Marjorie Charles‐Damte, RN

Setting and Subjects

This prospective cohort study was performed at Stroger Hospital of Cook County, a 500‐bed public teaching hospital in Chicago, IL, from March through May of 2007. The cohort was adult inpatients who were referred for SE on weekdays from 3 distinct patient care units (Figure 1). We used 2 sampling modes to balance practical constraints (short‐stay unit [SSU] patients were more localized and, therefore, easier to study) with clinical diversity. We consecutively sampled patients from our SSU, where adults with provisional cardiovascular diagnoses are admitted if they might be eligible for discharge with in 3 days.12 But we used random number tables with a daily unique starting point to randomly sample patients from the general medical wards and the coronary care unit (CCU). Patients were excluded if repositioning them for HCUE was potentially harmful. The study was approved by our hospital's institutional review board, and we obtained written informed consent from all enrolled patients.

Figure 1

SE Protocol

As part of enrolled patients' routine clinical care, SE images were acquired and interpreted in the usual fashion in our hospital's echocardiography laboratory, which performs SE on over 7,000 patients per year. Echocardiographic technicians acquired images with a General Electric Vivid 7 cardiac ultrasound machine (General Electric, Milwaukee, WI) equipped with a GE M4S 1.8 to 3.4 MHz cardiac transducer (General Electric). Technicians followed the standard adult transthoracic echocardiography scanning protocol to acquire 40 to 100 images on every patient using all available echocardiographic modalities: 2‐dimensional, M‐mode, color Doppler, continuous‐wave Doppler, pulse‐wave Doppler, and tissue Doppler.13 Blinded to HCUE results, attending physician cardiologist echocardiographers then interpreted archived images using computer software (Centricity System; General Electric) to generate final reports that were entered into patients' medical records. This software ensured that final reports were standardized, because echocardiographers' final qualitative assessments were limited to short lists of standard options; for example, in reporting left atrium (LA) size, echocardiographers chose from only 5 standard options: normal, mildly dilated, moderately dilated, severely dilated, and not interpretable. Investigators, who were also blinded to HCUE results, later abstracted SE results from these standardized report forms in patients' medical records. All echocardiographers fulfilled ASE training guidelines to independently interpret SE: a minimum of 150 personally‐performed and 300 personally‐interpreted echocardiographic examinations (training level 2).14

HCUE Training

Based on the recommendations of our cardiologist investigator (B.M.), we developed a training program for 1 hospitalist to become an HCUE instructor. Our instructor trainee (C.C.) was board‐eligible in internal medicine but had no previous formal training in cardiology or echocardiography. We a priori established that her training would continue until our cardiologist investigator determined that she was ready to train other hospitalists; this determination occurred after 5 weeks. She learned image acquisition by performing focused SE on 30 patients under the direct supervision of an echocardiographic technician. She also performed focused HCUE on 65 inpatients without direct supervision but with ongoing access to consult the technician to review archived images and troubleshoot difficulties with acquisition. She learned image interpretation by reading relevant chapters from a SE textbook15 and by participating in daily didactic sessions in which attending cardiologist echocardiographers train cardiology fellows in SE interpretation.

This hospitalist then served as the HCUE instructor for 8 other attending physician hospitalists who were board‐certified internists with no previous formal training in cardiology or echocardiography. The training program was limited to acquisition and interpretation of 2‐dimensional grayscale and color Doppler images for the 6 cardiac assessments under study (Table 1). The instructor marshaled pairs of hospitalists through the 3 components of the training program, which lasted a total of 27 hours.

First, hospitalists attended a 2‐hour lecture on the basic principles of HCUE. Slides from this lecture and additional images of normal and abnormal findings were provided to each hospitalist on a digital video disc. Second, each hospitalist underwent 20 hours of hands‐on training in 2‐hour sessions scheduled over 2 weeks. Willing inpatients from our hospital's emergency department were used as volunteers for these hand‐on training sessions. During these sessions the instructor provided practical suggestions to optimize image quality, such as transducer location and patient positioning. In the first 3 sessions, the minimum pace was 1 patient per hour; thereafter, the pace was increased to 1 patient per half‐hour. We chose 20 hours of hands‐on training and these minimum paces because they allowed each hospitalist to attain a cumulative experience of no less than 30 patientsan amount that heralds a flattening of the HCUE learning curve among medical trainees.9 Third, each pair of hospitalists received feedback from a cardiologist investigator (B.M.) who critiqued the quality and interpretation of images acquired by hospitalists during hands‐on training sessions. Since image quality varies by patient,16 hospitalists' images were compared side‐by‐side to images recorded by the instructor on the same patients. The cardiologist also critiqued hospitalists' interpretations of both their own images and additional sets of archived images from patients with abnormal findings.

HCUE Protocol

After completing the training program and blinded to the results of SE, the 8 hospitalists performed HCUE on enrolled patients within hours of SE. We limited the time interval between tests to minimize the effect that changes in physiologic variables, such as blood pressure and intravascular volume, have on the reliability of serial echocardiographic measurements.16 Hospitalists performed HCUE with a MicroMaxx 3.4 hand‐carried ultrasound machine equipped with a cardiology software package and a 1 to 5 MHz P17 cardiac transducer (Sonosite, Inc., Bothell, WA); simultaneous electrocardiographic recording, though available, was not used. While patients laid on their own standard hospital beds or on a standard hospital gurney in a room adjacent to the SE waiting room, hospitalists positioned them without assistance from nursing staff and recorded 7 best‐quality images per patient. Patients were first positioned in a partial (3045 degrees) left lateral decubitus position to record 4 grayscale images of the short‐axis and long‐axis parasternal and 2‐chamber and 4‐chamber apical views; 2 color Doppler images of the mitral inflow were also recorded from the long‐axis parasternal and the 4‐chamber apical views. Patients were then positioned supine to record 1 grayscale image of the inferior vena cava (IVC) from the transhepatic view. Hospitalists did not perform a history or physical exam on enrolled patients, nor did they review patients' medical records.

