The quality of the perioperative care that children receive has recently come into question, with high rates of hospital readmissions and emergency department (ED) visits reported after common procedures such as adenotonsillectomy (AT).1,2 Furthermore, United States and Canadian data would suggest that care for this surgical population is variable and does not always conform to published standards.3,4 Care may be influenced by the level of socioeconomic status (SES),5 but this does not appear to be a factor in Canadian or other healthcare systems with universal access.6,7 United States data on pediatric surgical patient readmissions suggest that the majority of these are preventable.8 While preliminary Canadian data support pain and dehydration as the main problems resulting in hospital visits following AT, 4 it is unclear how these revisit rates compare with other common surgeries. Furthermore, data similar to those developed for adult patients—i.e., identifying post-surgery discharge predictor variables for readmission, ED visits, and mortality—are lacking for pediatric patients.9,10

Beyond the quality of care for children undergoing surgery, questions about its safety, particularly in a day surgery setting, are being raised. Specifically, United States reports have raised concerns about home deaths following AT.11,12 These children are vulnerable to adverse respiratory-related events13 because of the surgery itself and the high prevalence of obstructive sleep apnea (OSA). Obstructive sleep apnea is the most common indication for AT in a pediatric setting14 and is associated with increased sensitivity to opioid-related respiratory depression when used to manage post-surgical pain. Opioid-related adverse events (ORADEs), e.g., constipation, are common and have been shown to impact recovery, healthcare costs, length of stay (LOS) and readmission rates in adults.15,16 More serious ORADEs (oversedation and respiratory depression) after surgery17 may result in anoxic brain injury or death.18 As opioids continue to be a cornerstone of pain management, it is important to understand the association between the post-discharge use of oral opioid analgesics and complications specific to AT and to surgeries in general.

The main objective of this population-based study was to estimate the rates of urgent readmission, emergency department (ED) visits, and mortality within 30 days of commonly performed pediatric surgeries, with an emphasis on AT. A secondary objective was to identify factors associated with an increased risk of each of these outcomes. In exploratory analysis, we also sought to determine if post-discharge opioid dispensation increased the risk of these outcomes. We hypothesized that children undergoing AT would have a higher rate of postoperative complications, mortality, and risk of opioid-related outcomes compared with other common pediatric surgical populations.

Methods

Ethical

This study was approved by the Children’s Hospital of Eastern Ontario Research Ethics Board and complies with the REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) statement.19

Historical patient cohort

We identified all hospitalizations with valid Ontario Health Insurance Plan (OHIP) enrolment discharged from April 1, 2002 to March 1, 2013 for children < 18 yr of age with a least one procedure coded. We excluded admissions for neonates, patients undergoing an unplanned surgery within 30 days of a previous hospital admission, and those who were discharged to any location other than the community. From this cohort, we determined the ten most common primary procedures by using the first five alphanumerics of the Canadian Classification of Health Interventions codes. We included only those interventions that were operating room surgical procedures, with some similar procedures being clustered (Appendix A). The study cohort consisted of the first admission for each person for one of the top ten surgeries during the study period. Patients with another top ten procedure coded as a secondary procedure (e.g., AT with ear tube insertion/removal) were excluded.

Databases used in the study

Four provincial population-based health administrative databases in Ontario, Canada (2014 population, 13.6 million) were linked deterministically using encrypted healthcare numbers. Briefly, the Discharge Abstract Database (DAD) captures demographic, diagnostic, and procedural data for all hospital and day surgery encounters and has been determined by medical record re-abstraction studies to be accurate and reliable.20,21 The Registered Persons Database (RPDB) provides basic demographic information (including date of death, neighbourhood, and date of last encounter with the healthcare system) on OHIP recipients. The National Ambulatory Care Reporting System (NACRS) captures all visits to emergency departments. Finally, the Ontario Drug Benefits (ODB) database records all drug prescriptions, including opioids, covered by the provincial drug plan.

Outcomes

Our study included three 30-day post-discharge outcomes: 1) urgent readmission (any unplanned admission in DAD), 2) unscheduled return to the emergency department (any emergency department visit in NACRS), and 3) all-cause death (as determined by RPDB). No data cells with numbers less than six were reported in detail to preserve confidentiality.

