FOCUS ON AIRWAY MANAGEMENT

DEFINING THE ‘‘LEARNING CURVE’’ FOR PARAMEDIC STUDENTENDOTRACHEAL INTUBATION

Henry E. Wang, MD, MPH, Samuel R. Seitz, MEd, RN, NREMT-P,David Hostler, PhD, NREMT-P, Donald M. Yealy, MD

ABSTRACT

Background. Proficiency in endotracheal intubation (ETI) isassumed to improve primarily with accumulated experienceon live patients. While the National Standard Paramedic Cur-riculum recommends that paramedic students (PSs) performat least five live ETIs, these training opportunities are lim-ited. Objective. To evaluate the effects of cumulative liveETI experience, elapsed duration of training, and clinicalsetting on PS ETI proficiency. Methods. The authors usedlongitudinal, multicenter data from 60 paramedic trainingprograms over a two-year period. The PSs reported out-comes (success/failure) for all live ETIs attempted in theoperating room (OR), the emergency department (ED), theintensive care unit (ICU), and other hospital or prehospi-tal settings. Fixed-effects logistic regression was used tomodel up to 30 consecutive ETI efforts by each PS, ac-counting for per-PS clustering. For each patient, the au-thors evaluated the association between ETI success andthe PS’s cumulative number of ETIs, adjusted for clini-cal setting, elapsed number of days from the first ETI en-counter, and the interaction (cumulative ETIs!elapsed days).Predicted probability plots were constructed depicting the“learning curve” overall and for each clinical setting. Re-sults. Between one and 74 ETIs (median 7; IQR 4–12) were

Received December 3, 2004, from the Department of EmergencyMedicine, University of Pittsburgh School of Medicine (HEW, DPH,DMY), Pittsburgh, Pennsylvania; and the Emergency Medicine Pro-gram, School of Health and Rehabilitative Sciences (SRS), Uni-versity of Pittsburgh, Pittsburgh, Pennsylvania. Revisions receivedDecember 14, 2004, and December 15, 2004; accepted for publicationDecember 15, 2004.

Presented at the National Association of EMS Physicians annualmeeting, Naples, Florida, January 2005.

Dr. Wang is supported by Clinical Scientist Training Award K08-HS013628 from the Agency for Healthcare Research and Quality.

Address correspondence and reprint requests to: Henry E. Wang,MD, MPH, Department of Emergency Medicine, University ofPittsburgh, 230 McKee Place, Suite 400, Pittsburgh, PA 15213. e-mail:<wanghe@upmc.edu>.

doi:10.1080/10903120590924645

performed by each of 802 PSs. Of 7,635 ETIs, 6,464 (87.4%)were successful. Stratified by clinical setting, 6,311 (82.7%)ETIs were performed in the OR, 271 (3.6%) in the ED, 64(0.8%) in the ICU, 86 (1.1%) in other in-hospital settings,and 903 (11.8%) in the prehospital setting. For the 7,398ETIs included in the multivariate analysis, cumulative num-ber of ETI was associated with increased adjusted odds ofETI success (odds ratio 1.067 per ETI; 95% CI: 1.044–1.091).ETI learning curves were steepest for the ICU and pre-hospital settings but lower than for other clinical settings.Conclusions. Paramedic student ETI success improves withaccumulated live experience but appears to vary across dif-ferent clinical settings. Strategies for PS airway educationmust consider the volume of live ETIs as well as the clinicalsettings used for ETI training. Key words: intubation; intra-tracheal; emergency medical services; learning; allied healthpersonnel.

PREHOSPITAL EMERGENCY CARE 2005;9:156–162

The acquisition of procedural competence is a chal-lenge faced by students in all areas of health care. Dur-ing training, paramedic students (PSs) must acquireproficiency in endotracheal intubation (ETI), a difficultand challenging procedure.1 PSs may learn ETI using acombination of mannequin practice and clinical expe-rience on live patients. Live ETI experience is tradition-ally gained under the guidance of anesthesiologists un-der controlled conditions in the operating room (OR).However, many factors limit the opportunities for PSsto acquire this type of training. Opportunities for PSs toperform ETI in other in-hospital or prehospital settingsare also limited.