Immediately following the HCUE, hospitalists replayed the recorded images as often as needed and entered final interpretations on data collection forms. Linear measurements were made manually with a caliper held directly to the hand‐carried ultrasound monitor. These measurements were then translated into qualitative assessments based on standard values used by our hospital's echocardiographers (Table 2).17 When a hospitalist could not confidently assess a cardiac abnormality, the final HCUE assessment was recorded as indeterminate. Hospitalists also recorded the time to perform each HCUE, which included the time to record 7 best‐quality images, to interpret the findings, and to fill out the data collection form.

Data Analysis

We based our sample size calculations on earlier reports of HCUE by noncardiologist trainees for assessment of left ventricular (LV) systolic function.7, 10 From these reports, we estimated a negative likelihood ratio of 0.3. In addition, we expected about a quarter of our patients to have LV systolic dysfunction (B.M., personal communication). Therefore, to achieve 95% confidence intervals (CIs) around the point estimate of a negative likelihood ratio that excluded 0.50, our upper bound for a clinically meaningful result, we needed a sample size of approximately 300 patients.18

We defined threshold levels of ordinal severity for the 6 cardiac abnormalities under study based on their clinical pertinence to hospitalists (Table 2). Here, we reasoned that abnormalities at or above these levels would likely lead to important changes in hospitalists' management of inpatients; abnormalities below these levels rarely represent cardiac disease that is worthy of an immediate change in management. Since even mild degrees of LV dysfunction have important diagnostic and therapeutic implications for most general medicine inpatients, particularly those presenting with heart failure,19 we set our threshold for LV dysfunction at mild or greater. In contrast, since neither mild nor moderate mitral regurgitation (MR) has immediate implications for medical or surgical therapy even if symptoms or LV dysfunction are present,20 we set our threshold for MR at severe. Similarly, though mild LA enlargement21 and mild LV hypertrophy22 have clear prognostic implications for patients' chronic medical conditions, we reasoned that only moderate or severe versions likely reflect underlying abnormalities that affect hospitalists' point‐of‐care decision‐making. Since cardiac tamponade is rarely both subclinical23 and due to a small pericardial effusion,24 we set our threshold for pericardial effusion size at moderate or large. Finally, we set our threshold IVC diameter, a marker of central venous volume status,25 at dilated, because volume overload is an important consideration in hospitalized cardiac patients.

Using these thresholds, investigators dichotomized echocardiographers' SE readings as normal or abnormal for each of the 6 cardiac abnormalities under study to serve as the reference standards. Hospitalists' HCUE results were then compared to the reference standards in 2 different ways. We first analyzed HCUE results as dichotomous values to calculate conventional sensitivity, specificity, and positive and negative likelihood ratios. Here we considered indeterminate HCUE results positive in a clinically conservative tradeoff that neither ignores indeterminate results nor risks falsely classifying them as negative.26 We then analyzed hospitalists' HCUE results as ordinal values for receiver operating characteristic (ROC) curve analysis. Here we considered an indeterminate result as 1 possible test result.27

To examine interobserver variability of HCUE, we first chose from the 6 possible assessments only those with a mean number of abnormal patients per hospitalist greater than 5. We reasoned that variability among assessments with lower prevalence would be predictably wide and inconclusive. We then expressed variability as standard deviations (SDs) around mean sensitivity and specificity for the 8 hospitalists.

The CIs for likelihood ratios were constructed using the likelihood‐based approach to binomial proportions of Koopman.28 The areas under ROC curves were computed using the trapezoidal rule, and the CIs for these areas were constructed using the algorithm described by DeLong et al.29 All analyses were conducted with Stata Statistical Software, Release 10 (StataCorp, College Station, TX).

Setting and Subjects

This prospective cohort study was performed at Stroger Hospital of Cook County, a 500‐bed public teaching hospital in Chicago, IL, from March through May of 2007. The cohort was adult inpatients who were referred for SE on weekdays from 3 distinct patient care units (Figure 1). We used 2 sampling modes to balance practical constraints (short‐stay unit [SSU] patients were more localized and, therefore, easier to study) with clinical diversity. We consecutively sampled patients from our SSU, where adults with provisional cardiovascular diagnoses are admitted if they might be eligible for discharge with in 3 days.12 But we used random number tables with a daily unique starting point to randomly sample patients from the general medical wards and the coronary care unit (CCU). Patients were excluded if repositioning them for HCUE was potentially harmful. The study was approved by our hospital's institutional review board, and we obtained written informed consent from all enrolled patients.

Figure 1

SE Protocol

As part of enrolled patients' routine clinical care, SE images were acquired and interpreted in the usual fashion in our hospital's echocardiography laboratory, which performs SE on over 7,000 patients per year. Echocardiographic technicians acquired images with a General Electric Vivid 7 cardiac ultrasound machine (General Electric, Milwaukee, WI) equipped with a GE M4S 1.8 to 3.4 MHz cardiac transducer (General Electric). Technicians followed the standard adult transthoracic echocardiography scanning protocol to acquire 40 to 100 images on every patient using all available echocardiographic modalities: 2‐dimensional, M‐mode, color Doppler, continuous‐wave Doppler, pulse‐wave Doppler, and tissue Doppler.13 Blinded to HCUE results, attending physician cardiologist echocardiographers then interpreted archived images using computer software (Centricity System; General Electric) to generate final reports that were entered into patients' medical records. This software ensured that final reports were standardized, because echocardiographers' final qualitative assessments were limited to short lists of standard options; for example, in reporting left atrium (LA) size, echocardiographers chose from only 5 standard options: normal, mildly dilated, moderately dilated, severely dilated, and not interpretable. Investigators, who were also blinded to HCUE results, later abstracted SE results from these standardized report forms in patients' medical records. All echocardiographers fulfilled ASE training guidelines to independently interpret SE: a minimum of 150 personally‐performed and 300 personally‐interpreted echocardiographic examinations (training level 2).14

HCUE Training

Based on the recommendations of our cardiologist investigator (B.M.), we developed a training program for 1 hospitalist to become an HCUE instructor. Our instructor trainee (C.C.) was board‐eligible in internal medicine but had no previous formal training in cardiology or echocardiography. We a priori established that her training would continue until our cardiologist investigator determined that she was ready to train other hospitalists; this determination occurred after 5 weeks. She learned image acquisition by performing focused SE on 30 patients under the direct supervision of an echocardiographic technician. She also performed focused HCUE on 65 inpatients without direct supervision but with ongoing access to consult the technician to review archived images and troubleshoot difficulties with acquisition. She learned image interpretation by reading relevant chapters from a SE textbook15 and by participating in daily didactic sessions in which attending cardiologist echocardiographers train cardiology fellows in SE interpretation.

