This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Permissions Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Varughese, A. M. Articles by Kurth, C. D. Search for Related Content PubMed PubMed Citation Articles by Varughese, A. M. Articles by Kurth, C. D. Related Collections Anesthetic Techniques Pediatrics

---

PEDIATRIC ANESTHESIOLOGY

Factors Predictive of Poor Behavioral Compliance During Inhaled Induction in Children

Anna M. Varughese, MD, MPH^*, Todd G. Nick, PhD, Joel Gunter, MD^*, Yu Wang, MS, and C. Dean Kurth, MD^*

From the *Department of Anesthesiology; Department of Pediatrics, Cincinnati Children's Hospital Medical Center, University of Cincinnati College of Medicine; and Center for Epidemiology and Biostatistics, Cincinnati Children's Hospital Medical Center, Cincinnati, Ohio.

Address correspondence and reprint requests to Anna M. Varughese, MD, MPH, Department of Anesthesiology, Cincinnati Children's Hospital Medical Center, 3333 Burnet Avenue, Cincinnati, OH 45229. Address e-mail to anna.varughese{at}cchmc.org.

Abstract

BACKGROUND: Preoperative identification of children at risk of emotional distress and poor behavioral compliance during inhaled induction of anesthesia allows targeted interventions to reduce distress, thereby enhancing the quality of the anesthetic experience. We sought to identify patient, procedural, and health care system factors predictive of poor behavioral compliance during induction.

METHODS: We studied 861 developmentally appropriate children ages 1–13 yr, The American Society of Anesthesiologists physical status I to III, presenting for inhaled induction of anesthesia. All inductions were performed in an induction room with parent(s) present. Behavioral compliance was assessed using the Induction Compliance Checklist (ICC), an observational scale consisting of 10 behaviors scored as the number of behaviors observed during induction; ICC 4 was considered poor behavioral compliance. A multivariable ordinal logistic regression model for behavioral compliance was generated and the performance of the multivariable model was evaluated by the c statistic.

RESULTS: Twenty-one percent of children exhibited poor behavioral compliance on induction. Factors increasing the odds of poor behavioral compliance were younger age (<4 yr, P < 0.0001), shorter preoperative preparation time (P = 0.004), and high anxiety levels in the preoperative clinic (modified-Yale preoperative anxiety scale >40; P = 0.016). Previous anesthesia experience increased the odds in school-age children (P = 0.046); this effect was ameliorated in children attending the preoperative tour (P = 0.018). The model using these factors demonstrated moderate discrimination between children with poor compliance and those with perfect compliance (ICC = 0) (c statistic = 0.75).

CONCLUSIONS: Factors predictive of poor behavioral compliance were age, previous anesthesia, preoperative tour attendance, preoperative preparation time and anxiety levels in the preoperative clinic. These factors, bundled into a predictive algorithm, may help identify children who could benefit from behavioral or pharmacological interventions and avoid use of interventions to those at low risk.

Inhaled induction of anesthesia is widely used for healthy children undergoing outpatient medical and surgical procedures. Behavioral and physiological studies suggest that the induction of anesthesia is the most stressful phase of the entire perioperative experience for these children and their families.1–3 Poor behavioral compliance with inhaled anesthetic induction often portends adverse clinical outcomes, including emergence delirium, and maladaptive postoperative behaviors such as separation and general anxiety, eating difficulties, and sleep disturbances.4–8

In many institutions, pediatric anesthesiologists use behavioral (parental presence, preoperative programs) and pharmacologic (sedative premedication) interventions in an attempt to improve behavioral compliance during induction. However, these interventions often incur undesirable effects, including delayed discharge from the postanesthesia care unit and hospital,9 increased preoperative and postanesthesia care unit nursing staff requirements, delayed surgical schedules, and increased health care costs.

Preoperative identification of children at risk for poor behavior with induction would permit administration of these interventions to those children who would derive benefit while avoiding the use of these interventions to children who would not benefit but still be at risk for undesirable effects. If a predictive model for poor behavioral compliance during anesthesia induction could be developed, it could be used in the preoperative clinic to target interventions to children at high risk while avoiding their use in children at low risk.

