Descriptive statistics including frequency (%) and interquartile ranges in the PRE and the no PRE groups were computed. Matched analysis included Wilcoxon signed rank test for continuous variables and McNemar χ2 test for categorical variables. To compare the main outcomes between 2 groups, comparisons on the outcomes of interest were adjusted for potential confounders using regression models.
Model for Any Hospital Stay and Number of Days Hospitalized After Surgery
The dataset included both outpatient and inpatient surgeries; data of outpatients on hospital stay contained many zeros (discharged the same day), whereas inpatients mostly contained positive counts (number of days hospitalized after surgery). To examine the relationship between the main exposure (PRE versus no PRE) and any hospital stay and number of days hospitalized after surgery, a hurdle model containing both binomial probability to predict nonzero counts (i.e., hospital stay after surgery) and truncated negative binomial-at-zero count for positive values (number of days hospitalized after surgery) was constructed.22 From the regression, there were 2 sets of predictors. The first predicted whether the patient had any hospital stay after surgery. The second predicted the number of days of hospitalization after surgery among those who were admitted. The model was refined by sequential backward elimination of nonsignificant variables guided by the likelihood ratio test, providing coefficients and their 95% confidence intervals. The exponential of their coefficients (odds ratio for hospital stay after surgery and count ratio for number of days hospitalized after surgery) were displayed and considered significant if the likelihood ratio test P values were <0.05.
Matched Analysis Model for Adjusted Excess Hospital Cost and Indirect Cost
To model the relationship between PRE and excess hospital cost and indirect cost, a multiple mixed-effects linear regression model was developed. The type of hospital payment system was also included in the analysis for cost to examine whether the effect of PRE was influenced by payment system. To fit the residual of linear distribution assumption, the so-called adjusted excess cost obtained by a combination of the log of excess cost beyond 2200 baht (for outpatients) and the log of excess cost beyond 3600 baht (for inpatients) was used for the final cost parameter. The exponential of their coefficients (cost ratio) and 95% confidence intervals were displayed and considered significant if the F test P values were <0.05.
To determine the impact of severity of PRE on hospital stay and cost, PRE was replaced with a variable indicating severity of PRE after obtaining the final model. The pairwise interactions with PRE and severity of PRE were evaluated for each final model.
Sample Size Calculation
A pilot study comparing number of days hospitalized and excess hospital cost between the 2 groups was performed before the main study was performed. Thirty-eight PRE and 38 no PRE children were included. The means and SDs of the number of days hospitalized of the respective group were 3.2 ± 4.2 and 2.2 ± 2.3 days, respectively. The means and SDs of the excess hospital costs were 21,808 ± 43,976 and 11,068 ± 6252 baht, respectively. To detect a difference of these magnitudes with a power of 80% and type I error of 5%, 181 children per group for number of days hospitalized and 135 children per group for excess hospital cost were required. Therefore, at least 202 children per group were required to compensate for 10% dropout in the study.
According to the low incidence of PRE (5%) in our hospital,23 14 months of data collection were performed to adequately obtain the required sample size.
A total of 1007 of 2455 eligible children were included in the study from November 2012 to December 2013 at Songklanagarind Hospital. Figure 1 shows a flow diagram of the study with the matching selection procedure for no PRE children. Sixteen PRE children could not be matched. Therefore, 215 matched pairs (215 children per group) were included for the matched analysis.
Among the PRE group, types of PRE were 190 desaturation, 23 upper airway obstruction, 19 laryngospasm, 10 wheezing, 3 endobronchial intubation, 3 respiratory depression, 2 esophageal intubation, 1 pulmonary aspiration, 1 accidental extubation, and 1 reintubation. A child could have >1 type of PRE. Forty-seven (22%) and 168 children (78%) had severe PRE (SpO2 ≤ 85%) and mild-to-moderate PRE (SpO2 > 85%), respectively.