This hospitalist then served as the HCUE instructor for 8 other attending physician hospitalists who were board‐certified internists with no previous formal training in cardiology or echocardiography. The training program was limited to acquisition and interpretation of 2‐dimensional grayscale and color Doppler images for the 6 cardiac assessments under study (Table 1). The instructor marshaled pairs of hospitalists through the 3 components of the training program, which lasted a total of 27 hours.

First, hospitalists attended a 2‐hour lecture on the basic principles of HCUE. Slides from this lecture and additional images of normal and abnormal findings were provided to each hospitalist on a digital video disc. Second, each hospitalist underwent 20 hours of hands‐on training in 2‐hour sessions scheduled over 2 weeks. Willing inpatients from our hospital's emergency department were used as volunteers for these hand‐on training sessions. During these sessions the instructor provided practical suggestions to optimize image quality, such as transducer location and patient positioning. In the first 3 sessions, the minimum pace was 1 patient per hour; thereafter, the pace was increased to 1 patient per half‐hour. We chose 20 hours of hands‐on training and these minimum paces because they allowed each hospitalist to attain a cumulative experience of no less than 30 patientsan amount that heralds a flattening of the HCUE learning curve among medical trainees.9 Third, each pair of hospitalists received feedback from a cardiologist investigator (B.M.) who critiqued the quality and interpretation of images acquired by hospitalists during hands‐on training sessions. Since image quality varies by patient,16 hospitalists' images were compared side‐by‐side to images recorded by the instructor on the same patients. The cardiologist also critiqued hospitalists' interpretations of both their own images and additional sets of archived images from patients with abnormal findings.

HCUE Protocol

After completing the training program and blinded to the results of SE, the 8 hospitalists performed HCUE on enrolled patients within hours of SE. We limited the time interval between tests to minimize the effect that changes in physiologic variables, such as blood pressure and intravascular volume, have on the reliability of serial echocardiographic measurements.16 Hospitalists performed HCUE with a MicroMaxx 3.4 hand‐carried ultrasound machine equipped with a cardiology software package and a 1 to 5 MHz P17 cardiac transducer (Sonosite, Inc., Bothell, WA); simultaneous electrocardiographic recording, though available, was not used. While patients laid on their own standard hospital beds or on a standard hospital gurney in a room adjacent to the SE waiting room, hospitalists positioned them without assistance from nursing staff and recorded 7 best‐quality images per patient. Patients were first positioned in a partial (3045 degrees) left lateral decubitus position to record 4 grayscale images of the short‐axis and long‐axis parasternal and 2‐chamber and 4‐chamber apical views; 2 color Doppler images of the mitral inflow were also recorded from the long‐axis parasternal and the 4‐chamber apical views. Patients were then positioned supine to record 1 grayscale image of the inferior vena cava (IVC) from the transhepatic view. Hospitalists did not perform a history or physical exam on enrolled patients, nor did they review patients' medical records.

Immediately following the HCUE, hospitalists replayed the recorded images as often as needed and entered final interpretations on data collection forms. Linear measurements were made manually with a caliper held directly to the hand‐carried ultrasound monitor. These measurements were then translated into qualitative assessments based on standard values used by our hospital's echocardiographers (Table 2).17 When a hospitalist could not confidently assess a cardiac abnormality, the final HCUE assessment was recorded as indeterminate. Hospitalists also recorded the time to perform each HCUE, which included the time to record 7 best‐quality images, to interpret the findings, and to fill out the data collection form.

Data Analysis

We based our sample size calculations on earlier reports of HCUE by noncardiologist trainees for assessment of left ventricular (LV) systolic function.7, 10 From these reports, we estimated a negative likelihood ratio of 0.3. In addition, we expected about a quarter of our patients to have LV systolic dysfunction (B.M., personal communication). Therefore, to achieve 95% confidence intervals (CIs) around the point estimate of a negative likelihood ratio that excluded 0.50, our upper bound for a clinically meaningful result, we needed a sample size of approximately 300 patients.18

We defined threshold levels of ordinal severity for the 6 cardiac abnormalities under study based on their clinical pertinence to hospitalists (Table 2). Here, we reasoned that abnormalities at or above these levels would likely lead to important changes in hospitalists' management of inpatients; abnormalities below these levels rarely represent cardiac disease that is worthy of an immediate change in management. Since even mild degrees of LV dysfunction have important diagnostic and therapeutic implications for most general medicine inpatients, particularly those presenting with heart failure,19 we set our threshold for LV dysfunction at mild or greater. In contrast, since neither mild nor moderate mitral regurgitation (MR) has immediate implications for medical or surgical therapy even if symptoms or LV dysfunction are present,20 we set our threshold for MR at severe. Similarly, though mild LA enlargement21 and mild LV hypertrophy22 have clear prognostic implications for patients' chronic medical conditions, we reasoned that only moderate or severe versions likely reflect underlying abnormalities that affect hospitalists' point‐of‐care decision‐making. Since cardiac tamponade is rarely both subclinical23 and due to a small pericardial effusion,24 we set our threshold for pericardial effusion size at moderate or large. Finally, we set our threshold IVC diameter, a marker of central venous volume status,25 at dilated, because volume overload is an important consideration in hospitalized cardiac patients.

Using these thresholds, investigators dichotomized echocardiographers' SE readings as normal or abnormal for each of the 6 cardiac abnormalities under study to serve as the reference standards. Hospitalists' HCUE results were then compared to the reference standards in 2 different ways. We first analyzed HCUE results as dichotomous values to calculate conventional sensitivity, specificity, and positive and negative likelihood ratios. Here we considered indeterminate HCUE results positive in a clinically conservative tradeoff that neither ignores indeterminate results nor risks falsely classifying them as negative.26 We then analyzed hospitalists' HCUE results as ordinal values for receiver operating characteristic (ROC) curve analysis. Here we considered an indeterminate result as 1 possible test result.27

To examine interobserver variability of HCUE, we first chose from the 6 possible assessments only those with a mean number of abnormal patients per hospitalist greater than 5. We reasoned that variability among assessments with lower prevalence would be predictably wide and inconclusive. We then expressed variability as standard deviations (SDs) around mean sensitivity and specificity for the 8 hospitalists.

