Q1	How satisfied are you with your pain management (was your pain well controlled)? 0–10	8.0 (6.0,9.0)	7.5 (5.0, 9.0)	0.11
Q2	How satisfied are you with your overall experience using the pain pump? 0–10	8.0 (7.0, 9.0)	8.0 (6.5, 10.0)	>0.99
Q3	How easy was your overall experience using the pain pump? 0–10	10.0 (8.0, 10.0)	9.0 (8.0, 10.0)	0.31
Q4	How easy was it for you to know when your lockout period was over? 0–10	10.0 (9.0,10.0)	6.0 (3.0, 9.0)	<0.0001 *
Q5	How worried were you about getting enough pain medicine? 0–10	2.0 (0, 5.0)	3.5 (0.5, 6.0)	0.26
		n (%) yes	n (%) yes
Q6	Did you have any difficulties using the PCA pump?	7 (11.7)	11 (18.3)	0.31
Q7	Did you feel you were provided adequate information regarding your PCA pump?	57 (95.0)	56 (93.3)	>0.99
Q8	Did you understand how to use the pain pump?	59 (98.3)	58 (96.7)	>0.99

^*Patients in the light group found it significantly easier to know when the lockout period was over, than patients in the control group (P < 0.0001).

The effect of the light cue on opioid consumption and treatment of side effects

Opioid Consumption

Collapsed across PODs, and after controlling for mean pain intensity, participants in the light group consumed significantly more opioid than those in the control group (adj. P = 0.016). The mixed model (Figure 2) showed an interaction of age group and intervention group, indicating that this effect was primarily due to differences in opioid consumption in children, 6–12 years of age. Specifically, children in the light group (Mdn 0.019, 0.012 – 0.036 mg/kg/hr) consumed more opioid than children in control group (Mdn 0.015, 0.006 – 0.025 mg/kg/hr) (P = 0.007). In addition, opioid consumption decreased from POD 0 to POD 2 (P < 0.0001) for both groups.

Figure 2.

Differences in opioid by age group across PODs.

The between-group differences in opioid consumption cannot be explained by differential usage of a basal. Specifically, 36 youth in the light group and 29 youth in the control group had a basal in place (P = 0.89). The between-group differences cannot be explained by differential use of non-opioid adjuvants. Specifically, across PODs no between-group differences were found for use of acetaminophen (light group: 60, 93.8% vs. control group: 59, 96.7%, P = 0.68), ketorolac (light group: 47, 73.4% vs. control group: 48, 78.7%, P = 0.54), or ibuprofen (light group: 5, 7.8% vs. control group: 5, 8.2%, P = 0.60).

Side effects requiring treatment

Data on side effects that required treatment are shown in Table 4. Significantly more patients in the light group were treated for nausea/vomiting on POD1 than were patients in the control group (P=0.01). No other group differences were found for nausea/vomiting or itching. No patients experienced over-sedation and no patients were administered naloxone for respiratory depression.

Table 4.

Side effects requiring treatment.

	Light Group n (%)	Control Group n (%)	P value
Nausea/vomiting
POD 0	19 (29.7)	17 (27.9)	0.85
POD 1	31 (48.4)	16 (26.2)	0.02
POD 2	14 (30.4)	18 (38.3)	0.51
Itching
POD 0	11 (17.2)	6 (9.8)	.23
POD 1	18 (28.1)	9 (14.8)	.08
POD 2	11 (17.2)	9 (14.8)	.81

^1.Data represents number of patients treated for each side effect listed, collapsed across POD.

The relationship between pain and opioid consumption by group

Pain and opioid consumption were unrelated for patients in the light group on POD 0–2, but were significantly correlated for patients in the control group on POD 0 (R² = 0.26, P < 0.0001) and POD 1 (R² = 0.21, P < 0.001). Although there were only 47 patients in total for POD 2, the control group had again a significant though small correlation between pain and opioid consumption (R² = 0.142, P =0.009). Data for POD 0 and 1 are shown in Figure. 3.

Figure 3.

Relationship between average pain intensity and average opioid consumption by intervention group and POD.

The effect of the light cue on PCA behavior

Injections:attempts ratio with focus on attempts

Injections:attempts data are shown in Figure 4. Over the study period, the number of attempts during daytime hours (6am – 5pm) ranged from 0 – 18/hour. The mixed model showed a significant between-group difference in the number of injections and attempts, such that the control group had a lower ratio than the light group (P < 0.001). This effect was due primarily to teens in the control group pressing the PCA button more frequently during the lockout period (Mdn 0.87, 0.68 – 1.00 injections:attempts) than teens in the light group (Mdn 1.00, 0.86 – 1.00 injections:attempts) (P < 0.001). Further analysis showed that this was true on POD 0 (P = 0.008) and POD 1 (P < 0.001).