Tables 1 and 2 compare baseline demographic data, respiratory-related, and anesthesia-related variables in children with and without PRE. The 2 groups were well balanced in their baseline characteristics except for anesthetic time where the PRE group tended to be under general anesthesia for a longer duration (P = 0.015). The major hospital payment system was universal coverage (72% in PRE and 75% in no PRE group). Table 3 provides the summary statistics of the charge data and analysis by category. PRE children had a higher proportion who required a postoperative oxygen device and mechanical ventilator (P < 0.001), had a longer PACU stay (P < 0.001), were more likely to require hospital stay after surgery (P = 0.004), and had a longer hospital stay (P < 0.001) compared with no PRE children. Causes of any hospital stay after surgery in outpatients were related to surgical condition (57%), anesthesia/PRE (31%), or patient’s condition (12%), for example, parental preference, difficult transportation, or private insurance. The cost variable, that is, hospital charge and excess cost of hospitalization in the PRE group, was significantly higher than in the no PRE group (P < 0.001), whereas indirect costs were not significantly different between PRE and no PRE groups (P = 0.23) (Table 3).
Analysis of Hospital Stay and Number of Days Hospitalized After Surgery
The following 8 variables having P ≤ 0.2 in the univariate analysis of any hospital stay after surgery (Supplemental Digital Content 1, http://links.lww.com/AA/B37) were included in the binomial part of the hurdle model but were not related to any hospital stay after surgery in the multivariate analysis: age, body mass index, snoring, choice of anesthesia, technique of anesthesia, intubating drug, gas mixed with oxygen, and narcotic. The following 4 variables having P ≤ 0.2 in the univariate analysis of number of days hospitalized after surgery (Supplemental Digital Content 1, http://links.lww.com/AA/B37) were included in the truncated negative binomial part of the hurdle model but were not related to number of days hospitalized after surgery in the multivariate analysis: ASA physical status, intubating drug, gas mixed with oxygen, and anesthetic time. Table 4 shows results of the multiple hurdle model predicting any hospital stay and number of days hospitalized after surgery. After adjusting for inpatients, ASA physical status, and anesthesia-related factors, PRE and severity of PRE were independent predictors for both any hospital stay (P = 0.01 and P = 0.024, respectively) and number of days hospitalized after surgery (P = 0.002 and P = 0.01, respectively). All pairwise interactions with PRE or severity of PRE for any hospital stay and length of stay after surgery were P ≥ 0.08 and P ≥ 0.36, respectively.
Analysis of Adjusted Excess Cost and Indirect Cost
The following 6 variables having P ≤ 0.2 in the univariate analysis (Supplemental Digital Content 2, http://links.lww.com/AA/B38) were included in the multiple mixed-effects linear regression but were not related to adjusted excess cost in the multivariate analysis: age, snoring, intubating drug, inhaled drug, gas mixed with oxygen, and narcotic. The following 7 variables having P ≤ 0.2 in the univariate analysis (Supplemental Digital Content 2, http://links.lww.com/AA/B38) were included in the multiple mixed-effects linear regression but were not related to indirect cost in the multivariate analysis: smoking, ASA physical status, technique of anesthesia, induction drug, intubating drug, narcotic, and anesthetic time. Table 5 shows results of the multiple mixed-effects linear regression analysis predicting log of adjusted excess hospital cost and log of indirect cost. After adjusting for inpatients, type of payment, and anesthesia-related factors, PRE, and severity of PRE were independent predictors for adjusted excess cost (P < 0.001 and P = 0.001, respectively) and for indirect costs (P = 0.001 and P = 0.002, respectively). All pairwise interactions with PRE or severity of PRE for adjusted excess cost were P ≥ 0.08, whereas PRE and severity of PRE showed significant effect modification with inpatients for indirect cost (P < 0.001).
This study examined excess hospital cost and indirect cost between children who had PRE and no PRE under a well-matching selection procedure with matching on outpatient/inpatient, type of surgery (similar surgical charge), ASA physical status, and date of surgery within a 6-month interval. Even though we allowed age differences in the matching procedure to be as high as 107 months, the median age difference between matched pairs in this study was only 13.5 months (P = 0.66). Overall, PRE had more impact on excess hospital costs (P < 0.001, i.e., accommodation, meals, medications, laboratory expenses, oxygen therapy, anesthesia charge, and nursing care service; Table 3) than on indirect costs (P = 0.23). The largest differences in charges were laboratory expenses, medical instrumentation, and hospital medications. We assumed that the more medication used, the more laboratory required and more material needed in the PRE group, whereas ventilator cost and cost of intensive care unit stay did not differ significantly between PRE and no PRE groups because of small numbers of severe PRE cases (22%).