Covariates

We determined the following data from the DAD: patient age and sex; hospital setting (tertiary care or community hospital); whether or not the admission was urgent; time in an intensive care unit; hospital LOS; and discharge disposition (home without services, home with services, and against medical advice). Being in receipt of ODB, typical among families on social assistance, was also considered as a risk factor for adverse postoperative outcomes.5 Diagnostic codes in DAD were used to calculate each patient’s Charlson comorbidity index (American Society of Anesthesiologists physical status was unavailable) based on admissions within one year prior to the index admission and chronic diagnoses at the index admission and using category weights from Schneweiss.22 We linked to the RPDB to determine residence (urban/rural) and the median household income quintile.

The capture of drug prescriptions among ODB patients, which represents approximately 20% of the general population,23 provides a unique opportunity to evaluate the postoperative risk associated with opioid prescriptions in the AT population. The status of each patient’s ODB coverage was determined by examining the ODB records during the one year preceding and 30 days following the index admission. Patients with claims were classified as having ODB coverage. Among ODB patients, those given opioids during the 30 days following discharge were classified as “ODB, opioids”, and otherwise, they were categorized as “ODB, no opioids”. Finally, in children who had undergone AT, the reason for each outcome was established based on ICD codes.

Statistical analyses

We determined outcome rates for ED visits, postoperative readmissions, and deaths for each of the top ten most common pediatric surgical procedures identified in the database. Because AT was the most common procedure, the reasons for patient outcomes were determined using the primary diagnoses, where possible. For each outcome, we first identified the 30 most common individual diagnoses. Because of their relevance to AT, respiratory-related individual diagnostic codes (abnormal breathing and asthma) and codes grouped together representing “upper airway infection” and “other respiratory” were also described. Diagnostic codes were grouped into nine AT-related categories (pain, upper airway infection, nausea/vomiting, volume depletion, bleeding, abnormal breathing, asthma, follow-up care, and unspecified complication) or unrelated (other) and stratified by ODB/opioid status.

In exploratory analysis, we used multivariable proportional hazards regression to evaluate the risk of each outcome associated with prescribed opioid analgesia in ODB patients. In this analysis, patient factors were treated as fixed baseline covariates, with the exception of the factor, prescribed opioids, which was entered as a binomial time-dependent covariate. We used fractional polynomials to determine the best transformation for patient age and hospital LOS. In each model, we clustered patients together within hospitals to account for within-hospital similarity.

Adjusted proportional hazards models were created separately for each outcome for each type of surgery to calculate parameter estimates for the influence of prescribed opioids on outcome risks after accounting for ODB status. We compared these parameter estimates between types of operations using 83% confidence intervals (CI) to achieve estimates with an α error equal to 5%.

Results

Participants and surgical populations

During April 1, 2002 to March 1, 2013, 1,621,210 Ontarians (< 18 yr and > one month of age) with a valid OHIP card were discharged from hospital (Fig. 1a). The following patients were excluded: 1,106,701 (68.3%) did not undergo one of the included procedures; 394 (0.02%) were discharged to locations other than the community; 739 (0.04%) were urgently readmitted within 30 days of a previous hospitalization; 83,684 (5.2%) patients underwent more than one of the included procedures; and 65,063 (4.0%) had a prior eligible surgery and were thus already enrolled in our patient-level analytic dataset. Thus, 364,629 patients (22.5% of all discharges) representing 165 distinct hospitals were involved in the study. Nine hospitals were classified as teaching hospitals and accounted for 74,839 (20.5%) of the patients. The ten most common procedures are listed in Appendix A. Tonsillectomy with or without adenoidectomy (AT) was the most common procedure. The AT cohort flow diagram related to ODB classification and opioid dispensation is shown in Fig. 1b.

