The Prevalence of and Risk Factors for Adverse Events in Children Receiving Patient-Controlled Analgesia by Proxy or Patient-Controlled Analgesia After Surgery

Abstract

BACKGROUND: 

Recent reports emphasize the risks associated with patient-controlled analgesia by proxy (PCA-P), yet data regarding such risks in children remain sparse. We compared the prevalence of clinically significant adverse events in children receiving PCA-P versus PCA, and examined factors that place children at increased risk.

METHODS: 

The records were reviewed of opioid-naïve children, ages birth to 18 yr, who received PCA or PCA-P after surgery. Data included demographics, comorbidities, perioperative information, pain, sedation, and respiratory assessments, oxygen saturation, analgesics, adverse outcomes, and interventions.

RESULTS: 

This study included 145 children who received PCA-P and 157 PCA. The PCA-P group was younger and had more comorbidities (P < 0.05). Opioid orders were similar, but pain scores and opioid dosages were lower, and fewer children received diazepam in the PCA-P group (P < 0.05). Clinically significant adverse events (i.e., those requiring intervention) occurred in 22% and 24% of patients in the PCA-P and PCA groups, respectively; however, more children in the PCA group had “threshold events” (minor intervention) and more in the PCA-P group had “rescue events” (opioid reversal or escalation of level of care). Respiratory events occurred earlier in the PCA-P group (P < 0.05). Factors associated with adverse events included orthopedic surgery, cognitive impairment, respiratory comorbidity, use of continuous basal opioid infusion, use of diazepam, and larger opioid doses on days 1, 2, and 3. Yet, cognitive impairment and opioid dose on day 1 were the only factors independently predictive of these events.

CONCLUSIONS: 

This study found that although a significant number of children receiving PCA and PCA-P experienced adverse events, there was no difference in the prevalence between groups. The PCA-P group was at greater risk for events requiring rescue interventions, perhaps due to the prevalence of underlying comorbidities. These findings emphasize the need for vigilant monitoring to facilitate early recognition and timely intervention of respiratory depression.

In Brief

IMPLICATIONS: This study describes the prevalence of and risk factors for adverse events in children receiving patient-controlled analgesia by proxy (PCA-P) and PCA. A significant number of children receiving these therapies experience clinically significant events, warranting close monitoring throughout the use of PCA-P and PCA.

Patient-controlled analgesia (PCA) has become a mainstay of postoperative pain management for many children because of its reported efficacy and safety.^1–3 This method of opioid delivery was modified to PCA by proxy (PCA-P; i.e., nurse- or caregiver-controlled analgesia) for young children or those with cognitive impairment, and found to be safe and effective in several studies.^4,5 However, data regarding the risks of these methods in children remain insufficient. Reports from the Joint Commission on Accreditation of Healthcare Organizations⁶ and the Institute for Safe Medication Practices (ISMP)^7,8 have raised safety concerns regarding the use of PCA and PCA-P. In a sentinel event alert, the Joint Commission on Accreditation of Healthcare Organizations cited data in which 15 of 460 deaths and/or serious adverse PCA events were the direct result of PCA-P (12 cases attributed to family members and 3 to health care workers), suggesting that well intentioned efforts by caregivers to maintain patient comfort can result in over-sedation, respiratory depression, and even death.⁶ Furthermore, the Anesthesia Patient Safety Foundation recently suggested that the risk of serious injury from PCA after surgery may be “underappreciated,”⁹ which has been validated by the recent increase in proportion of ASA Closed Claims associated with acute postoperative pain management.¹⁰ Lastly, in a recent survey, members of the Society of Pediatric Anesthesia recalled at least 154 patients over the past year who received naloxone and 18 over the past 5 yr who died due to opioid adverse effects.¹¹

Such data have sparked recommendations for specific safety policies for the use of PCA and PCA-P. The ISMP suggested that analgesia-related adverse events may be linked to insufficient patient monitoring,⁷ and the Anesthesia Patient Safety Foundation issued an initiative urging the consideration of continuous monitoring of oxygenation and ventilation for patients receiving PCA postoperatively.⁹ The ISMP further suggested that practitioners identify patients at risk for opioid-related respiratory depression and determine the level of enhanced monitoring required during PCA use in these populations.⁸

Despite these recommendations, there remains no consensus regarding appropriate physiologic monitoring of adults or children receiving PCA or PCA-P. Furthermore, it remains unknown whether risks are greater for populations who cannot self-report their pain, and for whom judgments are being made for treatment, such as those receiving PCA-P. Further data are warranted to identify the relative risks of respiratory depression in children receiving PCA or PCA-P, and to define the best monitoring practices for these populations. The purpose of this study was to compare the prevalence of adverse outcomes in opioid-naïve children receiving PCA to those receiving PCA-P for routine postoperative analgesia. Additionally, this study examined risk factors for adverse respiratory events in children who received these methods of opioid delivery.