Any Hospital Stay and Prolonged Number of Days Hospitalized After Surgery
Predictors for hospital stay and prolonged number of days hospitalized after surgery are shown in Table 4. The more severe the PREs, the higher the risk for any hospital stay and prolonged number of days hospitalized after surgery (P = 0.024 and P = 0.01, respectively). Most PRE patients had desaturation as a result of their condition. Even though most desaturation was not severe (75%), 39% (24/61) of outpatient children with PRE were hospitalized due to desaturation in combination with wheezing, upper airway obstruction, or respiratory depression.
ASA physical status III (versus ASA I) was an independent predictor for any hospital stay after surgery (P = 0.042). The Thai Anesthesia Incidents Study4 reported that high ASA physical status (III–V) was related to desaturation and cardiac arrest in pediatric anesthesia, which could have impact on both length of stay and cost. Use of a facemask compared with other techniques was a protective factor, that is, it decreased the number of days hospitalized after surgery (P = 0.001). Some studies reported that the use of a facemask or laryngeal mask airway compared with tracheal intubation significantly decreased the risk of respiratory complications in pediatric anesthesia,7,24,25 which might shorten the length of hospital stay. Anesthetic time 1 to 3 hours and >3 hours increased the odds of hospital stay after surgery 3-fold and 17-fold but was not a predictor for longer hospitalization stay after surgery.
Excess Hospital Cost and Indirect Cost
Severity of PRE was an independent predictor for 27% to 48% higher adjusted excess costs regardless of type of hospital payment (P = 0.001). The associations of PRE with higher excess hospital cost and with higher indirect cost are likely due to the longer hospitalization after surgery and the higher income loss of patients with PRE. Among inpatients, PRE was not associated with increased indirect cost (cost ratio, 2.89/2.86 ~ 1.01 [0.85–1.20]) presumably because income loss in inpatients may not differ among PRE and no PRE groups. ASA physical status and anesthesia-related factors were predictors for higher adjusted excess costs but not for indirect cost. Universal coverage had more impact on excess cost than other payment methods because the majority of excess hospital cost was paid by the government via universal coverage, whereas most of the indirect costs were not covered by universal coverage and were paid by the patients themselves.
ASA physical status and anesthesia-related factors had more impact on adjusted excess hospital cost than on indirect cost whereas nonanesthetic factors, that is, inpatients and young age (< 1 year), had more impact on indirect cost than on adjusted excess hospital costs (Table 5). Induction with propofol was associated with higher excess costs compared with sevoflurane in our study (P < 0.001), which is consistent with a study by Montes and Bohn.26
Prolonged admission among PRE children would reduce the accessibility for other patients to be admitted. Hospital policy to efficiently manage hospital beds for urgent patients is needed. Because relative weight for certain type of operation with PRE do not yet exist, universal coverage pays the hospital an average rate based on diagnosis-related group weighting per case assuming cases are no PRE. Thus, a loss of 30% to 48% reimbursement was found when PRE developed. The compensatory mechanisms to prevent hospital bankruptcy should be developed imperatively.
Strengths and Limitations
There are several strengths of our study. First, this prospective cohort study demonstrated excess hospital costs and indirect costs comparing children with and without PRE in noncardiac surgery, which has rarely been done. Second, children in the 2 groups were well matched on several characteristics to reduce confounding. Third, we performed matched analysis in both univariate part and multivariate part (mixed-effects model) for cost. A hurdle model with truncated negative binomial for hospital stay could not be done by matched analysis because in some of the matched pairs, 1 case may have had a zero count (discharged same day), whereas the other may not have, which would have resulted in both cases being excluded from the analysis. Last, adequate sample size in combination with appropriate study design provided high validity in our study. The external validity to other public health sectors should be credible according to high internal validity.
Even though we attempted to examine both excess hospital costs and indirect costs, there were some other resources used by patients or their families, (e.g. cost of accommodation of parents) or ongoing costs after discharge (e.g.cost of follow-up and cost of procedure related to PRE) which could not be obtained. Moreover, cost analysis regarding hospital total margin or contribution margin per patients27 could not be performed because revenues do not exist in the hospital information system and not all no PRE children were included in the analysis.
PRE in pediatric anesthesia for noncardiac surgery was associated with increased odds of hospital stay after surgery, 2 times prolonged hospitalization after surgery, 30% higher excess hospital cost overall, and 58% higher indirect cost among outpatients.
Name: Maliwan Oofuvong, MD.