The National Standard Paramedic Curriculum rec-ommends that PSs successfully perform at least fiveETIs prior to graduation.1 However, empiric linksbetween PS ETI proficiency and accumulated “live”experience (i.e., the “learning curve”) do not exist. Inthis study, we sought to determine whether PS ETI suc-cess is associated with accumulated live ETI experience,adjusted for elapsed duration of training and clinicalsetting.

156

Preh

osp

Emer

g Ca

re D

ownl

oade

d fro

m in

form

ahea

lthca

re.c

om b

y 74

.69.

7.84

on

05/3

0/10

For p

erso

nal u

se o

nly.

Wang et al. LEARNING CURVE FOR PARAMEDIC ETI 157

METHODS

Study DesignThe University of Pittsburgh Institutional ReviewBoard (IRB) approved this secondary data set analy-sis. The original collection of PS data for research pur-poses was governed by a preexisting IRB approval fromInver Hills Community College, Inver Grove Heights,Minnesota.

Study Setting and PopulationData for this study were drawn from clinical perfor-mance data compiled by the FISDAP system (FISDAP,Inc., St. Paul, MN). FISDAP is a proprietary Internet-based system used to log PS clinical and proceduralexperience. PSs use the program to self-report informa-tion for each patient encounter, including the patient’schief complaint and demographic information, the re-sults of physical assessment, and the results of proce-dures attempted by the student. FISDAP is presentlyused by more than 175 paramedic training programsin North America. All FISDAP data are pooled on anational basis.

Study ProtocolFor this analysis, we included data on PS encountersinvolving ETI for the study period May 1, 1999, toDecember 31, 2003. We included only data for studentswho consented to data release for research purposes.FISDAP contains a system for site instructors to in-ternally audit PS entries. While there are no specifiedsystemwide methods for such auditing, site instructorsusually ensure that students’ Internet entries are con-sistent with those recorded on paper logs. We includedonly data that had been audited by site instructors.

Measurements or Key Outcome MeasuresParameters reported by PSs included the date of en-counter, the type of airway attempted, whether theprocedure was observed or performed by the PS, theclinical setting, the number of attempts, and the out-come of the airway placement effort (success or failure).Airway outcome was self-reported by the PS and waspresumed to be verified by the instructor or clinicalpreceptor. Types of airway included ETI, laryngealmask airway (LMA), Combitube, and cricothyroido-tomy. Clinical settings were broadly categorized as pre-hospital, OR, emergency department (ED), intensivecare unit (ICU), and other in-hospital.

We included only ETI cases; we excluded all non-ETIairway placements (LMA, Combitube, etc.). Because itis a specialized technique much different from and lessutilized than orotracheal ETI, we also excluded all na-sotracheal intubations.2 We excluded procedures per-

formed on mannequins or cadavers and where the PSonly observed the procedure. The definition of an ETIattempt (i.e., insertion of blade vs. insertion of tube)was not set a priori, and therefore this parameter wasnot used in the analysis.

The FISDAP system did not record data on many clin-ical covariates (for example, Glasgow Coma Score, levelof consciousness, or degree of relaxation) or anatomicfactors (for example, obesity, short neck). While basicphysiology and chief complaint information were avail-able for each patient in a separate database, we couldnot successfully link the airway and physiology datasets. Therefore, we could not further assess associationsbetween the ETI effort and patient physiologic param-eters. Student demographic data (age, sex, race, priorcertifications, etc.) were available but frequently dis-played missing values (more than 40%) and thus werenot analyzed.