METHODS

With approval from the IRB at The University of Michigan, all children who received PCA-P or PCA between January 2006 and January 2007 were identified from our anesthesia pain database (which captures all cases of PCA, PCA-P, epidural and pain consultations on a daily basis), and a random sample was generated using the random sample selection function in SPSS. Children with chronic pain conditions, those taking opioids preoperatively, those requiring mechanical ventilation and sedation postoperatively, those in chronic renal or liver failure, or those who received epidural analgesia were excluded.

Per routine practice, all children were evaluated by the Pediatric Pain Service at least twice daily, and oxygen saturation (Spo₂) was monitored via continuous pulse oximetry for the duration of PCA/PCA-P treatment. Pulse oximetry alarms generated automated nurse call light notification after 15 s, a page to the bedside nurse and charge nurse after 1 min, and an emergency nurse group page after 3 min of sustained oxygen desaturation. Institutional policy stipulated that administration of bolus doses via PCA-P were to be provided only by nurses or physicians. Preprinted PCA order forms that included dosing ranges and defined rescue criteria were used in all cases. Rescue orders defined the specific variables [i.e., sedation scores, respiratory rate (RR), and Spo₂] based on the child’s age and baseline information at which to; (1) hold PCA and page the pain service, (2) administer supplemental O₂ and page pain service, and (3) hold PCA, administer naloxone and STAT page the primary service and the pain service.

Data were obtained from the electronic perioperative records, preoperative history and physical examination, PCA/PCA-P orders and flowsheets, pain team documentation, nursing flowsheets, medication administration records, and progress notes. The following data were recorded: demographics, comorbidities, surgical procedure, perioperative analgesic dosages, initial PCA/PCA-P orders, timing and nature of order changes, pain assessment tools used, and frequency of pain, sedation, and respiratory assessments. The following data were captured hourly for the first 4 h, every 4–8 h through 24 h, and then daily through 72 h after surgery: total opioid, benzodiazepine, and other analgesic doses administered, highest pain and sedation scores, lowest RR and Spo₂, and use of supplemental O₂.

All medical record sources and the pain database were used to document the occurrence, nature, and timing of adverse events, which included: oxygen desaturation, bradypnea, over-sedation, and any other critical event. The following definitions were used to evaluate adverse outcomes in this study: oxygen desaturation was coded as minor (>5% to <10% decrease from baseline) and major (≥10% decrease) for general reporting; bradypnea was defined as RR ≤16 for age<6 mo; RR ≤12 for age 6 mo to 2 yr; RR ≤10 for age 3–10 yr; RR ≤8 for age >10 yr; over-sedation was defined as a University of Michigan Sedation Scale score >2 (i.e., difficult to arouse to unarousable; deep sedation)^12–14 in conjunction with an intervention and pain service note regarding excessive somnolence. Clinically significant events were coded as (1) “Threshold Events,” which included oxygen desaturation, bradypnea or over-sedation documented in conjunction with a first-line intervention such as decreasing or discontinuing the opioid or benzodiazepine dose, addition of supplemental O₂, patient stimulation; and (2) “Rescue Events,” which included those treated with naloxone, airway management (insertion of an artificial airway and/or bag/mask ventilation), or escalation of care [i.e., admission to an intensive care unit (ICU) or moderate care unit]. Nausea, vomiting, and pruritis and concomitant order changes related to these side effects were not recorded for the purposes of this study.

A sample size calculation determined that 139 cases per group (α = 0.80; β = 0.05) were required to reject the hypothesis that 25% of those receiving PCA-P would experience hypoxemia⁵ compared with 12.5% of those receiving PCA. Morphine equivalents were calculated for all opioids administered (i.e., IV or oral) using standard equivalency formulas (e.g., hydromorphone:morphine = 0.15:1).¹⁵ Nonparametric data including the presence/absence of comorbid conditions were compared between groups using χ² and Fisher’s exact tests where applicable. Parametric data including analgesic dosages were analyzed using analysis of variance or unpaired t-tests as appropriate. Factors that were significantly associated with adverse events by univariate analysis were subsequently entered into a logistic regression model to determine independent risk factors. Data are presented as n (%) or mean ± sd as appropriate. P values of <0.05 were accepted as statistically significant.