The CIs for likelihood ratios were constructed using the likelihood‐based approach to binomial proportions of Koopman.28 The areas under ROC curves were computed using the trapezoidal rule, and the CIs for these areas were constructed using the algorithm described by DeLong et al.29 All analyses were conducted with Stata Statistical Software, Release 10 (StataCorp, College Station, TX).

Setting and Subjects

This prospective cohort study was performed at Stroger Hospital of Cook County, a 500‐bed public teaching hospital in Chicago, IL, from March through May of 2007. The cohort was adult inpatients who were referred for SE on weekdays from 3 distinct patient care units (Figure 1). We used 2 sampling modes to balance practical constraints (short‐stay unit [SSU] patients were more localized and, therefore, easier to study) with clinical diversity. We consecutively sampled patients from our SSU, where adults with provisional cardiovascular diagnoses are admitted if they might be eligible for discharge with in 3 days.12 But we used random number tables with a daily unique starting point to randomly sample patients from the general medical wards and the coronary care unit (CCU). Patients were excluded if repositioning them for HCUE was potentially harmful. The study was approved by our hospital's institutional review board, and we obtained written informed consent from all enrolled patients.

Figure 1

SE Protocol

As part of enrolled patients' routine clinical care, SE images were acquired and interpreted in the usual fashion in our hospital's echocardiography laboratory, which performs SE on over 7,000 patients per year. Echocardiographic technicians acquired images with a General Electric Vivid 7 cardiac ultrasound machine (General Electric, Milwaukee, WI) equipped with a GE M4S 1.8 to 3.4 MHz cardiac transducer (General Electric). Technicians followed the standard adult transthoracic echocardiography scanning protocol to acquire 40 to 100 images on every patient using all available echocardiographic modalities: 2‐dimensional, M‐mode, color Doppler, continuous‐wave Doppler, pulse‐wave Doppler, and tissue Doppler.13 Blinded to HCUE results, attending physician cardiologist echocardiographers then interpreted archived images using computer software (Centricity System; General Electric) to generate final reports that were entered into patients' medical records. This software ensured that final reports were standardized, because echocardiographers' final qualitative assessments were limited to short lists of standard options; for example, in reporting left atrium (LA) size, echocardiographers chose from only 5 standard options: normal, mildly dilated, moderately dilated, severely dilated, and not interpretable. Investigators, who were also blinded to HCUE results, later abstracted SE results from these standardized report forms in patients' medical records. All echocardiographers fulfilled ASE training guidelines to independently interpret SE: a minimum of 150 personally‐performed and 300 personally‐interpreted echocardiographic examinations (training level 2).14

HCUE Training

Based on the recommendations of our cardiologist investigator (B.M.), we developed a training program for 1 hospitalist to become an HCUE instructor. Our instructor trainee (C.C.) was board‐eligible in internal medicine but had no previous formal training in cardiology or echocardiography. We a priori established that her training would continue until our cardiologist investigator determined that she was ready to train other hospitalists; this determination occurred after 5 weeks. She learned image acquisition by performing focused SE on 30 patients under the direct supervision of an echocardiographic technician. She also performed focused HCUE on 65 inpatients without direct supervision but with ongoing access to consult the technician to review archived images and troubleshoot difficulties with acquisition. She learned image interpretation by reading relevant chapters from a SE textbook15 and by participating in daily didactic sessions in which attending cardiologist echocardiographers train cardiology fellows in SE interpretation.

This hospitalist then served as the HCUE instructor for 8 other attending physician hospitalists who were board‐certified internists with no previous formal training in cardiology or echocardiography. The training program was limited to acquisition and interpretation of 2‐dimensional grayscale and color Doppler images for the 6 cardiac assessments under study (Table 1). The instructor marshaled pairs of hospitalists through the 3 components of the training program, which lasted a total of 27 hours.

First, hospitalists attended a 2‐hour lecture on the basic principles of HCUE. Slides from this lecture and additional images of normal and abnormal findings were provided to each hospitalist on a digital video disc. Second, each hospitalist underwent 20 hours of hands‐on training in 2‐hour sessions scheduled over 2 weeks. Willing inpatients from our hospital's emergency department were used as volunteers for these hand‐on training sessions. During these sessions the instructor provided practical suggestions to optimize image quality, such as transducer location and patient positioning. In the first 3 sessions, the minimum pace was 1 patient per hour; thereafter, the pace was increased to 1 patient per half‐hour. We chose 20 hours of hands‐on training and these minimum paces because they allowed each hospitalist to attain a cumulative experience of no less than 30 patientsan amount that heralds a flattening of the HCUE learning curve among medical trainees.9 Third, each pair of hospitalists received feedback from a cardiologist investigator (B.M.) who critiqued the quality and interpretation of images acquired by hospitalists during hands‐on training sessions. Since image quality varies by patient,16 hospitalists' images were compared side‐by‐side to images recorded by the instructor on the same patients. The cardiologist also critiqued hospitalists' interpretations of both their own images and additional sets of archived images from patients with abnormal findings.

HCUE Protocol

After completing the training program and blinded to the results of SE, the 8 hospitalists performed HCUE on enrolled patients within hours of SE. We limited the time interval between tests to minimize the effect that changes in physiologic variables, such as blood pressure and intravascular volume, have on the reliability of serial echocardiographic measurements.16 Hospitalists performed HCUE with a MicroMaxx 3.4 hand‐carried ultrasound machine equipped with a cardiology software package and a 1 to 5 MHz P17 cardiac transducer (Sonosite, Inc., Bothell, WA); simultaneous electrocardiographic recording, though available, was not used. While patients laid on their own standard hospital beds or on a standard hospital gurney in a room adjacent to the SE waiting room, hospitalists positioned them without assistance from nursing staff and recorded 7 best‐quality images per patient. Patients were first positioned in a partial (3045 degrees) left lateral decubitus position to record 4 grayscale images of the short‐axis and long‐axis parasternal and 2‐chamber and 4‐chamber apical views; 2 color Doppler images of the mitral inflow were also recorded from the long‐axis parasternal and the 4‐chamber apical views. Patients were then positioned supine to record 1 grayscale image of the inferior vena cava (IVC) from the transhepatic view. Hospitalists did not perform a history or physical exam on enrolled patients, nor did they review patients' medical records.