Data AnalysisWe used fixed-effects logistic regression to model thehypothesized relationship.3"7 This approach is an ac-cepted approach for modeling learning curves and per-mitted us to adjust for multiple relevant covariateswhile accounting for per-PS clustering effects. Ordi-nary logistic regression is not appropriate where thedata are clustered; for example, in this data set, eachPS performed multiple ETIs and accounted for multi-ple observations.3 We did not use generalized estimat-ing equations (GEE) because this technique producesestimates that are population-averaged; we were inter-ested in subject-specific estimates.6 GEE is also inap-propriate when the cluster sizes are relatively large, asin this analysis. We did not use random-effects logisticregression because the regression estimates were sen-sitive to the number of quadrature points used in theapproximation and violated the Hausman specificationtest.8 We elected not to use cumulative sum techniques,which describe performance trends for individual sub-jects only.9,10

We modeled ETI success as the primary binary out-come. The key independent variable was cumulativeETIs, defined as the PS’s cumulative number of ETI en-counters (including the current patient encounter). Weadjusted for the covariate clinical setting (prehospital,OR, ED, ICU, or other in-hospital) because of the varia-tions in clinical and training conditions in each of theselocations.

We speculated that elapsed time, independent of thenumber of ETI encounters, may have positive effects(e.g., from additional didactic and clinical experience)or negative effects (e.g., skill decay due to the spo-radic nature of ETI opportunities). We also hypothe-sized that elapsed time may interact with the number ofETIs attempted. Therefore, we incorporated as covari-ates elapsed days since first ETI encounter (calculated

Preh

osp

Emer

g Ca

re D

ownl

oade

d fro

m in

form

ahea

lthca

re.c

om b

y 74

.69.

7.84

on

05/3

0/10

For p

erso

nal u

se o

nly.

158 PREHOSPITAL EMERGENCY CARE APRIL / JUNE 2005 VOLUME 9 / NUMBER 2

FIGURE 1. Number of endotracheal intubation (ETI) attempts by 802 paramedic students.

as the number of elapsed days from the first ETI en-counter to the current encounter) as well as the interac-tion (elapsed days ! cumulative ETIs).

Because only a small number of students (33, 4.1%)performed more than 30 ETIs, we modeled only thefirst 30 ETI encounters for each PS. The excluded casesaccounted for only a small fraction (237, 3.1%) of the to-tal ETIs. Because goodness-of-fit approaches for fixed-effects models are not well developed, as recommendedby prior authors we applied the Hosmer-Lemeshowtest on a model fitted with the same parameters butusing ordinary logistic regression.3 We also examined!D and !" statistics of the ordinary logistic regressionmodel.3 We considered higher-order parameterizationsof predictors in the model by using fractional polyno-mial regression (fracpoly command in Stata).11

To depict the “learning curve” (the graphical relation-ship between predicted ETI success and cumulative ETIencounters), we used parameter estimates from the re-gression analysis to calculate the predicted probabilityof ETI success for each successive ETI encounter. Weassumed that the fixed effect was zero. We constructedsimilar plots stratified by clinical setting.

All analyses were conducted using Stata v.8.2 (Stata-Corp LP, College Station, TX).

RESULTS

Of 2,063 students from 120 programs, 891 studentsfrom 60 programs consented to the release of datafor research purposes. Of these 891 students, 802

attempted a total of 7,635 ETIs. No ETIs were reportedby 89 students.

Paramedic student ETI characteristics are summa-rized in Figure 1 and Table 1. The PSs attempted be-tween one and 74 ETIs. Of 802 PSs, 556 achieved tenETI encounters, 175 achieved 20 ETI encounters, and71 achieved more than 20 ETI encounters. Elapsedtime from the first ETI encounter ranged from 0 to552 days. Pooled overall ETI success was approximately87.5% (95% confidence interval [CI]: 86.7–88.2%). ETIsuccess rates were highest for ED, OR, and other in-hospital settings, and lowest for ICU and prehospitalsettings.

TABLE 1. Characteristics of Paramedic Student EndotrachealIntubation (ETI) Efforts

Parameter Value

Number of ETI encounters per studentRange 1–74Mean 9.5Median (IQR) 7 (4–12)

Elapsed days since first ETI encounterRange 0–552Mean 9Median (IQR) 6 (3–12)

Intubation success, % (95% CI)Overall (n = 7, 635) 87.5% (86.7–88.2%)Operating room (n = 6, 311) 89.3% (88.5–90.0%)Emergency department (n = 271) 90.0% (86.4–93.6%)Intensive care unit (n = 64) 68.8% (57.1–80.4%)Other in-hospital (n = 86) 94.2% (89.1–99.2%)Prehospital (n = 903) 74.8% (71.9–77.6%)

IQR = interquartile range; CI = confidence interval.