Fig. 1
figure 1

A) Study cohort flow diagram. B) ODB/Opioid Substudy Cohort Flow Diagram. ODB = Ontario drug benefits. Note: M&T= myringotomy and tube; ODB=Ontario Drug Benefits

Full size image

Demographic characteristics for the full cohort are described in Table 1. The mean age was seven years and females represented 40% of the surgeries but varied by type of surgery (Appendix B, available as Electronic Supplementary Material). Patients rarely had any significant coded comorbidity (Charlson index = 0 in 98.3% of cases). Most patients (84.6%) underwent same-day surgery, and the mean hospital LOS was a third of a day (Table 1). The highest rates of postoperative admission to a special care unit (i.e., a medical, surgical, medical-surgical, or neurological intensive care unit) were for appendectomy (0.6%), hernia repair (0.4%), and AT patients (0.2%) (Appendix B, available as Electronic Supplementary Material). Most patients (89.5%) had no ODB coverage. Among patients receiving ODB (n = 38,405), most resided in the lowest household income quintile neighbourhoods, and the majority (69.4%) did not receive opioids after discharge. The ODB patients prescribed opioid analgesics were older and more likely to have undergone AT.

Table 1 Demographics of pediatric Ontarians undergoing one of ten most common surgeries, years 2002-2013
Full size table

Outcomes

The overall risks of 30-day urgent readmission and emergency room visits were 1.5% (n = 5,475) and 9.2% (n = 33,472), respectively. Appendectomy patients (urgent readmission 3.8%; ED visit 12.8%) and AT patients (urgent readmission 2.7%; ED visit 12.4%) had the highest crude risks of postoperative complications, while circumcision patients had a high risk of ED visits (11.6%) (Table 2). Among AT patients, the most common diagnosis for readmission was related to volume depletion (7.7-10.6%), followed by pain (2.7-5.8%) and nausea/vomiting (0.7-3.9%). Among AT patients, the most common diagnoses for ED visits were pain and upper airway infection (8.0-11.5%), followed by nausea/vomiting (2.1-4.1%) and volume depletion (3.2-3.6%) (Table 3). In the entire study cohort, only ten deaths (0.27 deaths per 10,000 children) were reported within 30 days post discharge (95% confidence interval [CI], 0.18 to 0.39). Surgical populations reporting deaths (n ≤ 5), in descending order of prevalence, were tympanic membrane repair, hernia repair, circumcision, dental restorations, myringotomy & tube (M&T), and AT, but a confidentiality agreement prevented any further analysis surrounding these deaths (Table 4).

Table 2 Province of Ontario raw outcome risk by pediatric surgical study group, years 2002-2013
Full size table
Table 3 Reasons for hospital visit within 30 days following adenotonsillectomy in children as per ODB / opioid substudy groups, years 2002 to 2013 (n = 111,085)
Full size table
Table 4 Thirty-day mortality and associated urgent readmission for “Abnormal Breathing” rates by surgical study group, years 2002-2013 (n = 364, 629)
Full size table

Outcome-associated risk factors

Factors, besides type of surgery, associated with at least a modest increase in the risk (hazard ratio [HR] > 1.15) of readmission and ED visits included previous prolonged LOS ≥ four days, inpatient admission, associated comorbidities (Table 5), and extremes of age, with higher risks noted in the very young (age ≤ three years) and older (age ≥ 14 yr) patients (Fig. 2A and B, available as Electronic Supplementary Material). In addition, compared with children not covered by ODB, the risk of either readmission and/or ED visits was higher across most surgeries for ODB patients not prescribed opioid analgesia. The risk of ED visits was higher still for ODB patients prescribed opioids (Table 6). Compared with most other surgeries, AT generally had a lower outcome risk for ODB children, whether or not they were prescribed opioids. For the population cohort, ODB children had increased crude risks for urgent readmissions and ED visits—i.e., for ODB patients not prescribed opioids (relative risk increase, 28.3% and 26.5%, respectively) as well as for ODB patients prescribed opioids (relative risk increase, 109.5% and 60.5%, respectively) (Table 7, available as Electronic Supplementary Material).