Immediately following the HCUE, hospitalists replayed the recorded images as often as needed and entered final interpretations on data collection forms. Linear measurements were made manually with a caliper held directly to the hand‐carried ultrasound monitor. These measurements were then translated into qualitative assessments based on standard values used by our hospital's echocardiographers (Table 2).17 When a hospitalist could not confidently assess a cardiac abnormality, the final HCUE assessment was recorded as indeterminate. Hospitalists also recorded the time to perform each HCUE, which included the time to record 7 best‐quality images, to interpret the findings, and to fill out the data collection form.

Data Analysis

We based our sample size calculations on earlier reports of HCUE by noncardiologist trainees for assessment of left ventricular (LV) systolic function.7, 10 From these reports, we estimated a negative likelihood ratio of 0.3. In addition, we expected about a quarter of our patients to have LV systolic dysfunction (B.M., personal communication). Therefore, to achieve 95% confidence intervals (CIs) around the point estimate of a negative likelihood ratio that excluded 0.50, our upper bound for a clinically meaningful result, we needed a sample size of approximately 300 patients.18

We defined threshold levels of ordinal severity for the 6 cardiac abnormalities under study based on their clinical pertinence to hospitalists (Table 2). Here, we reasoned that abnormalities at or above these levels would likely lead to important changes in hospitalists' management of inpatients; abnormalities below these levels rarely represent cardiac disease that is worthy of an immediate change in management. Since even mild degrees of LV dysfunction have important diagnostic and therapeutic implications for most general medicine inpatients, particularly those presenting with heart failure,19 we set our threshold for LV dysfunction at mild or greater. In contrast, since neither mild nor moderate mitral regurgitation (MR) has immediate implications for medical or surgical therapy even if symptoms or LV dysfunction are present,20 we set our threshold for MR at severe. Similarly, though mild LA enlargement21 and mild LV hypertrophy22 have clear prognostic implications for patients' chronic medical conditions, we reasoned that only moderate or severe versions likely reflect underlying abnormalities that affect hospitalists' point‐of‐care decision‐making. Since cardiac tamponade is rarely both subclinical23 and due to a small pericardial effusion,24 we set our threshold for pericardial effusion size at moderate or large. Finally, we set our threshold IVC diameter, a marker of central venous volume status,25 at dilated, because volume overload is an important consideration in hospitalized cardiac patients.

Using these thresholds, investigators dichotomized echocardiographers' SE readings as normal or abnormal for each of the 6 cardiac abnormalities under study to serve as the reference standards. Hospitalists' HCUE results were then compared to the reference standards in 2 different ways. We first analyzed HCUE results as dichotomous values to calculate conventional sensitivity, specificity, and positive and negative likelihood ratios. Here we considered indeterminate HCUE results positive in a clinically conservative tradeoff that neither ignores indeterminate results nor risks falsely classifying them as negative.26 We then analyzed hospitalists' HCUE results as ordinal values for receiver operating characteristic (ROC) curve analysis. Here we considered an indeterminate result as 1 possible test result.27

To examine interobserver variability of HCUE, we first chose from the 6 possible assessments only those with a mean number of abnormal patients per hospitalist greater than 5. We reasoned that variability among assessments with lower prevalence would be predictably wide and inconclusive. We then expressed variability as standard deviations (SDs) around mean sensitivity and specificity for the 8 hospitalists.

The CIs for likelihood ratios were constructed using the likelihood‐based approach to binomial proportions of Koopman.28 The areas under ROC curves were computed using the trapezoidal rule, and the CIs for these areas were constructed using the algorithm described by DeLong et al.29 All analyses were conducted with Stata Statistical Software, Release 10 (StataCorp, College Station, TX).

Setting and Study Design

We conducted a prospective, firm‐based clinical trial approved by the institutional review board at Cook County Hospital, a 464‐bed public teaching hospital. We evaluated all central venous catheters inserted by residents on the general medicine service from November 15, 2004, to March 31, 2005. The internal medicine residency program assigns residents to 1 of 3 firms for their entire 3 years of training. We designated 1 firm as the intervention group; the other 2 firms constituted the control group.

Educational Intervention

At the beginning of each 4‐week general medicine ward rotation, intervention‐firm residents attended an educational and simulation laboratory session. Control‐firm residents received the usual ward orientation. We conducted 6 sessions, with total attendance of 40 intervention‐firm residents, or approximately 7 residents per session. A chief medical resident experienced in catheter placement and an attending internist led and supervised each 2‐hour training session. The sessions were conducted at the Simulation Laboratory of Rush University and included a presentation about indications for central venous catheter insertion, insertion techniques, common complications, and practice placing catheters in mannequins. During the hands‐on session, each participant observed the expert insert a triple‐lumen catheter in the mannequin's internal jugular and subclavian veins. Then, with supervision, each participant practiced catheter insertion using recommended infection‐control practices (eg, use of gloves, mask, and large drape, and chlorhexidine skin preparation).

Resident Survey

Before each session, we administered a survey that assessed residents' knowledge of insertion techniques and their confidence in placing catheters at each venous insertion site. To measure change in the confidence level of residents, we distributed an abbreviated survey 2 additional times, immediately after the session and at the end of the study period. We measured confidence with answers to survey questions, which were rated on a 5‐point Likert scale, from strongly disagree to strongly agree. In addition to measuring the change in residents' confidence, the final survey repeated knowledge assessment questions, evaluated residents' attitudes regarding venous insertion sites, and asked about potential strategies to improve insertion practices.

Central Venous Catheter Detection and Monitoring

At the end of each day, residents reported catheter insertions to chief residents during routine sign‐out rounds. If a catheter had been inserted, the chief resident interviewed the resident about type of catheter, venous insertion site, duration of attempt, patient location, immediate complications, number of inserters, inserter attendance at an educational session, inserter specialty, and professional designation (eg, resident, fellow, attending), indication for insertion, and adherence to infection‐control practices. For all insertion attempts, the research team reviewed the medical record and recorded patient characteristics that might influence venous insertion site (eg, thrombocytopenia, coagulopathy, and body mass index) and evaluated patients for insertion‐related complications.