Preh

osp

Emer

g Ca

re D

ownl

oade

d fro

m in

form

ahea

lthca

re.c

om b

y 74

.69.

7.84

on

05/3

0/10

For p

erso

nal u

se o

nly.

Wang et al. LEARNING CURVE FOR PARAMEDIC ETI 159

TABLE 2. Results of a Fixed-Effect Logistic Regression ModelDepicting the Association between Paramedic StudentEndotracheal Intubation (ETI) Success and Cumulative

ETI Experience

Parameter "

OddsRatio# 95% CI

Cumulative ETI experience 0.065 1.067 (1.044–1.091)(per patient)

Elapsed days since first 7.795 ! 10"5 1.000 (0.998–1.002)ETI encounter

Elapsed days ! cumulative 1.000 ! 10"6 1.000 (1.000–1.000)ETI (interaction)

Clinical SettingPrehospital† N/A 1.000 N/AOperating room 1.368 3.927 (3.017–5.113)Emergency department 0.967 2.638 (1.526–4.559)Intensive care unit "0.158 0.854 (0.335–2.175)Other in-hospital 2.006 7.434 (2.232–24.758)

#The odds ratio for cumulative ETI experience represents the incremental effectper ETI encounter (patient).†Prehospital setting is the baseline value for clinical setting.CI = confidence interval.

The multivariate model included 7,398 ETIs (Table 2).Adjusted for clinical setting, elapsed days from the firstto the current ETI encounter, and the interaction (cumu-lative ETIs ! elapsed days), cumulative ETI experiencewas independently associated with PS ETI success. Theodds ratio of 1.067 (95% CI: 1.044–1.091) represents theincremental effect of each additional ETI encounter onpredicted ETI success. Therefore, compared with thefirst ETI effort, the predicted odds of ETI success was1.914 (95% CI: 1.534–2.390) after ten ETI encounters,3.664 (2.352–5.710) after 20 ETI encounters, and 7.015

FIGURE 2. Relationship between endotracheal intubation (ETI) success and cumulative experience—all clinical settings pooled. Curve depictsthe predicted probability of successful ETI (solid line) and 95% confidence interval (dashed lines). Only the first 30 ETI efforts were modeled.

(3.607–13.644) after 30 ETI encounters. The main andinteraction effects of elapsed days were not statisticallysignificant.

Compared with the baseline linear model, frac-tional polynomial regression did not identify anyhigher-ordered models with superior fit. The Hosmer-Lemeshow test indicated acceptable model fit (p =0.70).

The “learning curve” for ETI success increased from77.8% to 95.8% over 30 ETI procedures (Figure 2). Learn-ing curves stratified by clinical setting are depicted inFigure 3. Predicted ETI success appeared to increasewith cumulative ETI experience across all clinical set-tings. The largest ETI success increases were observedfor the prehospital and ICU settings, but initial pre-dicted success rates for these settings were low (<60%).In contrast, while the learning curves of OR, ED, andother in-hospital settings began at higher levels (on theorder of 75–90%), improvements in predicted ETI suc-cess were lower.

DISCUSSION

We found that PS ETI success is associated with accu-mulated live ETI experience. This relationship existseven when adjusted for elapsed duration of trainingand the clinical setting of each ETI. Our analysis alsosuggests that elapsed time from the initiation of live ETItraining is not related to PS ETI success.

The concept of “learning” presumes that perfor-mance improves with repetition of a task.12 ETI is both

Preh

osp

Emer

g Ca

re D

ownl

oade

d fro

m in

form

ahea

lthca

re.c

om b

y 74

.69.

7.84

on

05/3

0/10

For p

erso

nal u

se o

nly.