The purpose of our study was to identify factors predicting poor behavioral compliance during induction of anesthesia and to use them to develop a predictive model applicable to the preoperative clinic. Model development included patient (e.g., age, gender), procedural (e.g., type of surgery), and health care system (e.g., number of hours fasted, sedative premedication) factors.

METHODS

Patients
After IRB approval and informed parental consent, we conducted a single-center prospective observational cohort study in 861 children from age 1–13 yr with normal neurocognitive development and ASA physical status I, II or III, undergoing outpatient or outpatient-admit surgery. Patients were selected by systematic random sampling (five subjects per day, evenly distributed from the operating room schedule) between June 25, 2003 and April 5, 2005.

Protocol
Patients were observed during the course of routine practice at our institution. A voluntary preoperative tour is offered to all patients and families a week in advance of surgery. Healthy children scheduled for outpatient and outpatient-admit procedures receive their preanesthetic evaluation on the day of surgery.10 Families arrive 90 min before surgery is scheduled to start. Preoperative fasting guidelines include cessation of clear liquids and solid food 4 and 8 hr before surgery, respectively. The decision to administer midazolam 0.3–0.5 mg/kg PO was left to the discretion of the anesthesiologist.

In keeping with our standard practice, induction of anesthesia was performed using oxygen, nitrous oxide and sevoflurane administered by mask. Anesthesia was induced in a small room adjacent to each operating room equipped with an anesthesia machine and monitor, with one or both parents present. This "induction room" is a non-sterile environment which permits the parent(s) to be present in street clothes.

Measures
The dependent variable (primary outcome measure) for this study was behavioral compliance during induction, assessed using the Induction Compliance Checklist (ICC).11 The ICC, an observational scale with good reliability (r = 0.978), contains 10 negative behavioral groupings:

• Crying, tears in eyes,
  
• Turns head away from mask,
  
• Verbal refusal, says "no,"
  
• Verbalization indicating fear or worry, "where's mommy?" or "will it hurt?"
  
• Pushes mask away with hands, pushes nurse or anesthesiologist with hands/feet,
  
• Covers mouth/nose with hands/arms or buries face,
  
• Hysterical crying, may scream,
  
• Kicks/flails legs/arms, arches back, and/or general struggling,
  
• Requires physical restraint,
  
• Complete passivity, either rigid or limp.
  

The ICC score represents the sum of the groupings checked as present during induction; high scores correlate with poor behavioral compliance.11 A perfect induction (no negative behaviors) is scored as 0; the highest possible score is 10. When the mask was introduced to the child's face, a research assistant, standing at the foot of the bed with an unobstructed view of the subject's face, scored the ICC. Pilot data on ICC scores collected simultaneously by two independent research assistants in 100 subjects at our institution confirmed excellent inter-rater reliability (intra-class correlation coefficient = 0.96, 95% CI 0.95–0.98).

For purposes of analysis, ICC scores were stratified into three categories: Perfect (ICC = 0), Moderate, and Poor. As the particularly maladaptive behaviors "hysterical crying," "kicking and flailing," and "requires physical restraint," were rare for ICC <4 and common for ICC 4 (Fig. 1), we selected an ICC 4 as the definition of "Poor" behavioral compliance. "Moderate" compliance was, by default, defined as ICC = 1–3 and represented subjects exhibiting milder negative behaviors.

View larger version (15K): [in this window] [in a new window]   Figure 1. Plot of the distribution of subjects with extremely maladaptive behaviors by Induction Compliance Checklist (ICC) score. For each ICC score, the black bar represents the number of subjects with that ICC score who exhibit none of behaviors 7, 8 or 9; the light gray bar represents the number of subjects who exhibit one of behaviors 7, 8 and 9, and the dark gray bar represents the number of subjects who exhibit two or more of behaviors 7, 8, and 9. The negative behaviors corresponding to the ICC item are listed in Table 3. ICC: Induction Compliance Checklist.

A variety of independent variables possibly predictive of poor behavioral compliance were recorded, including age, gender, race, ASA physical status, preoperative anxiety, previous anesthesia exposure, type of procedure, preoperative preparation time, number of hours fasted, participation in the preoperative tour, and use of midazolam premedication.