Contribution: This author conducted and designed the study. She also helped collect data and analyze and prepare the manuscript.
Attestation: Maliwan Oofuvong attests to the integrity of the original data and the analysis reported in this manuscript and has approved the final manuscript. She is also the archival author.
Name: Alan Frederick Geater, PhD.
Contribution: This author helped design the study. He also helped analyze the data and approved the manuscript.
Attestation: Alan Frederick Geater attests to the integrity of the original data and the analysis reported in this manuscript and has approved the final manuscript.
Name: Virasakdi Chongsuvivatwong, MD, PhD.
Contribution: This author helped design the study. He also helped analyze data and approved the manuscript.
Attestation: Virasakdi Chongsuvivatwong attests to the analysis reported in this manuscript and has approved the final manuscript.
Name: Thavat Chanchayanon, MD.
Contribution: This author helped design the study and approved the manuscript.
Attestation: Thavat Chanchayanon attests to the analysis reported in this manuscript and has approved the final manuscript.
Name: Bussarin Sriyanaluk.
Contribution: This author helped collect the data and approved the manuscript.
Attestation: Bussarin Sriyanaluk attests to the integrity of the original data and has approved the final manuscript.
Name: Boonthida Saefung.
Contribution: This author helped collect the data and approved the manuscript.
Attestation: Boonthida Saefung has approved the final manuscript.
Name: Kanjana Nuanjun.
Contribution: This author helped collect the data and approved the manuscript.
Attestation: Kanjana Nuanjan has approved the final manuscript.
This manuscript was handled by: James A. DiNardo, MD.
We would like to thank the Faculty of Medicine, Prince of Songkla University, for cofunding this research project. This study forms part of the dissertation of the first author to fulfill the requirement for PhD in Epidemiology at Prince of Songkla University.
1. Oofuvong M, Geater AF, Chongsuvivatwong V, Chanchayanon T, Worachotekamjorn J, Sriyanaluk B, Saefung B, Nuanjun K. Comparison of intelligence, weight and height in children after general anesthesia with and without perioperative desaturation in non-cardiac surgery: a historical and concurrent follow-up study. SpringerPlus. 2014;3:164
2. Bhananker SM, Ramamoorthy C, Geiduschek JM, Posner KL, Domino KB, Haberkern CM, Campos JS, Morray JP. Anesthesia-related cardiac arrest in children: update from the Pediatric Perioperative Cardiac Arrest Registry. Anesth Analg. 2007;105:344–50
3. Morray JP. Cardiac arrest in anesthetized children: recent advances and challenges for the future. Paediatr Anaesth. 2011;21:722–9
4. Bunchungmongkol N, Somboonviboon W, Suraseranivongse S, Vasinanukorn M, Chau-in W, Hintong T. Pediatric anesthesia adverse events: the Thai Anesthesia Incidents Study (THAI Study) database of 25,098 cases. J Med Assoc Thai. 2007;90:2072–9
5. Mamie C, Habre W, Delhumeau C, Argiroffo CB, Morabia A. Incidence and risk factors of perioperative respiratory adverse events in children undergoing elective surgery. Paediatr Anaesth. 2004;14:218–24
6. Tait AR, Voepel-Lewis T, Burke C, Kostrzewa A, Lewis I. Incidence and risk factors for perioperative adverse respiratory events in children who are obese. Anesthesiology. 2008;108:375–80
7. von Ungern-Sternberg BS, Boda K, Chambers NA, Rebmann C, Johnson C, Sly PD, Habre W. Risk assessment for respiratory complications in paediatric anaesthesia: a prospective cohort study. Lancet. 2010;376:773–83
8. Edler AA, Mariano ER, Golianu B, Kuan C, Pentcheva K. An analysis of factors influencing postanesthesia recovery after pediatric ambulatory tonsillectomy and adenoidectomy. Anesth Analg. 2007;104:784–9
9. Nafiu OO, Burke CC, Chimbira WT, Ackwerh R, Reynolds PI, Malviya S. Prevalence of habitual snoring in children and occurrence of peri-operative adverse events. Eur J Anaesthesiol. 2011;28:340–5