Fig 4.

injection:attempts ratio over time for intervention and age groups: Focus on lockout period attempts.

Injections:attempts ratio with focus on pattern of relationship between injections and attempts

PCA injections:attempts across each POD were dichotomized as 1:1 (i.e. over a 24h period, every button press resulted in medication delivery; there were no attempts) or less than 1:1 (i.e. over a 24h period, at least 1 attempt occurred). Significantly more patients in the light group (51%) had a 1:1 injections:attempts ratio over the study period, than participants in the control group (31%) (P < 0.001). Looking at individual PODs, the difference was significant on POD 0 (P = 0.05) and POD 1 (P = 0.03). A histogram illustrating the number of patients with a 1:1 injections:attempts ratio vs. those with a less than 1:1 is shown in Figure 5.

Figure 5.

Histogram illustrating injection:attempts ratios over each POD by group.

The effect of light cue on anxiety

Groups did not differ on state anxiety on any POD. While 65% of the entire sample reported elevated state-anxiety (defined as a STAI T-score ≥70) on POD 0, state-anxiety dropped across POD 1 and 2. Very few patients’ state-anxiety remained elevated across the entire study period (POD 0 – POD 2). Among them, 5.8% (7/120) remained elevated at POD 1 (light group: 6.7% (4/60); control group: 5.0% (3/60)). Of these, 3.7% (3/82) remained elevated at POD 2 (light group: 2.3% (1/44); control group: 4.30% (2/47)).

Discussion

The overall aim of this RCT was to evaluate clinical outcomes associated with PCA use when the availability and delivery of pain medication were cued. The findings show that the cues achieved the purpose of the PCA design. Knowing when the lockout period was over was rated as much less difficult by patients in the light group than by those in the control group. Indeed, patients in the light group made significantly fewer attempts during lockout, than those in the control group. Although this knowledge did not influence patient satisfaction with either pain management or with the PCA pump itself, the cues were not innocuous. The overall pattern of findings suggests that patients in the light group self-administered opioid in response to the light, rather than in response to pain, and this was particularly true for children 12 years and under. In contrast to the control group, patients in the light group consumed more opioid, their opioid consumption was unrelated to their pain, and the proportion of patients with a 1:1 injections:attempts ratio across the study period was almost two times higher in this group. This is the first study to examine the effects of cueing medication availability in a pediatric post-surgical sample, and it extends our understanding of factors that influence PCA usage behaviors in children and adolescents.

Satisfaction with the PCA pump was high for both groups. Although consistent with previous studies of PCA use in pediatric samples, there is room for improvement.²¹ Studies of PCA use in adults have shown that PCAs can be associated with negative expectations and experiences, such as mistrust in the PCA pump itself¹¹. Hecker and Albert showed that cuing medication availability was associated with improved patient and nurse satisfaction with the PCA pump itself, perceptions of pain intensity/pain relief, and with reduced anxiety.¹⁰ Recently, Patak et al. found that adults can be frustrated and dissatisfied with their PCA experience.¹¹ In a study focused on patients’ use of PCAs after surgery, half of the patients did not know if their button push would result in medication delivery, and 22% of those patients believed that “not knowing” made their pain worse. When presented with potential technological improvements to the PCA design, 70% of the sample preferred “…a light on the PCA button to be lit up when the pump was available to deliver more medicine” (p 328). Based on these studies, why we did not see improvement in patient satisfaction associated with the cued condition is unclear. As no studies similar to that by Patak et al. have been conducted with pediatric patients, whether children would express perspectives similar to those of adults is difficult to know. Future studies should strive to understand more fully pediatric patients’ PCA experiences and preferences.

Consistent with our hypothesis, patients in the light group consumed more opioid and required more treatment for side effects (nausea/vomiting) than those in the control group. In fact, children in the light group consumed almost 25% more opioid than children in the control group. Despite these differences, it is important to note that the amount consumed is in line with other reports of opioid consumption in pediatrics,²² and the increased amount was not associated with increased serious adverse events. In their study, Patak et al. acknowledged that cues to medication availability could potentially increase PCA use and thereby increase adverse events and complications from opioids.¹¹ However, they hypothesized that the lockout itself and the requirement that patients are awake to push the button, may be sufficient enough to prevent over usage. The current study demonstrates that these factors are not enough to prevent increased PCA use in children. Behaviorally-based explanations may offer insight into why this increased use occurred.