Preoperative anxiety was measured using the modified Yale Preoperative Anxiety Scale (m-YPAS).12 The m-YPAS, an observational instrument that can be completed in <1 min, contains 27 items in five categories (activity, emotional expressivity, state of arousal, vocalization, and use of parents). It has good to excellent inter- and intra-rater reliability (weighted = 0.73—0.91) and good validity (r = 0.64) for measuring anxiety in the preoperative area and on entrance to the induction room.12 Pilot data on m-YPAS scores collected by two independent research assistants in 100 subjects at our institution confirmed excellent inter-rater reliability for the m-YPAS in our environment (intra-class correlation coefficient = 0.94, 95% CI 0.91–0.96). The m-YPAS score was recorded by a research assistant when subjects first entered the preoperative examination room. Subjects with m-YPAS scores >40 were classified as anxious.13

Type of procedures was classified as either surgical (skin incision or excision of tissue) or non-surgical (e.g., endoscopy, magnetic resonance imaging, myringotomy and ventilating tubes). Preoperative preparation time was defined as the time from patient arrival in the preoperative clinic to patient transfer to the induction room. Number of hours fasted was defined as the time from the last ingestion of solids or liquids to start of induction.

Sample Size
The sample size of more than 800 subjects was initially based on the usual rule-of-thumb for sample size in multivariable logistic regression [10 events (or non-events, whichever is fewer) are required for each parameter].14 The parameters correspond to degrees of freedom (df) and include both the independent variables (predictors) and their higher order terms (interactions and nonlinear terms). In a pilot study with ICC >6 as the threshold for poor behavioral compliance, we found the incidence of poor compliance to be 10%, implying that 100 subjects would be required for each parameter in the model. Final sample size (n = 800+) was based on the ability to model at least eight parameters; the final decision on the number of parameters which could be reliably modeled was based on the observed number of subjects in each ICC category.

Statistical Analysis
Descriptive statistics were computed for demographic and clinical characteristics. Continuous variables are presented as median and interquartile range; categorical variables are presented as proportions.

A multivariable Proportional Odds Model (POM) was fitted to the categorized ICC score (Perfect, Moderate, or Poor) with covariates "age" (continuous, months), "race" (Caucasian versus non-Caucasian), "gender" (male versus female), "type of procedure" (surgical versus non-surgical), "ASA physical status" (I vs II or III), "m-YPAS" (40 vs >40), "preoperative tour" (Yes versus No), "previous anesthesia" (Yes versus No), "midazolam premedication" (Yes versus No), "hours fasted" (continuous, hours), and "preoperative preparation time" (continuous, minutes).

POM is an ordinal logistic regression model. It is a generalization of the Wilcoxon-Mann-Whitney rank-sum test and is the most popular method for ordinal outcomes.15 Applicability of the assumptions of the POM is tested by checking the consistency of the means of the predictors stratified by outcome levels.16

The number of interaction terms was limited to four based on the sample size considerations given above. Based on our intuition that previous anesthetic experience might be a significant modifier of other predictive factors, interaction terms to account for the interaction of previous anesthesia with age, preoperative tour, m-YPAS score and midazolam premedication were considered and tested for their significance in the model. In the interest of model simplicity, interaction terms with P > 0.20 were removed from the final model.

Restricted cubic splines17,18 were applied to the three continuous predictor variables ("age," "hours fasted" and "preoperative preparation time") to account for possible nonlinearity. In a model using splines, the continuous independent variable is divided into two or more domains separated by knots. Within each domain, the relationship between the independent variable(s) and the dependent (outcome) variable is defined by a different polynomial; knots represent the inflection points in the curve representing the regression model. As for the interaction terms above, nonlinear (spline) terms were removed from the final model for P > 0.20.

Besides the P value for each predictor, adjusted odds ratios (ORs) and their 95% Confidence Intervals (CI) were used to investigate the effect of a predictor changing from one level to another while controlling the other predictors. For continuous independent variables, the adjusted OR is reported for a specified interval. For dichotomized categorical independent variables, the adjusted odds ratio is reported for the two levels of the variable. The performance of the model was evaluated by the c statistic [numerically equivalent to the area under the receiver-operator characteristic curve].

All computations were performed using S-Plus software, version 7.0 for Windows. Missing values in the independent variables were imputed using the single imputation method implemented in the function "aregImpute" in the Hmisc Library^*; this method of imputation is based on predictive multivariable regression models using the available subject variables. The function "lrm" in the Design Library was used for the proportional odds model. Both the Hmisc Library and the Design 23, 2007. Library were developed by F.E. Harrell and are included in S-Plus 7.0.