160 PREHOSPITAL EMERGENCY CARE APRIL / JUNE 2005 VOLUME 9 / NUMBER 2

FIGURE 3. Relationship between endotracheal intubation (ETI) success and cumulative experience—stratified by clinical setting. The solid linerepresents the overall (pooled) predicted rate. Only the first 30 ETI efforts were modeled. OR = operating room; ED = emergency department;ICU = intensive care unit.

a key and complex procedure.13 While it is importantthat PSs achieve minimal proficiency in ETI prior tocertification, many barriers may prevent students fromattaining adequate live experience. Paramedics and PSshave only limited opportunities for ETI in the clinicalsetting.14 OR training is ideal but is difficult to obtainbecause of competition from other students (such asmedical students and residents) and medicolegal con-cerns. Even if a PS achieves “mastery” of ETI, once inclinical practice only a few select paramedics have op-portunities to return to the OR for skills maintenance.15

Even when presented with ETI opportunities, PSsmay be constrained by other factors that impedethe learning process. For example, PSs may not bepermitted to attempt ETI on patients who have dif-ficult airways. Thus, while PSs may learn basic ETItechnique, they may not acquire skills necessary formanaging more difficult airways, conditions likely tobe encountered in clinical prehospital practice. Clinicalconsiderations (for example, the deterioration of apatient) may also constrain the number of ETI attemptspermitted to the PS. In addition, the clinical techniquesused to manage the airway may also impact thelearning experience; for example, ETI conditions maybe very different depending on whether an instructorchooses to use neuromuscular-blocking agents or asedative agent only.

An interesting observation was that the PS ETI learn-ing curves differed when stratified by clinical setting.These trends may be explained by the different patient

populations, clinical techniques, and teaching condi-tions present in each setting. For example, in the ORsetting, most patients are clinically stable and deeplysedated or paralyzed prior to ETI. These patients maybe easier to intubate, and PSs may be afforded op-portunities for repeated attempts if initially unsuccess-ful. In the ED, rapid-sequence intubation (RSI—theuse of neuromuscular-blocking agents) is used widely,but this population may include “ill” patients withsome degree of instability.16 Therefore, while the PSmay face airway relaxation similar to the OR, he orshe may be allowed fewer attempts to achieve suc-cessful ETI. Likewise, the markedly lower ICU andprehospital curves may be explained by the nature ofthese patients and the limited array of pharmacologi-cally assisted ETI techniques used in these settings.17,18

Most importantly, PSs may have acquired very dif-ferent learning experiences given the heterogeneity ofinstructors and patients across these varied clinicalsettings.19

We did not have adequate information to determinewhy the “other in-hospital” curve rests higher than theother curves. However, only 20 PSs performed the 86ETIs in this subset, and thus this may simply reflect theskills of this student subset. We emphasize that this dataset contained information concerning neither the clini-cal condition of the intubated patient nor the techniquesand drugs used to facilitate ETI, and thus we could notcontrol for these factors, which may be relevant to PSETI success.20

Preh

osp

Emer

g Ca

re D

ownl

oade

d fro

m in

form

ahea

lthca

re.c

om b

y 74

.69.

7.84

on

05/3

0/10

For p

erso

nal u

se o

nly.

Wang et al. LEARNING CURVE FOR PARAMEDIC ETI 161

An interesting question is whether PSs should be ex-posed to a particular sequence of clinical settings for ETItraining. The slope of the learning curve depicts the de-gree of skill gained per additional ETI in each clinicalsetting. The learning curves for the OR and ED settingsappear to plateau after 20–25 ETIs. In contrast, the pre-hospital and ICU learning curves appear steeper anddo not plateau across the span of 30 ETIs. However, theinitial predicted ETI success rates for prehospital andICU settings are low (50–60%); these levels may repre-sent unacceptable risks to live patients. Therefore, PSsin this cohort may need to acquire baseline experiencewith 15–20 ETI encounters in the OR or ED prior to at-tempting more difficult ETIs in the prehospital or ICUsettings.