Principles of both operant and classical conditioning have been utilized to understand the complex nature of PCA behaviors. For example, Katz et al. examined factors that influenced PCA lockout attempts in a sample of women undergoing major surgery.⁹ They found that preoperative intrusive thoughts and avoidant behaviors were directly associated with increased attempts, but were not associated with untreated pain, anxiety, or negative affect. The authors suggested several behaviorally based explanations. First, self-administration of opioid via the PCA pump can be considered escape/avoidance learning. Patients may press a PCA button to escape from the fear of pain or may press in anticipation and fear of future pain. This behavioral pattern is consistent with negative reinforcement of behavior, and would be applicable to PCA pump usage regardless of whether medication availability is cued or not. As the authors point out, patients are commonly instructed to press the button prior to movements that may exacerbate pain, which is also true for pediatric patients.^23,24 Second, while not testable under their paradigm, the authors suggested that any environmental cues (e.g., the PCA button itself) may serve as a “safety signal” which cues the absence of an aversive state. Alternatively, the authors hypothesize that, via classical conditioning, any environmental cue could take on secondary reinforcing properties when paired with a potent unconditioned stimulus, such as morphine. Based on this hypothesis, the cue would elicit an analgesic effect even in the absence of a medication dose, and would likely be associated with decreased pain and PCA usage over time.

To explain the observed benefits of cueing medication availability, Hecker and Albert suggested that the cue may have acted as a discriminative stimulus (S^d)¹⁰. A S^d is said to “set the occasion” for a behavior being followed by a reinforcer. Its “presence” serves as a cue that the behavior (pushing the PCA button) will be reinforced, whereas its absence serves as a cue that the behavior will not be followed by a reinforcer. This explanation is similar to Katz et al.’s hypothesized safety signal, but would be best characterized by increased responding in the presence of the cue (button illuminated), with decreased responding in the absence (button dark). Finally, others have noted that PCA parameters make self-administration of opioid medication akin to a fixed-interval (FI) schedule of reinforcement, because the cue may increase awareness of the timing associated with the FI.⁶

The overall pattern of results in the current study suggests that in pediatric patients, especially children under 12 years, the cueing may result in stimulus control of behavior. The key is in the lack of a relationship between pain and opioid consumption in the light group, coupled with the number of those patients with a 1:1 injections:attempts ratio. Almost twice the number of patients in the light group had this behavioral pattern despite no between-group differences in anxiety or pain. Why stimulus control of behavior would affect more children than adolescents is unclear. Future studies should continue to examine age-group differences in PCA use behavior.

Finally, our hypothesis that the cues would reduce anxiety was not supported. The state-anxiety levels reported by patients in our study are consistent with reports of elevated pre-operative anxiety in children.²⁵ Given the demonstrated relationship between perioperative psychological distress and PCA behaviors as well as pain and opioid consumption in both children and adults,^8,9 it was possible that the cues would benefit patients by reducing anxiety. It is also likely that the STAI did not capture other relevant forms of anxiety that could be affected by the light cue, such as fear of pain, pain catastrophizing, or anxiety sensitivity. Further study of perioperative anxiety that includes nuanced aspects of anxiety in relation to PCA use in children is needed.

Strengths and Limitations

Since the inception of PCA technology in the late 1940s, this is only the second study to examine the results of cueing medication availability on PCA behaviors, and the first of its kind in pediatrics. Nonetheless, the results must be taken in the context of study limitations. We did not design the study to examine all factors that may have been affected by PCA cues, such as preoperative anticipated pain, as found to influence PCA behaviors in adolescents.⁸ It is plausible that other distress-related factors may have given a different perspective on the relationship between the light cue, anxiety, pain and PCA usage. Further exploration of such factors may also explain why the adolescents in the control group frequently pressed the PCA button during lockout. Additionally, although the control group had no overt cues to the lockout status of the PCA pump, the pump itself emitted a noise during medication delivery. Not unlike the auditory cue, this sound (although nearly imperceptible and not pointed out to families) could have served as a cue that medication was actually delivered and the lockout period started. This alone, however, would not explain the overall pattern of between-group differences. While unlikely, it is possible that prescribing providers’ awareness of the study may have introduced bias. However, while it is plausible that prescribing providers could have been aware of the study, it is highly unlikely that providers would be aware of group assignment. The PCA profile is not visible on the main screen of the PCA, and the provider would have had to notice that the green light on the PCA button was not illuminated. Finally, It is not known how often parents reminded or prompted their child to push the PCA button, regardless of group or patient age. It is possible that some parents more concerned about pain may have prompted their child (children and adolescents) to push the PCA button more often than other parents. It is not known if the light cue may have resulted in less prompting, or if the cue made no difference at all.

This study expands our understanding of factors that affect PCA usage in pediatric samples, and contributes to the evidence that pressing a PCA button is a complex behavior that is subject to the principles of operant conditioning. The results suggest that the light cue may result in increased opioid consumption, unrelated to pain for younger patients. While providers may want to consider this cue for adolescents, at this time it cannot be recommended for children 12 years and under.