RESULTS

Twenty-nine of the 861 subjects were excluded from analysis [enrollment errors (n = 2), incomplete data (n = 25) and change in induction technique from inhaled to IV (n = 2)], leaving 832 subjects in the final sample. Only 19 data values were missing from the complete data set of 832 subjects (6 preoperative preparation time, 4 hours fasted, 4 ASA physical status, 2 procedure type, 2 premedication, and 1 m-YPAS). Estimates for these 19 missing values were imputed as outlined above (Methods). Table 1 summarizes the patient, procedural and health care system characteristics of the study population. A majority of the study population were healthy (ASA I or II) Caucasians undergoing surgery without midazolam premedication.

The distribution of ICC scores in the study population appears in Table 2. A "Perfect" induction (ICC = 0) was observed in more than one-half of subjects (57%). For the remaining 43% of subjects with "Non-Perfect" inductions (ICC >0), Table 3 presents the frequencies of the various behavioral items by ICC score. Among subjects with ICC scores of 4, 72% to 100% exhibited the particularly maladaptive behaviors of "kicking/flailing/ struggling" (behavior #8), and "requires physical restraint" (behavior #9). "Hysterical crying" (behavior #7) was also observed more commonly in subjects with ICC 4 (13%–73%). Figure 1 presents the distribution of subjects with none, one, or two or more of these three negative behaviors by ICC score. Most subjects (88%) with ICC scores 4 exhibited 2 or more of these maladaptive behaviors, compared to only 10% of subjects with ICC = 3. Based on the association between an ICC score 4 and these negative behaviors, we defined "Poor" behavioral compliance as an ICC score 4, and "Moderate" behavioral compliance as an ICC score between 1 and 3. Moderate and Poor behavioral compliance occurred in 21% and 22% of subjects, respectively. Table 4 presents the distribution of candidate predictors by ICC category (Perfect, Moderate, or Poor).

Based on the considerations outlined above (see Methods), the 172 subjects in the smallest ICC category (Moderate) provided adequate sample size to model 17 df. The candidate interaction terms "previous anesthesia" and "midazolam premedication" and "previous anesthesia" and "m-YPAS" were excluded from consideration based on the criteria given above in Methods (P = 0.92 and 0.42 respectively). Similarly, nonlinear terms for "hours fasted" and "preoperative preparation time" were also excluded from consideration (P = 0.21 and 0.56 respectively). The final model thus contained the 11 predictors and two interactions listed in Table 5, representing 17 df [10 df from linear terms, 3 df from nonlinear terms ("age" spline function with four knots), 1 df from linear interaction terms, and 3 df from nonlinear interaction terms ("age" spline function with 4 knots)]. Including the two intercepts required by the POM with a three level outcome, there were thus 19 parameters to be estimated from the regression model. The data as modeled satisfied the criteria for applicability of the POM outlined above (see Methods).

The model demonstrated moderate discrimination between subjects (c statistic = 0.75; validated c statistic = 0.73). The model formulae to predict the probability of poor behavioral compliance and the probability of moderate or poor behavioral compliance ("Non-Perfect" behavioral compliance) are provided in the Appendix.

Using the model, Figure 2 illustrates the effect of age and preoperative preparation time on the probability of Non-Perfect Induction (ICC >0) and the probability of Poor Induction (ICC 4). The probability of Poor or Non-Perfect compliance increased markedly as age decreased below 4 yr. The probability of Poor or Non-Perfect compliance decreased gradually as preoperative preparation time increased.

View larger version (12K): [in this window] [in a new window]   Figure 2. Plot of predicted probabilities. (A) Probability of Non-Perfect behavioral compliance (ICC >0) versus "age"; (B) Probability of Non-Perfect behavioral compliance (ICC >0) versus "preoperative preparation time"; (C) Probability of Poor behavioral compliance (ICC 4) versus "age"; (D) Probability of Poor behavioral compliance (ICC 4) versus "preoperative preparation time." For each figure, the factors not plotted were set as follows: "age" = 4.42 yr (53 mo, sample median), "gender" = "female," "race" = "non-Caucasian," "preoperative preparation time" = 76 min (sample median), modified Yale Preoperative Anxiety Scale ("m-YPAS") 40 (non-anxious), "midazolam premedication" = "no," "ASA physical status" = "I," "hours fasted" = 11.5 h (sample median), "preoperative tour" = "no," "previous anesthesia" = "no," and "procedure type" = "non-surgical." ICC: Induction Compliance Checklist.