10. Hintong T, Klanarong S, Suksompong S, Chua-in W, Chatmongkolchat S, Werawatganon T. The Thai Anesthesia Incident Monitoring Study (Thai AIMS) of oxygen desaturation in the post-anesthetic care unit. J Med Assoc Thai. 2008;91:1531–8
11. Stepanova M, Mishra A, Venkatesan C, Younossi ZM. In-hospital mortality and economic burden associated with hepatic encephalopathy in the United States from 2005 to 2009. Clin Gastroenterol Hepatol. 2012;10:1034–41.e1
12. Whitmore RG, Schwartz JS, Simmons S, Stein SC, Ghogawala Z. Performing a cost analysis in spine outcomes research: comparing ventral and dorsal approaches for cervical spondylotic myelopathy. Neurosurgery. 2012;70:860–7
13. Xue FS, Luo LK, Tong SY, Liao X, Deng XM, An G. Study of the safe threshold of apneic period in children during anesthesia induction. J Clin Anesth. 1996;8:568–74
14. Oofuvong M, Geater AF, Chongsuvivatwong V, Pattaravit N, Nuanjun K. Risk over time and risk factors of intraoperative respiratory events: a historical cohort study of 14,153 children. BMC Anesthesiol. 2014;14:13
15. Drummond MF, Sculpher MJ, Torrance GW, O’Brien BJ, Stodda GLDrummond MF, Sculpher MJ, Torrance GW, O’Brien BJ, Stoddart GL. Critical assessment of economic evaluation. Methods for the Economic Evaluation of Health Care Programmes. 20053rd ed. Oxford, UK Oxford University Press:1–373 In:
16. Dimai HP, Redlich K, Schneider H, Siebert U, Viernstein H, Mahlich J. [Direct and indirect costs of fractures due to osteoporosis in Austria.]. Gesundheitswesen. 2012;74:e90–8
17. Hamer HM, Spottke A, Aletsee C, Knake S, Reis J, Strzelczyk A, Oertel WH, Rosenow F, Dodel R. Direct and indirect costs of refractory epilepsy in a tertiary epilepsy center in Germany. Epilepsia. 2006;47:2165–72
18. Ivanova JI, Birnbaum HG, Kidolezi Y, Qiu Y, Mallett D, Caleo S. Direct and indirect costs associated with epileptic partial onset seizures among the privately insured in the United States. Epilepsia. 2010;51:838–44
19. Pato PA, Cebrian PE, Cimas H, I, Lorenzo Gonzalez JR, Rodriguez C, I, Gude SF. Analysis of direct, indirect, and intangible costs of epilepsy. Neurologia. 2011;26:32–8
20. Pike C, Birnbaum HG, Schiller M, Sharma H, Burge R, Edgell ET. Direct and indirect costs of non-vertebral fracture patients with osteoporosis in the US. Pharmacoeconomics. 2010;28:395–409
21. Rabenda V, Manette C, Lemmens R, Mariani AM, Struvay N, Reginster JY. The direct and indirect costs of the chronic management of osteoporosis: a prospective follow-up of 3440 active subjects. Osteoporos Int. 2006;17:1346–52
22. Hu MC, Pavlicova M, Nunes EV. Zero-inflated and hurdle models of count data with extra zeros: examples from an HIV-risk reduction intervention trial. Am J Drug Alcohol Abuse. 2011;37:367–75
23. Uakritdathikarn T, Chongsuvivatwong V, Geater AF, Vasinanukorn M, Thinchana S, Klayna S. Perioperative desaturation and risk factors in general anesthesia. J Med Assoc Thai. 2008;91:1020–9
24. Tait AR, Malviya S, Voepel-Lewis T, Munro HM, Seiwert M, Pandit UA. Risk factors for perioperative adverse respiratory events in children with upper respiratory tract infections. Anesthesiology. 2001;95:299–306
25. Tait AR, Pandit UA, Voepel-Lewis T, Munro HM, Malviya S. Use of the laryngeal mask airway in children with upper respiratory tract infections: a comparison with endotracheal intubation. Anesth Analg. 1998;86:706–11
26. Montes RG, Bohn RA. Deep sedation with inhaled sevoflurane for pediatric outpatient gastrointestinal endoscopy. J Pediatr Gastroenterol Nutr. 2000;31:41–6
27. Eappen S, Lane BH, Rosenberg B, Lipsitz SA, Sadoff D, Matheson D, Berry WR, Lester M, Gawande AA. Relationship between occurrence of surgical complications and hospital finances. JAMA. 2013;309:1599–606
Supplemental Digital Content
© 2015 International Anesthesia Research Society