Table 5 Influence of other patient and hospitalization factors on outcome risk in Ontario, years 2002-2013
Full size table
Table 6 Outcome risk by Ontario drug benefits status with or without opioids for the ten most common pediatric surgeries in Ontario, years 2002-2013
Full size table

Discussion

In the province of Ontario, Canada, elective AT was the most common pediatric surgical procedure during 2002-2013. Compared with other common day surgeries, AT patients had the highest risk for a postoperative urgent readmission and were comparable to circumcision patients in terms of risk for an ED visit. The majority of visits following AT were related to potentially preventable pain, nausea/vomiting, and dehydration. Fortunately, death within 30 days of surgery was rare for all surgeries. Prolonged LOS, comorbidities, previous inpatient admission, and age were found to increase the risk of readmission and ED visits following surgery. The ODB coverage was also a risk factor for postoperative hospital visits, particularly when opioids were dispensed.

Our findings are similar to pediatric and adult perioperative literature, with some notable exceptions. Our reported overall 30–day ED visit rate of 9.2% was higher than the overall 30-day postoperative complication rate of 7.4% reported for 2010 by the National Surgical Quality Improvement Program Pediatric® (NSQIP-P) database.24 Nevertheless, the latter collects data only on 30-day readmissions and not on ED visit rates, and it considers only wound/organ system complications and not pain and nausea/vomiting. This may explain why they found otolaryngologic procedures to be among the lowest as regards complications; whereas, we found AT to have the highest rate of post-discharge complications and healthcare resource utilization among day surgeries. Inclusion of inpatient admissions or those considered to have chronic complex comorbidities,25 as in the present study, may explain our higher rates compared with studies specific to the pediatric AT population.3,26 Nevertheless, rates of post-AT hospital visits related to surgical complications, including upper airway infections and bleeding, were similar to other studies3,4,25–27 and similar to the low (3.8%) NSQIP-P post-otolaryngology 30-day complication rate.24 As with other United States3,25–28 and one Canadian-based study,4 pain, nausea/vomiting, and dehydration were among the most common reasons for a post-AT hospital visit/readmission. Capture of all types of associated adverse events following AT likely provides a more comprehensive picture and should be considered in efforts to improve outcomes following surgery.29

Our 30-day all-cause pediatric mortality rates among the top ten most common surgeries were lower than previous reports from 2003-2010, which ranged from 11.0-18.8 per 10,000 children.24,30 This may reflect our predominantly day surgery case mix. Our data show that concerns about excessive postoperative deaths after AT,11,12,26 when compared with other common surgeries, are unfounded in Ontario. This may be due to differences in the surgical setting, healthcare access, and a different population case mix than in the United States, with different comorbidities and associated risk factors. While all Canadian children undergo AT in hospital settings, in the United States, many children now routinely receive care in ambulatory surgery centres25,31 where pressures to “efficiently operate and discharge patients” may negatively impact patient outcomes.11 In addition, hospital co-payments and insurance deductibles in the United States may influence the willingness to seek medical attention. Further, a higher proportion of patients undergoing AT in the United States are African American, a known risk factor for OSA.32 Finally, higher United States mortality rates may be attributed to the 2001 adoption of pain as the “fifth vital sign” by The Joint Commission, which may increase the postoperative complication rate in this at-risk population due to overzealous opioid administration.33

Based on our research, there are other factors beyond surgical procedure associated with the risk of readmission. Those related to prolonged LOS, associated comorbidities (Charlson score > 2), urgent inpatient surgery status, and age may all reflect associated medical complexity. Also, a prolonged LOS may be the result of an unmeasured complication, a known adult risk factor for postoperative morbidity.34 These factors are consistent with hospital-based tools derived in Ontario to predict the risk of postoperative readmission in adults9,10 and reflect experience in children.3,4,35,36 The impact of SES on postoperative outcomes in Ontario requires further study; however, our findings suggest that neighbourhood income quintile does not significantly impact outcomes after pediatric surgery, which is similar to previous reports in universal access health systems.6,7,37 Nevertheless, ODB status, which is a family-level (as opposed to geographic) measurement of low SES and a known risk factor for inferior surgical care and outcomes,5,6,27,38 identified children who were particularly vulnerable to readmission or ED visits. This association may reflect children on ODB having procedures that required more complex care at home (e.g., urological surgeries) or were associated with conditions (e.g., upper airway infections most common following dental and M&T procedures—data not shown) that could not be managed in the community due to known limited access to primary care providers among lower SES families. Otherwise, this could reflect factors associated with a lower SES, such as lack of informal support networks or low household health literacy which can negatively impact recovery from illness. Either way, stratification of future risk and development of processes in pediatric surgery should include appropriate measurement and considerations of SES in the context of family-centred care.