Table 5 presents the adjusted ORs, their 95% CI's and P values for each predictor included in the model. Statistically significant predictors of poor behavioral compliance were younger age (P < 0.0001), shorter preoperative preparation time (P = 0.004), and m-YPAS >40 (P = 0.016). "Previous anesthesia" was also a significant predictor of poor behavioral compliance, but was modified by both "age" (Table 5, Fig. 3) and "preoperative tour" (Table 5, Fig. 4). Previous anesthesia experience increased the odds of poor behavioral compliance in older, school-age subjects but not in toddlers [6-yr-old subject: OR = 2.78, 95% CI = (1.49, 5.20) vs 2 yr old subject: OR = 1.15, 95% CI = (0.74, 1.80)]. Odds of poor behavioral compliance decreased in subjects with previous anesthetic experience who attended the preoperative tour [OR = 0.26, 95% CI = (0.08, 0.78)]; the preoperative tour had no effect in subjects without prior anesthetic experience [OR = 1.26, 95% CI = (0.62, 2.56)].

View larger version (11K): [in this window] [in a new window]   Figure 3. Plot of predicted probabilities. (A) Probability of Non-Perfect behavioral compliance (ICC >0) versus "age," with and without previous anesthetic experience; (B) Probability of Poor behavioral compliance (ICC 4) versus "age," with and without previous anesthetic experience. The remaining factors in the model were set as follows: "gender" = "female," "race" = "non-Caucasian," "preoperative preparation time" = 76 min (sample median), "m-YPAS" 40 (non-anxious), "midazolam premedication" = "no," "ASA physical status" = "I," "hours fasted" = 11.5 h (sample median), "preoperative tour" = "no," and "procedure type" = "non-surgical."

View larger version (14K): [in this window] [in a new window]   Figure 4. (A) Plot of the probability of Non-Perfect behavioral compliance versus "previous anesthesia" and "preoperative tour." (B) Plot of the probability of poor behavioral compliance versus "previous anesthesia" and "preoperative tour." The remaining factors in the model were set as follows: "age" = 4.42 yr (53 mo, sample median), "gender" = "female," "race" = "non-Caucasian," "preoperative preparation time" = 76 min (sample median), "m-YPAS" 40 (non-anxious), "midazolam premedication" = "no," "ASA physical status" = "I," "hours fasted" = 11.5 h (sample median), and "procedure type" = "non-surgical."

DISCUSSION

Younger age (<4 yr), shorter preoperative preparation time, and preoperative anxiety (m-YPAS >40) were all associated with increased odds of Poor or Non-Perfect behavioral compliance with induction of anesthesia. Previous anesthesia increased the odds of poor behavioral compliance in school-age children but not in pre-schoolers; this effect was ameliorated in school-age children who attended the preoperative tour. Gender, race, type of surgical procedure, ASA physical status, midazolam premedication, and hours fasted were not predictive of Poor or Non-Perfect behavioral compliance during induction of anesthesia.

Our results are consistent with previous studies describing improved compliance with increasing age19–21 and with the stages of neurocognitive development. In the Piagetian taxonomy,22 toddlers and preschool children are in the sensorimotor and preoperational stage of cognitive development, respectively, in which explicit memory of previous events is restricted and understanding is limited to immediate physical interaction. As children enter school, they begin the concrete operational stage, where they are able to form logical mental models of reality based on previous experience and reason. Adolescents are in the logical operational stage, when thinking is increasingly flexible and understanding becomes more abstract and complete. Toddlers and pre-schoolers do not understand the purpose and benefits of anesthesia and may easily interpret induction as threatening. Compliance significantly improves in school-age children, as they are able to form mental models of surgery and perceive possible benefits from anesthesia. However, previous anesthesia may have a negative impact on their behavioral compliance because they are able to associate previous anesthetic induction with unpleasant experiences related to hospitalization, surgery, and anesthesia. The negative effect of previous anesthetic experience may decrease in older school-age children as unpleasant previous experiences are incorporated into a more complete understanding of the benefits of surgery and anesthesia.