RESULTS

Eight hundred and thirty-three children from our database met inclusion criteria over the study period, and the selection process provided a sample of 302 children (145 PCA-P and 157 PCA). Children who received PCA-P were younger and had more comorbid conditions compared with those who received PCA (Table 1). There were no differences in the initial opioid orders between groups, with morphine ordered in 95% and 91% of children in the PCA-P and PCA groups, respectively, and hydromorphone in the remaining children. A continuous basal infusion (CBI) was ordered in 71% and 70% of the PCA-P and PCA groups, and there were no differences between groups in the orders for CBI rates (0.007 ± 0.005 vs 0.008 ± 0.006 morphine mg · kg⁻¹ · h⁻¹), bolus doses (0.019 ± 0.004 vs 0.018 ± 0.047 mg/kg), or 4 h limits (0.26 ± 0.05 mg · kg⁻¹ · 4 h⁻¹, both groups). The duration of PCA/ PCA-P use was shorter for the PCA-P group (1.9 ± 1.3 days) compared with the PCA group (2.4 ± 1.7; P = 0.002).

Table 1: Description of the Sample

Self-reported scores (i.e., either 0–10 numbers or Faces Scales) were used to assess pain in 151 (97%) and 59 (41%) children in the PCA and PCA-P groups, respectively; and behavioral assessment using the FLACC tool¹⁶ was documented in 59 (38%) and 131 (91%) children in these groups. There were no differences between groups in the frequency of pain, sedation, or respiratory assessments that were recorded every 2–4 h (median) for 48 h, and every 4 h on postoperative day (POD) 3. Spo₂ was documented hourly for the first 48 h and every other hour during POD 3. RR was documented every 4 h and quality of ventilation every 8 h through POD 3.

Pain scores were lower for the first 3 PODs in the PCA-P group; however, depth of sedation was similar (Table 2). Furthermore, children in the PCA-P group received lower opioid doses on all 3 PODs, and fewer received diazepam on PODs 2 and 3 compared with the PCA group (Table 3). There was a similar prevalence of respiratory events (bradypnea, minor, and major oxygen desaturation); however, more children in the PCA group experienced threshold events and more in the PCA-P group had rescue events including: 6 cases of airway obstruction requiring chin lift/jaw thrust maneuver and placement of an oral/nasal airway (n = 4) or endotracheal intubation (n = 2); 4 cases of repeated hypoventilation and oxygen desaturation treated with naloxone (n = 3) or intubation (n = 1) and monitoring in the ICU; and 1 case of stridor treated with racemic epinephrine and ICU observation. None of these children required resuscitation, and all events resolved without sequelae. Supplemental O₂ was the most common intervention for children with events (Table 4). Respiratory events occurred earlier postoperatively for children who received PCA-P compared with PCA (Table 4). Two patients in the PCA-P group received bolus doses from a parent, despite institutional policy disallowing this practice. One case was discovered after the child became excessively sedated and experienced respiratory depression, whereas the other experienced no adverse events. In both cases, parents were reminded to not push the button, and there were no further sequelae.

Table 2: Postoperative Assessments in the Groups

Table 3: Analgesic Dosages Administered in the Groups

Table 4: Adverse Events in the Groups

Factors associated with adverse events are presented in Table 5. These factors were entered into a stepwise (backward) logistic regression model, and of these, the following were found to be independently predictive of overall adverse events: cognitive impairment (P = 0.039; Wald statistic 4.27) and morphine equivalents on POD 1 (P < 0.001; Wald 21.44).

Table 5: Factors Associated with Adverse Events

DISCUSSION

This retrospective study found that 24% and 22% of monitored, opioid-naïve children who received postoperative PCA or PCA-P, respectively, experienced clinically significant adverse events (i.e., those requiring intervention). Children in the PCA group experienced more threshold events (i.e., managed with first-line interventions including decreasing the opioid dose or use of supplemental oxygen), whereas those in the PCA-P group had more rescue events (i.e., associated with use of naloxone, airway management, or admission to the moderate care unit/ICU). Lastly, cognitive impairment and opioid dose on POD 1 were independent predictors of clinically significant adverse events.

Although this study found no difference in the prevalence of adverse events between the PCA-P and PCA groups, clinically significant events occurred in almost one-quarter of all children, and rescue events occurred in <3.6%. Data regarding hypoxemia and bradypnea in children receiving PCA or PCA-P are sparse. Monitto et al. found that 25% of children receiving PCA-P required supplemental oxygen to maintain Spo₂ ≥95%, and 1.7% required naloxone for apnea and oxygen desaturation.⁵ More recently, Anghelescu et al. found that <1% of pediatric oncology patients receiving PCA-P required naloxone for adverse effects; however, hypoxemia data were unavailable from their study, as children were not routinely monitored with pulse oximetry.⁴ Overdyk et al. recently described a 41% and 12% incidence of oxygen desaturation and bradypnea, respectively, in adults monitored continuously during the first 24 h of PCA postoperatively.¹⁷ Only one patient in their sample required airway intervention and none required naloxone, which may have been curtailed with continuous monitoring and alarm conditions present in their setting. Differences in findings in the above studies may be related to populations, definitions of respiratory depression, and monitoring.