The thresholds for ETI “proficiency” have been de-fined differently for different medical specialties. Forexample, the anesthesia literature has recommendedthat anesthesia residents perform 20–57 ETIs to reach90% success.9,21"23 Nurse anesthetists are required toperform 200 ETIs, prior to graduation.24 Emergencymedicine residents are recommended to perform atleast 35 ETIs prior to graduation.25 The National Stan-dard Paramedic Curriculum recommends that PSs per-form at least five successful ETIs prior to graduation.1While this analysis was not expressly designed for thispurpose, our data suggest that PSs may require ex-posure to more than 15–25 live ETI encounters acrossa range of clinical settings to achieve success ratesabove 90%. This potential skill threshold may be af-fected by heterogeneity in the clinical settings usedfor training. We emphasize that this analysis depictshow ETI skill improves with cumulative experience—not the minimum thresholds that denote “ETI profi-ciency.” However, if this minimum experience figurewere adopted, only 10–15% of the students in this se-ries would have achieved this standard. We note that inthe current series 255 PSs (31.8%) did not attain the na-tional paramedic curriculum standard of five successfulETIs.

Of note, our analysis may depict the true learn-ing curve for novice intubators. Prior depictions ofETI learning curves have been based on anesthe-sia residents, many of whom may have obtained atleast minimal ETI experience during medical schoolstudies.9,21"23 Since PSs typically do not have priortraining in advanced life support techniques, our serieslikely represents a group with no prior ETI experience.

LIMITATIONS

We did not have data regarding clinical or anatomicfactors that may have affected the difficulty of the ETIeffort.20 As discussed previously, we could not success-fully link our ETI data to the data set containing basicphysiologic data. We similarly did not have access todata regarding program demography. We encountered

large numbers of missing values for PS demographyand therefore could not assess their potential relation-ships. We accepted these limitations since there was noother similarly large data set suitable for testing the hy-pothesized relationship. Prospective efforts to validateour findings must consider these factors.

All outcomes in this study were self-reported andmay have been subject to self-reporting bias. How-ever, most of the ETIs performed in this series occurredin controlled (i.e., OR) settings where ETI placementswere likely verified by an instructor or a preceptor.We note that the pooled ETI success rate in this co-hort (87.5%) appears similar to those reported by priorstudies of paramedic ETI.26,27

This analysis is based on a data set with a low subjectparticipation rate. Also, students who were more adeptintubators may have performed more ETIs and re-ported information more diligently. Despite these facts,we still had access to more than 7,600 ETIs performedby more than 800 PSs. A data set of this magnitudewould be extremely difficult to replicate on a prospec-tive basis. Furthermore, our intent with this data setwas strictly to model a hypothesized relationship be-tween ETI success and cumulative live ETI experience.We did not seek to describe the epidemiologic aspectsof PS ETI training.

This analysis assumes that acquisition of ETI skill re-sults primarily from experience on live patients. How-ever, students may acquire ETI skill using a varietyof other platforms such as mannequins, human simu-lators, and cadavers.28"33 Evaluation of the effects ofthese training modalities and their interactions withlive ETI experience was beyond the scope of this study.We found that elapsed time, and its interaction with ETIencounters, was not a significant predictor of ETI suc-cess. However, in this series most of the ETI encountersoccurred within a relatively short time frame. There-fore, we cannot make inferences regarding skill reten-tion from these data.

Finally, the learning curves in this analysis representpredicted trends in this population for a defined rangeof ETI experience. While this analysis accounts for aPS’s number of prior ETI encounters, it does not accountfor the student’s quality of performance on those encoun-ters. While we defined ETI proficiency in terms of ETIsuccess, ETI proficiency may be potentially describedin other dimensions. For example, avoidance of tech-nical errors (prolonged or multiple laryngoscopy, tubemisplacement or tube dislodgement, etc.) is importantduring ETI.34,35 Manner of laryngoscopy and appropri-ate decision making are other important dimensions ofETI, but measures for these constructs have not beendeveloped. These intermediate measures are likely im-portant but must be studied prospectively. We electedto defer modeling of these conceptual constructs be-cause of the preliminary nature of this analysis and therecognized limitations of the data set.