These stages of cognitive development may explain why attending the preoperative tour had little impact on behavioral compliance except in school-age children with anesthetic experience. Younger children lack the cognitive ability to incorporate the information from the tour into their model of reality and receive little benefit from attendance. Children without a previous anesthetic do not have an internal model of the perioperative experience which can be fruitfully modified by the preoperative tour.

As noted in previous work,13 we found poor behavioral compliance during induction in children who were extremely anxious upon arrival to the preoperative clinic. Both child and parental anxiety are predictive of compliance during induction.13 We chose to consider only child anxiety because the m-YPAS, an observational tool to assess child anxiety, can be performed in less than a minute.

We expected that increased time spent waiting in the preoperative clinic before surgery would be associated with increased anxiety and poor compliance during induction. Instead, we found the converse: compliance improved as waiting time increased. One explanation for this finding is that faster preparation may increase anxiety because of sensory overload from rapid interactions with multiple providers, a feeling of being "out of control," and a lack of time to integrate information and receive reassurance.

We did not find that midazolam premedication decreased the odds of poor compliance with induction. Previous reports of the effect of midazolam premedication on behavioral compliance during induction have been inconsistent.23–26 There are many possible explanations for our results with regard to midazolam premedication. First, in our study, premedication was administered to children judged to be at risk of poor behavioral compliance during the preoperative interview. If this assessment was accurate, and if premedication was effective, then there would have been no difference in outcome for subjects receiving or not receiving premedication. If so, the assessment during the preoperative interview must have been based on considerations other than the factors found to be significant in our model, as the effect of each factor on premedication was corrected in the model. If, on the other hand, the assessment during the preoperative interview was not accurate in identifying at-risk subjects, then administration of premedication was, in essence, random and any possible benefit therefrom would have been received equally by subjects at-risk and not at-risk for poor behavioral compliance. Second, it is possible that premedication was effective in improving behavioral compliance, but only to a minor degree, and not sufficient to alter ICC category. If premedication improved ICC by only one or two points, then subjects receiving premedication might have had "better" inductions than otherwise expected, but still be within the same ICC category. Finally, premedication may have been ineffective because of inadequate dosing or poor timing of its administration.

Several aspects of our study design merit comment. Other studies have used an ICC >6 as the definition of a poor induction11; however, in our sample ICC scores >6 were quite rare (7%). The frequent occurrence of more than one of the particularly maladaptive behaviors "hysterical crying," "kicking and flailing," and "requires physical restraint" in subjects with ICC 4, and the virtual absence of their multiple occurrence in subjects with ICC <4, suggest implicit content validity for a definition of Poor behavioral compliance as an ICC 4.

Because of the assumptions inherent in the POM, comparisons are equally valid for the odds of Non-Perfect (ICC >0) versus Perfect behavioral compliance (ICC = 0) and the odds of Poor (ICC 4) versus Moderate or Perfect behavioral compliance (ICC <4). Thus, our predictive model works equally well for either goal for anesthetic induction: a perfect induction (absolutely no negative behaviors) or avoidance of a stormy induction (poor compliance).

The limitations of our study include its observational nature, our use of stand-alone induction rooms with parental presence, the absence of assessment of parental anxiety, and failure to include anesthesiologist experience and temperament. Our results may not be generalizable to anesthetic inductions conducted in operating rooms or without parents being present. It is entirely possible that additional psycho-behavioral dimensions of the parent/child/anesthesiologist unit might contribute as significant predictors of compliance during induction of anesthesia. However, our intention was to describe factors which could be used to rapidly and unobtrusively identify children who might benefit from interventions to improve behavioral compliance during induction. Assessment of parental anxiety is time consuming and requires the co-operation of the parent. Inclusion of anesthesiologist experience and temperament in the model would have been problematic because these factors would introduce the possibility of incorrect risk categorization if the assigned anesthesiologist were changed between assessment and induction.

For many minor procedures, the psychological impact of anesthesia induction may be greater than the physical insult from the procedure itself. Induction of anesthesia is one of the most psychologically challenging events in the perioperative period, and a stormy induction may predict postoperative maladaptive behaviors.