Although several factors were associated with adverse events in this study, opioid consumption and the presence of cognitive impairment were independent predictors of these events. Previously, CBI was associated with increased opioid consumption, greater sedation, and respiratory depression in children aged 6–12 yr.¹⁸ Our data similarly found that use of CBI was associated with adverse events, yet earlier studies^1,2 refuted this notion, likely due to differences in populations and opioid dosages. PCA-P was not a risk factor for overall adverse events, but children receiving this therapy were twice as likely to receive rescue interventions when events occurred. Since this group had more underlying comorbid conditions, including cognitive impairment, compared with the PCA group, this finding suggests either an increased risk for severe events or, perhaps, a lower clinical threshold for treating these events in children with comorbidities. Interestingly, the PCA-P group had lower pain scores, lower opioid, and benzodiazepine use, despite similar surgical procedures as the PCA group. Weldon et al. described similar findings in their sample of developmentally delayed children and suggested that these differences may, in part, result from perceptions that this group is more vulnerable to the respiratory depressant effects of opioids.² Our finding that cognitive impairment independently predicted adverse events, despite lower opioid consumption, lends support to this assumption. Additionally, children with cognitive impairment in this study were similar in age and underwent similar procedures compared with those without cognitive impairment. It is therefore likely that some other physiologic difference placed these children at increased risk for opioid-related adverse events.

Our data, as well as those from Overdyk et al.,¹⁷ suggest that pulse oximetry provides good information regarding threshold criteria, which may prevent worsening respiratory depression. Although the mean time to events in the present study was 16–27 h, events occurred through the third postoperative day, challenging previous notions that continuous monitoring be used only during high risk conditions such as the first 6–24 h of PCA-use postoperatively.¹⁹ Noninvasive capnography has been shown to promote earlier detection of respiratory depression compared with pulse oximetry in adults and children,^20–23 particularly during the use of supplemental oxygen.²⁰ Despite its potential inaccuracies in nonintubated patients, Overdyk et al. suggest that capnography provides more reliable RR information and has potential value in providing trend analysis during PCA use in adults.¹⁷ The value of capnography in promoting safe care of children receiving PCA or PCA-P merits further investigation. Lastly, close monitoring of sedation depth during PCA use has been suggested, since excessive sedation may be a precursor to respiratory depression during opioid use.^19,24 Half of the children in this study exhibited moderate levels of sedation during each of the first three PODs, lending support to the notion of careful sedation assessment during PCA and PCA-P use in children. Additionally, Taylor et al. found that, during the first 24 h of PCA use postoperatively, 73% of adults reached sedation levels similar to patients sedated for colonoscopy, emphasizing the need for vigilant monitoring during PCA use.²⁵

Importantly, there are several limitations posed by this study design. This report is based on documentation and not continuous data capture, so that the duration of desaturation events remains unknown. Additionally, there is the potential for reporting bias despite the nurse notification system in place over the course of this study. Future research using continuous data capture¹⁷ may provide greater insight into the incidence and duration of respiratory depression in children. Small samples of children with certain comorbidities, and the possibility of differences in their clinical management, reduced the power of this study to determine their association with adverse events. Additionally, it remains unknown how adverse events in children receiving intermittent IV or oral opioids, or no opioids postoperatively, would compare to our findings. These data are unavailable in healthy children. Lastly, study findings cannot be extrapolated to children with chronic pain or those receiving long-standing opioid therapy, and it remains unknown whether PCA-P with parental dosing poses more or less risk than nurse/physician initiated dosing.

In summary, findings from this study suggest that a significant number of opioid-naïve children receiving PCA and PCA-P experience clinically significant adverse events. Although there were no differences in the prevalence of events between groups, children who received PCA-P were at increased risk for events that required opioid-reversal, airway intervention, and escalation of care. These findings suggest the need for pulse oximetry and sedation monitoring during PCA/PCA-P use to facilitate early recognition of respiratory depression and timely interventions to avoid undesirable outcomes.

ACKNOWLEDGMENTS

The authors thank Robin Snyder and Sky Yang for their work in collecting data for this study.