We prospectively monitored patients for mechanical (ie, pneumothorax or hematoma), thromboembolic, or infectious complications. To evaluate for pneumothorax, postinsertion chest radiographs were reviewed by a physician‐investigator, and radiologists' interpretations and progress notes were reviewed. To evaluate for infectious or other mechanical complications, progress notes also were reviewed. We required radiographic confirmation of venous thromboembolism. To categorize potential bloodstream infections, we used Centers for Disease Control and Prevention definitions.9 All medical record and radiograph reviews were performed by investigators who were masked to patient firm assignment. We monitored patients until catheter removal or hospital discharge. After patient discharge, we reviewed the electronic record, including emergency room visits and repeat hospitalizations, for 30 days after the earlier of hospital discharge or catheter removal.

Statistics

Because we were aware that temporary dialysis catheters are sometimes placed in femoral veins to preserve the subclavian or internal jugular venous sites for more permanent tunneled intravascular catheters, our prespecified plans were to compare venous insertion sites between intervention and control groups after excluding temporary dialysis catheters. To more completely describe catheter use, we also collected data on temporary dialysis catheters, and we present the results both with and without inclusion of data on temporary dialysis catheters. If multiple residents attempted to insert a catheter, we would have used the group that the final inserter was in to determine intervention versus control group assignment; however, this never occurred.

To determine resident confidence in inserting catheters, we collapsed the responses of agree and strongly agree and of disagree and strongly disagree into single categories; thus, frequency of agreement was evaluated as a dichotomous outcome. To test whether residents' confidence changed between the 3 surveys, we analyzed responses using the matched‐pair signed rank test, with the initial survey used as the referent.

We dichotomized certain continuous variables using the following cut points: body mass index 30 kg/m²; coagulopathy, international normalized ratio (INR) > 1.5; thrombocytopenia, platelets < 100 10⁹/L. Data were entered into a relational database (Microsoft Access, Microsoft Inc., Redmond, WA) and merged analyzed using Stata software, version 8.2 (Stata Corporation, College Station, TX).

Patient and Catheter Characteristics

Fifty‐four catheters were inserted in 48 patients during the study period, 16 (30%) in the intervention group and 38 (70%) in the control group. Mean number of catheters inserted per resident for each 4‐week rotation was 0.24; therefore, on average, a resident would insert 1 catheter every 4 general‐medicine rotations. Most catheters were inserted between 7:00 _AM and 5:00 _PM; the most common reason for insertion was to administer intravenous medications to a patient without intravenous access, followed by the need for a temporary dialysis catheter. Most catheters were inserted by the medicine team rather than radiology or a subspecialty service (Table 1). Most patient characteristics and reasons for insertion were similar between groups; however, more patients in the control group had thrombocytopenia (Table 1).

Insertion Practices

Femoral venous insertion was the most common type of catheter insertion (67%), followed by internal jugular (26%) and subclavian (7%); there were no differences in insertion site between the intervention and control groups (Table 2). When we excluded temporary dialysis catheters, 39% of central venous catheters were inserted in the internal jugular vein. Although a smaller proportion of catheters inserted by the intervention group were placed in a femoral vein, the difference was not significant (Table 2).

For most insertions, residents reported using sterile gloves (94%) and a large drape (80%); however, most did not report use of a sterile gown (48%), mask (46%), or cap (15%). Residents in the intervention group were more likely to report use of a mask, and there was a trend toward increased use of large drapes (Table 2). No patient characteristics predicted femoral venous insertion (data not shown).

Complications

The most frequent complication was arterial puncture (n = 4); all four occurred during femoral venous insertion attempts. Compared to subclavian or internal jugular venous placement, there was a trend toward more mechanical complications among femoral catheters (Table 3). One episode of clinical sepsis occurred, in an intervention‐group patient who had femoral and internal jugular catheters, and no pneumothoraxes or episodes of venous thromboembolism occurred (Table 3). The overall incidence of bloodstream infection was 2.7 per 1000 central‐line days; there was no difference between the intervention and control groups (9.2 versus 0 per 1000 central‐line days; P = .29).

Survey Responses

Before the educational session, many residents did not recognize that femoral venous catheter insertions had a higher risk of arterial puncture or venous thrombosis (Table 4); by the final survey, residents were more likely to recognize the higher risk of these complications during femoral venous insertions. Most residents recognized the higher risk of infectious complications at the femoral site (Table 4).

Residents overwhelmingly responded that the lecture was useful (95%), that mannequins provided a valuable skill‐building exercise (90%), and that the session should be incorporated into the training program (95%). Immediately after the session, residents had increased confidence about inserting a central venous catheter at any venous site, especially for internal jugular or subclavian insertions. By the final survey, the confidence of residents about inserting catheters in the internal jugular or subclavian veins had returned to baseline but had increased for femoral‐site insertions (Table 4).

Most residents in the intervention group agreed that the educational session motivated them to remove unnecessary catheters, improve insertion‐related infection‐control practices, and place the catheter in an internal jugular or subclavian vein; some agreed because of the educational session, they were less likely to place a central venous catheter. Some reported successfully inserting a central venous catheter in the subclavian or internal jugular vein for the first time (Table 4).

Setting and Study Design

We conducted a prospective, firm‐based clinical trial approved by the institutional review board at Cook County Hospital, a 464‐bed public teaching hospital. We evaluated all central venous catheters inserted by residents on the general medicine service from November 15, 2004, to March 31, 2005. The internal medicine residency program assigns residents to 1 of 3 firms for their entire 3 years of training. We designated 1 firm as the intervention group; the other 2 firms constituted the control group.

Educational Intervention

At the beginning of each 4‐week general medicine ward rotation, intervention‐firm residents attended an educational and simulation laboratory session. Control‐firm residents received the usual ward orientation. We conducted 6 sessions, with total attendance of 40 intervention‐firm residents, or approximately 7 residents per session. A chief medical resident experienced in catheter placement and an attending internist led and supervised each 2‐hour training session. The sessions were conducted at the Simulation Laboratory of Rush University and included a presentation about indications for central venous catheter insertion, insertion techniques, common complications, and practice placing catheters in mannequins. During the hands‐on session, each participant observed the expert insert a triple‐lumen catheter in the mannequin's internal jugular and subclavian veins. Then, with supervision, each participant practiced catheter insertion using recommended infection‐control practices (eg, use of gloves, mask, and large drape, and chlorhexidine skin preparation).