We are currently in the process of developing and validating a simple algorithm to identify children at risk of poor behavioral compliance during anesthesia induction based on the predictive model developed in this study; such an algorithm would permit targeted application of interventions intended to improve compliance, while avoiding unnecessary diversion of resources or administration of drugs to children at low risk.

APPENDIX

The formula used to predict the probability of Poor (versus Perfect or Moderate) behavioral compliance is given by

and the formula used to predict the probability of Moderate or Poor (versus Perfect) behavioral compliance is given by

where:

Xβ = 0.03132 gender + 0.1012 race + 0.2476 asa – 0.1603 procedure – 0.0345 npotime – 0.006507 prepare + 0.5579 mypas – 0.2182 premed + 1.3028 prevanes – 0.6705 age + 0.005359 (age – 1.1667)₊³ + 0.0008844 (age – 3.25)₊³ – 0.01153 (age – 5.5833)₊³ + 0.005283 (age – 10.4542)₊³ + 0.2289 pretour + prevanes [–0.5955 age + 0.05224 (age – 1.1667)₊³ – 0.1255 (age – 3.25)₊³ + 0.08606 (age – 5.5833)₊³ – 0.01277 (age – 10.4542)₊³] – 1.5893 prevanes*pretour and (a)₊ = a if a > 0 otherwise (a)₊ = 0.

The labels and values for the variable names are listed below

To account for the nonlinear effect of age, restricted cubic splines with four knots were applied. By default, four knots (1.1667, 3.25, 5.5833, 10.4542) were used, corresponding to the 5%, 35%, 65% and 95%-iles for age (in yr), respectively.

To use the above formulae, plug in the values of the corresponding predictor variables in the expression of Xβ, and select the appropriate formula to calculate the predicted probability of ICC 4 (vs ICC <4) or ICC >0 (vs ICC = 0).

Footnotes

*http:/biostat.mc.vanderbilt.edu/s/Hmisc, http://biostat.mc.vanderbilt.edu/s/Design, last accessed on July 23, 2007.

http:/biostat.mc.vanderbilt.edu/s/Design, http://biostat.mc.vanderbilt.edu/s/Design, last accessed on July 23, 2007.

Accepted for publication April 8, 2008.

Supported by Department of Anesthesiology and Center for Epidemiology and Biostatistics, Department of Pediatrics, Cincinnati Children's Hospital Medical Center, Cincinnati, OH 45229.

Presented at the American Society of Anesthesiologists Annual Meeting, October 2005.