Preh

osp

Emer

g Ca

re D

ownl

oade

d fro

m in

form

ahea

lthca

re.c

om b

y 74

.69.

7.84

on

05/3

0/10

For p

erso

nal u

se o

nly.

162 PREHOSPITAL EMERGENCY CARE APRIL / JUNE 2005 VOLUME 9 / NUMBER 2

CONCLUSIONS

Paramedic student ETI success improves with accumu-lated live experience but appears to vary across differ-ent clinical settings. Strategies for PS airway educationmust consider the volume of live ETIs as well as theclinical settings used for ETI training.

The authors acknowledge David Page, MS, NREMT-P, Inver HillsCommunity College, St. Paul, Minnesota, and Michael Johnson, BA,FISDAP, Inc., St. Paul, Minnesota, for providing access to and as-sistance with the data set for this study. The authors also acknowl-edge Joseph P. Costantino, DrPH, Professor, Biostatistics, Universityof Pittsburgh, for his statistical assistance and guidance.

References1. National Highway Traffic Safety Administration. Emergency

Medical Technician Paramedic: National Standard Curricu-lum (EMT-P). Available at: http://www.nhtsa.dot.gov/people/injury/ems/EMT-P/. Accessed October 8, 2004.

2. Roppolo LP, Vilke GM, Chan TC, et al. Nasotracheal intubation inthe emergency department, revisited. J Emerg Med. 1999;17:791–9.

3. Hosmer DW, Lemeshow S. Applied Logistic Regression. Seconded. New York: John Wiley & Sons, 2000.

4. Kleinbaum DG, Klein M, Pryor ER. Logistic Regression: A self-learning Text. 2nd ed. New York: Springer, 2002.

5. Allison PD, SAS Institute. Logistic Regression Using the SASSystem: Theory and Application. Cary, NC: SAS Institute,1999.

6. Hardin JW, Hilbe J. Generalized Estimating Equations. BocaRaton, FL: Chapman & Hall/CRC, 2003.

7. Hsiao C. Analysis of Panel Data. 2nd ed. Cambridge; New York:Cambridge University Press, 2003.

8. Stata Corporation. Stata Cross-sectional Time-series: ReferenceManual. Release 8. College Station, TX: Stata Press, 2003.

9. de Oliveira Filho GR. The construction of learning curves for basicskills in anesthetic procedures: an application for the cumulativesum method. Anesth Analg. 2002;95:411–6.

10. Lim TO, Soraya A, Ding LM, Morad Z. Assessing doctors’ com-petence: application of CUSUM technique in monitoring doctors’performance. Int J Qual Health Care. 2002;14:251–8.

11. Stata Corporation. Stata Base Reference Manual. Release 8. Col-lege Station, TX: Stata Corporation, 2003.

12. Ramsay CR, Wallace SA, Garthwaite PH, Monk AF, Russell IT,Grant AM. Assessing the learning curve effect in health tech-nologies. Lessons from the nonclinical literature. Int J TechnolAssess Health Care. 2002;18(1):1–10.

13. Pollock MJ, Brown LH, Dunn KA. The perceived importance ofparamedic skills and the emphasis they receive during EMS ed-ucation programs. Prehosp Emerg Care. 1997;1:263–8.

14. Burton JH, Baumann MR, Maoz T, Bradshaw JR, Lebrun JE. Endo-tracheal intubation in a rural EMS state: procedure utilization andimpact of skills maintenance guidelines. Prehosp Emerg Care.2003;7:352–6.

15. Wayne MA, Friedland E. Prehospital use of succinylcholine: a20-year review. Prehosp Emerg Care. 1999;3:107–9.

16. Sakles JC, Laurin EG, Rantapaa AA, Panacek EA. Airway man-agement in the emergency department: a one-year study of 610tracheal intubations. Ann Emerg Med. 1998;31:325–32.

17. Karch SB, Lewis T, Young S, Hales D, Ho CH. Field intubation oftrauma patients: complications, indications, and outcomes. Am JEmerg Med. 1996;14:617–9.