Resident Survey

Before each session, we administered a survey that assessed residents' knowledge of insertion techniques and their confidence in placing catheters at each venous insertion site. To measure change in the confidence level of residents, we distributed an abbreviated survey 2 additional times, immediately after the session and at the end of the study period. We measured confidence with answers to survey questions, which were rated on a 5‐point Likert scale, from strongly disagree to strongly agree. In addition to measuring the change in residents' confidence, the final survey repeated knowledge assessment questions, evaluated residents' attitudes regarding venous insertion sites, and asked about potential strategies to improve insertion practices.

Central Venous Catheter Detection and Monitoring

At the end of each day, residents reported catheter insertions to chief residents during routine sign‐out rounds. If a catheter had been inserted, the chief resident interviewed the resident about type of catheter, venous insertion site, duration of attempt, patient location, immediate complications, number of inserters, inserter attendance at an educational session, inserter specialty, and professional designation (eg, resident, fellow, attending), indication for insertion, and adherence to infection‐control practices. For all insertion attempts, the research team reviewed the medical record and recorded patient characteristics that might influence venous insertion site (eg, thrombocytopenia, coagulopathy, and body mass index) and evaluated patients for insertion‐related complications.

We prospectively monitored patients for mechanical (ie, pneumothorax or hematoma), thromboembolic, or infectious complications. To evaluate for pneumothorax, postinsertion chest radiographs were reviewed by a physician‐investigator, and radiologists' interpretations and progress notes were reviewed. To evaluate for infectious or other mechanical complications, progress notes also were reviewed. We required radiographic confirmation of venous thromboembolism. To categorize potential bloodstream infections, we used Centers for Disease Control and Prevention definitions.9 All medical record and radiograph reviews were performed by investigators who were masked to patient firm assignment. We monitored patients until catheter removal or hospital discharge. After patient discharge, we reviewed the electronic record, including emergency room visits and repeat hospitalizations, for 30 days after the earlier of hospital discharge or catheter removal.

Statistics

Because we were aware that temporary dialysis catheters are sometimes placed in femoral veins to preserve the subclavian or internal jugular venous sites for more permanent tunneled intravascular catheters, our prespecified plans were to compare venous insertion sites between intervention and control groups after excluding temporary dialysis catheters. To more completely describe catheter use, we also collected data on temporary dialysis catheters, and we present the results both with and without inclusion of data on temporary dialysis catheters. If multiple residents attempted to insert a catheter, we would have used the group that the final inserter was in to determine intervention versus control group assignment; however, this never occurred.

To determine resident confidence in inserting catheters, we collapsed the responses of agree and strongly agree and of disagree and strongly disagree into single categories; thus, frequency of agreement was evaluated as a dichotomous outcome. To test whether residents' confidence changed between the 3 surveys, we analyzed responses using the matched‐pair signed rank test, with the initial survey used as the referent.

We dichotomized certain continuous variables using the following cut points: body mass index 30 kg/m²; coagulopathy, international normalized ratio (INR) > 1.5; thrombocytopenia, platelets < 100 10⁹/L. Data were entered into a relational database (Microsoft Access, Microsoft Inc., Redmond, WA) and merged analyzed using Stata software, version 8.2 (Stata Corporation, College Station, TX).

Patient and Catheter Characteristics

Fifty‐four catheters were inserted in 48 patients during the study period, 16 (30%) in the intervention group and 38 (70%) in the control group. Mean number of catheters inserted per resident for each 4‐week rotation was 0.24; therefore, on average, a resident would insert 1 catheter every 4 general‐medicine rotations. Most catheters were inserted between 7:00 _AM and 5:00 _PM; the most common reason for insertion was to administer intravenous medications to a patient without intravenous access, followed by the need for a temporary dialysis catheter. Most catheters were inserted by the medicine team rather than radiology or a subspecialty service (Table 1). Most patient characteristics and reasons for insertion were similar between groups; however, more patients in the control group had thrombocytopenia (Table 1).

Insertion Practices

Femoral venous insertion was the most common type of catheter insertion (67%), followed by internal jugular (26%) and subclavian (7%); there were no differences in insertion site between the intervention and control groups (Table 2). When we excluded temporary dialysis catheters, 39% of central venous catheters were inserted in the internal jugular vein. Although a smaller proportion of catheters inserted by the intervention group were placed in a femoral vein, the difference was not significant (Table 2).

For most insertions, residents reported using sterile gloves (94%) and a large drape (80%); however, most did not report use of a sterile gown (48%), mask (46%), or cap (15%). Residents in the intervention group were more likely to report use of a mask, and there was a trend toward increased use of large drapes (Table 2). No patient characteristics predicted femoral venous insertion (data not shown).

Complications

The most frequent complication was arterial puncture (n = 4); all four occurred during femoral venous insertion attempts. Compared to subclavian or internal jugular venous placement, there was a trend toward more mechanical complications among femoral catheters (Table 3). One episode of clinical sepsis occurred, in an intervention‐group patient who had femoral and internal jugular catheters, and no pneumothoraxes or episodes of venous thromboembolism occurred (Table 3). The overall incidence of bloodstream infection was 2.7 per 1000 central‐line days; there was no difference between the intervention and control groups (9.2 versus 0 per 1000 central‐line days; P = .29).

Survey Responses

Before the educational session, many residents did not recognize that femoral venous catheter insertions had a higher risk of arterial puncture or venous thrombosis (Table 4); by the final survey, residents were more likely to recognize the higher risk of these complications during femoral venous insertions. Most residents recognized the higher risk of infectious complications at the femoral site (Table 4).

Residents overwhelmingly responded that the lecture was useful (95%), that mannequins provided a valuable skill‐building exercise (90%), and that the session should be incorporated into the training program (95%). Immediately after the session, residents had increased confidence about inserting a central venous catheter at any venous site, especially for internal jugular or subclavian insertions. By the final survey, the confidence of residents about inserting catheters in the internal jugular or subclavian veins had returned to baseline but had increased for femoral‐site insertions (Table 4).