The impact of opioids on outcomes also requires consideration. We were only able to assess the impact of dispensed opioids in ODB patients, which may limit the generalizability of our findings; however, similar associations between opioids and increased readmissions and ED visits have been reported in adults.15 Not surprisingly, opioids were associated with increased nausea/vomiting among ODB AT patients, which could drive increased rates of return to the acute care setting. Furthermore, older ODB patients were dispensed opioids, a known risk factor for the increased pain experienced, which could be exacerbated in a setting like AT surgery where undertreatment of pain is well recognized.4 Regardless of the causal pathway underlying the association between opioids and readmissions or ED visits, future care and research should address the role that opioids may play in promoting adverse outcomes in at-risk groups, including those on social assistance, to delineate their appropriate use.

Our results support ongoing initiatives to heighten the quality of perioperative AT in Ontario.4 Future considerations for such initiatives should include appendectomy, M&T, circumcision, and dental surgical populations based on the absolute number and prevalence of postoperative adverse outcomes reported here. Simpler tools to predict the risks of adverse outcomes, including the prevalence of underlying OSA, are needed, and screening should occur in all surgical populations, not just those undergoing AT. An adaptation of the United States “Perioperative Surgical Home” model39 with standardized interventions, including improved parental education, may reduce the incidence of preventable postoperative pain and dehydration following AT.40 The ODB status and opioid dispensation should be included in a pediatric tool to identify at-risk patients. Finally, a provincial repository of AT cases is needed to allow for cross-referencing with health administrative databases to ensure all cases are accurately captured for research purposes.

The strengths of our study include accurate identification of all readmissions, ED visits, and deaths following the ten most common pediatric surgeries in a universal access healthcare system. As a province-wide study, our results are highly generalizable, as Ontario accounts for over one-third of the Canadian population. That said, our study has some limitations that deserve mention. There is no validated administrative database methodology to determine if all-cause 30-day death or unplanned hospital readmission/visits were related to the index surgery, but this is a standard outcome in the literature. A confidentiality agreement makes it unclear why tympanoplasty, among other surgical populations, was associated with a strikingly higher postoperative mortality rate compared with AT; a coincidence cannot be ruled out. Our databases included only children with a valid OHIP number; however, non-OHIP patients account for only a small fraction of children treated annually. The Charlson comorbidity index has not been validated in children,41 and it is unknown how it correlates with the American Society of Anesthesiologists health status grade. Nevertheless, modest correlation has been shown in certain adult surgical populations.42,43 Although the diagnoses for ED visits following AT were accurately identified, ICD-10 codes have variable accuracy. Nevertheless, the pattern of our results is similar to previous reports. By excluding patients who underwent combined same-day AT and M&T, it is possible that we may not have accounted for all outcomes of interest; however, the number is relatively small—only 10-15% of AT patients undergo M&T. We did not comment on the preventability of hospital visits following other procedures because it was beyond the scope of this paper. Further, we were not able to capture anoxic brain injuries not leading to death, but these are suggested to be low in number.11,12,44 It is unknown whether prescribed opioids were actually ingested.45

To conclude, among children undergoing day surgery, AT is a healthcare resource-intensive procedure in Ontario’s universal access healthcare system. Death after AT is fortunately rare and rates are lower than those reported in the United States, which may be a reflection of differences in surgical population, healthcare delivery, and postoperative procedures. Opioid dispensation in ODB surgical populations is generally associated with increased postoperative hospital revisits; however, this was not unique to AT and therefore requires future study to determine whether a generalizable causal relationship exists. Quality improvement initiatives in the AT population and other surgical populations at high risk of postoperative complications are warranted.