18. Easley RB, Segeleon JE, Haun SE, Tobias JD. Prospective study ofairway management of children requiring endotracheal intuba-tion before admission to a pediatric intensive care unit. Crit CareMed. 2000;28:2058–63.

19. Dewey J. Experience and Education. 60th Anniversary Ed. WestLafayette, IN: Kappa Delta Pi, 1998.

20. Wang HE, Kupas DF, Paris PM, Bates RR, Yealy DM. Multivariatepredictors of failed out-of-hospital endotracheal intubation. AcadEmerg Med. 2003;10:717–24.

21. Konrad C, Schupfer G, Wietlisbach M, Gerber H. Learning man-ual skills in anesthesiology: is there a recommended number ofcases for anesthetic procedures? Anesth Analg. 1998;86:635–9.

22. Kopacz DJ, Neal JM, Pollock JE. The regional anesthesia “learningcurve.” What is the minimum number of epidural and spinalblocks to reach consistency? Reg Anesth. 1996;21:182–90.

23. Charuluxananan S, Kyokong O, Somboonviboon W,Pothimamaka S. Learning manual skills in spinal anesthesiaand orotracheal intubation: is there any recommended numberof cases for anesthesia residency training program? J Med AssocThai. 2001;84 (suppl 1):S251–S255.

24. Council on Accreditation of Nurse Anesthesia Educational Pro-grams. Standards for Accreditation of Nurse Anesthesia Educa-tional Programs. Park Ridge, IL: CANAEP, 2004.

25. Accreditation Council for Graduate Medical Education: Emer-gency Medicine—Guidelines for Procedures and Resuscitations.Available at: http://www.acgme.org/acWebsite/RRC 110/110guidelines.asp#res. Accessed November 12, 2004.

26. Wang HE, Kupas DF, Paris PM, Bates RR, Yealy DM. Preliminaryexperience with a prospective, multi-centered evaluation of out-of-hospital endotracheal intubation. Resuscitation. 2003;58:49–58.

27. Wang HE, O’Connor RE, Schnyder ME, Barnes TA, Megargel RE.Patient status and time to intubation in the assessment of prehos-pital intubation performance. Prehosp Emerg Care. 2001;5:10–8.

28. Hagberg CA, Greger J, Chelly JE, Saad-Eddin HE. Instructionof airway management skills during anesthesiology residencytraining. J Clin Anesth. 2003;15:149–53.

29. Stringer KR, Bajenov S, Yentis SM. Training in airway manage-ment. Anaesthesia. 2002;57:967–83.

30. Stewart RD, Paris PM, Pelton GH, Garretson D. Effect of var-ied training techniques on field endotracheal intubation successrates. Ann Emerg Med. 1984;13:1032–6.

31. Stratton SJ, Kane G, Gunter CS, et al. Prospective study ofmanikin-only versus manikin and human subject endotrachealintubation training of paramedics. Ann Emerg Med. 1991;20:1314–8.

32. Morhaim DK, Heller MB. The practice of teaching endotra-cheal intubation on recently decreased patients. J Emerg Med.1991;9:515–8.

33. Kovacs G, Bullock G, Ackroyd-Stolarz S, Cain E, Petrie D. A ran-domized controlled trial on the effect of educational interventionsin promoting airway management skill maintenance. Ann EmergMed. 2000;36:301–9.

34. Katz SH, Falk JL. Misplaced endotracheal tubes by paramedics inan urban emergency medical services system. Ann Emerg Med.2001;37:32–7.

35. Dunford JV, Davis DP, Ochs M, Doney M, Hoyt DB. Incidenceof transient hypoxia and pulse rate reactivity during paramedicrapid sequence intubation. Ann Emerg Med. 2003;42:721–8.

Preh

osp

Emer

g Ca

re D

ownl

oade

d fro

m in

form

ahea

lthca

re.c

om b

y 74

.69.

7.84

on

05/3

0/10

For p

erso

nal u

se o

nly.