REFERENCES

1. Kain ZN, Mayes LC, O'Connor TZ, Cicchetti DV. Preoperative anxiety in children. Predictors and outcomes.[see comment]. Arch Pediatr Adolesc Med 1996;150:1238–45[Abstract/Free Full Text]
2. Kain ZN, Mayes LC, Caramico LA, Silver D, Spieker M, Nygren MM, Anderson G, Rimar S. Parental presence during induction of anesthesia. A randomized controlled trial. Anesthesiology 1996;84:1060–7[Web of Science][Medline]
3. Schwartz BH, Albino JE, Tedesco LA. Effects of psychological preparation on children hospitalized for dental operations. J Pediatr 1983;102:634–8[Web of Science][Medline]
4. Kotiniemi LH, Ryhanen PT, Moilanen IK. Behavioural changes in children following day-case surgery: a 4-week follow-up of 551 children. Anaesthesia 1997;52:970–6[Web of Science][Medline]
5. Kain ZN, Wang SM, Mayes LC, Caramico LA, Hofstadter MB. Distress during the induction of anesthesia and postoperative behavioral outcomes. Anesth Analg 1999;88:1042–7[Abstract/Free Full Text]
6. Maranets I, Kain ZN. Preoperative anxiety and intraoperative anesthetic requirements. Anesth Analg 1999;89:1346–51[Abstract/Free Full Text]
7. Eckenhoff JE. Relationship of anesthesia to postoperative personality changes in children. AMA Am J Dis Child 1953;86:587–91[Abstract/Free Full Text]
8. Meyers EF, Muravchick S. Anesthesia induction techniques in pediatric patients: a controlled study of behavioral consequences. Anesth Analg 1977;56:538–42[Abstract/Free Full Text]
9. Viitanen H, Annila P, Viitanen M, Tarkkila P. Premedication with midazolam delays recovery after ambulatory sevoflurane anesthesia in children. [see comment]. Anesth Analg 1999;89:75–9[Abstract/Free Full Text]
10. Wittkugel EP, Varughese AM. Pediatric preoperative evaluation–a new paradigm. Int Anesthesiol Clin 2006;44:141–58[Medline]
11. Kain ZN, Mayes LC, Wang SM, Caramico LA, Hofstadter MB. Parental presence during induction of anesthesia versus sedative premedication: which intervention is more effective? Anesthesiology 1998;89:1147–56[Web of Science][Medline]
12. Kain ZN, Mayes LC, Cicchetti DV, Bagnall AL, Finley JD, Hofstadter MB. The Yale preoperative anxiety scale: how does it compare with a "gold standard"? Anesth Analg 1997;85:783–8[Abstract]
13. Kain ZN, Caldwell-Andrews AA, Maranets I, Nelson W, Mayes LC. Predicting which child-parent pair will benefit from parental presence during induction of anesthesia: a decision-making approach. Anesth Analg 2006;102:81–4[Abstract/Free Full Text]
14. Babyak MA. What you see may not be what you get: a brief, nontechnical introduction to overfitting in regression-type models. Psychosom Med 2004;66:411–21[Abstract/Free Full Text]
15. Bender R, Grouven U. Ordinal logistic regression in medical research. J R Coll Physicians Lond 1997;31:546–51[Web of Science][Medline]
16. Harrell FE Jr, Margolis PA, Gove S, Mason KE, Mulholland EK, Lehmann D, Muhe L, Gatchalian S, Eichenwald HF. Development of a clinical prediction model for an ordinal outcome: the World Health Organization multicentre study of clinical signs and etiological agents of pneumonia, sepsis and meningitis in young infants. WHO/ARI Young Infant Multicentre Study Group. Stat Med 1998;17:909–44[Web of Science][Medline]
17. Harrell FE Jr, Lee KL, Pollock BG. Regression models in clinical studies: determining relationships between predictors and response. J Natl Cancer Inst 1988;80:1198–202[Abstract/Free Full Text]
18. Riester KA, Peduzzi P, Holford TR, Ellison RT 3rd, Donta ST. Statistical evaluation of the role of Helicobacter pylori in stress gastritis: applications of splines and bootstrapping to the logistic model. J Clin Epidemiol 1997;50:1273–9[Web of Science][Medline]
19. Holm-Knudsen RJ, Carlin JB, McKenzie IM. Distress at induction of anaesthesia in children. A survey of incidence, associated factors and recovery characteristics. Paediatr Anaesth 1998;8:383–92[Web of Science][Medline]
20. Vetter TR. The epidemiology and selective identification of children at risk for preoperative anxiety reactions. Anesth Analg 1993;77:96–9[Abstract/Free Full Text]
21. Lumley MA, Melamed BG, Abeles LA. Predicting children's presurgical anxiety and subsequent behavior changes. J Pediatr Psychol 1993;18:481–97[Abstract/Free Full Text]
22. Dixon S. Theories, concepts, and cultural dimensions of development encounters with children: pediatric behavior and development. 3rd ed. St. Louis: Mosby, 2000:15–43
23. Morgan-Hughes JO, Bangham JA. Pre-induction behaviour of children. A review of placebo-controlled trials of sedatives. Anaesthesia 1990;45:427–35[Web of Science][Medline]
24. Beeby DG, Hughes JO. Behaviour of unsedated children in the anaesthetic room. Br J Anaesth 1980;52:279–81[Abstract/Free Full Text]
25. Mansson ME, Fredrikzon B, Rosberg B. Comparison of preparation and narcotic-sedative premedication in children undergoing surgery. Pediatr Nurs 1992;18:337–42[Medline]
26. Vetter TR. A comparison of midazolam, diazepam, and placebo as oral anesthetic premedicants in younger children. J Clin Anesth 1993;5:58–61[Web of Science][Medline]

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Permissions Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Varughese, A. M. Articles by Kurth, C. D. Search for Related Content PubMed PubMed Citation Articles by Varughese, A. M. Articles by Kurth, C. D. Related Collections Anesthetic Techniques Pediatrics