Most residents in the intervention group agreed that the educational session motivated them to remove unnecessary catheters, improve insertion‐related infection‐control practices, and place the catheter in an internal jugular or subclavian vein; some agreed because of the educational session, they were less likely to place a central venous catheter. Some reported successfully inserting a central venous catheter in the subclavian or internal jugular vein for the first time (Table 4).

Setting and Study Design

We conducted a prospective, firm‐based clinical trial approved by the institutional review board at Cook County Hospital, a 464‐bed public teaching hospital. We evaluated all central venous catheters inserted by residents on the general medicine service from November 15, 2004, to March 31, 2005. The internal medicine residency program assigns residents to 1 of 3 firms for their entire 3 years of training. We designated 1 firm as the intervention group; the other 2 firms constituted the control group.

Educational Intervention

At the beginning of each 4‐week general medicine ward rotation, intervention‐firm residents attended an educational and simulation laboratory session. Control‐firm residents received the usual ward orientation. We conducted 6 sessions, with total attendance of 40 intervention‐firm residents, or approximately 7 residents per session. A chief medical resident experienced in catheter placement and an attending internist led and supervised each 2‐hour training session. The sessions were conducted at the Simulation Laboratory of Rush University and included a presentation about indications for central venous catheter insertion, insertion techniques, common complications, and practice placing catheters in mannequins. During the hands‐on session, each participant observed the expert insert a triple‐lumen catheter in the mannequin's internal jugular and subclavian veins. Then, with supervision, each participant practiced catheter insertion using recommended infection‐control practices (eg, use of gloves, mask, and large drape, and chlorhexidine skin preparation).

Resident Survey

Before each session, we administered a survey that assessed residents' knowledge of insertion techniques and their confidence in placing catheters at each venous insertion site. To measure change in the confidence level of residents, we distributed an abbreviated survey 2 additional times, immediately after the session and at the end of the study period. We measured confidence with answers to survey questions, which were rated on a 5‐point Likert scale, from strongly disagree to strongly agree. In addition to measuring the change in residents' confidence, the final survey repeated knowledge assessment questions, evaluated residents' attitudes regarding venous insertion sites, and asked about potential strategies to improve insertion practices.

Central Venous Catheter Detection and Monitoring

At the end of each day, residents reported catheter insertions to chief residents during routine sign‐out rounds. If a catheter had been inserted, the chief resident interviewed the resident about type of catheter, venous insertion site, duration of attempt, patient location, immediate complications, number of inserters, inserter attendance at an educational session, inserter specialty, and professional designation (eg, resident, fellow, attending), indication for insertion, and adherence to infection‐control practices. For all insertion attempts, the research team reviewed the medical record and recorded patient characteristics that might influence venous insertion site (eg, thrombocytopenia, coagulopathy, and body mass index) and evaluated patients for insertion‐related complications.

We prospectively monitored patients for mechanical (ie, pneumothorax or hematoma), thromboembolic, or infectious complications. To evaluate for pneumothorax, postinsertion chest radiographs were reviewed by a physician‐investigator, and radiologists' interpretations and progress notes were reviewed. To evaluate for infectious or other mechanical complications, progress notes also were reviewed. We required radiographic confirmation of venous thromboembolism. To categorize potential bloodstream infections, we used Centers for Disease Control and Prevention definitions.9 All medical record and radiograph reviews were performed by investigators who were masked to patient firm assignment. We monitored patients until catheter removal or hospital discharge. After patient discharge, we reviewed the electronic record, including emergency room visits and repeat hospitalizations, for 30 days after the earlier of hospital discharge or catheter removal.

Statistics

Because we were aware that temporary dialysis catheters are sometimes placed in femoral veins to preserve the subclavian or internal jugular venous sites for more permanent tunneled intravascular catheters, our prespecified plans were to compare venous insertion sites between intervention and control groups after excluding temporary dialysis catheters. To more completely describe catheter use, we also collected data on temporary dialysis catheters, and we present the results both with and without inclusion of data on temporary dialysis catheters. If multiple residents attempted to insert a catheter, we would have used the group that the final inserter was in to determine intervention versus control group assignment; however, this never occurred.

To determine resident confidence in inserting catheters, we collapsed the responses of agree and strongly agree and of disagree and strongly disagree into single categories; thus, frequency of agreement was evaluated as a dichotomous outcome. To test whether residents' confidence changed between the 3 surveys, we analyzed responses using the matched‐pair signed rank test, with the initial survey used as the referent.

We dichotomized certain continuous variables using the following cut points: body mass index 30 kg/m²; coagulopathy, international normalized ratio (INR) > 1.5; thrombocytopenia, platelets < 100 10⁹/L. Data were entered into a relational database (Microsoft Access, Microsoft Inc., Redmond, WA) and merged analyzed using Stata software, version 8.2 (Stata Corporation, College Station, TX).

Patient and Catheter Characteristics

Fifty‐four catheters were inserted in 48 patients during the study period, 16 (30%) in the intervention group and 38 (70%) in the control group. Mean number of catheters inserted per resident for each 4‐week rotation was 0.24; therefore, on average, a resident would insert 1 catheter every 4 general‐medicine rotations. Most catheters were inserted between 7:00 _AM and 5:00 _PM; the most common reason for insertion was to administer intravenous medications to a patient without intravenous access, followed by the need for a temporary dialysis catheter. Most catheters were inserted by the medicine team rather than radiology or a subspecialty service (Table 1). Most patient characteristics and reasons for insertion were similar between groups; however, more patients in the control group had thrombocytopenia (Table 1).

Insertion Practices

Femoral venous insertion was the most common type of catheter insertion (67%), followed by internal jugular (26%) and subclavian (7%); there were no differences in insertion site between the intervention and control groups (Table 2). When we excluded temporary dialysis catheters, 39% of central venous catheters were inserted in the internal jugular vein. Although a smaller proportion of catheters inserted by the intervention group were placed in a femoral vein, the difference was not significant (Table 2).

For most insertions, residents reported using sterile gloves (94%) and a large drape (80%); however, most did not report use of a sterile gown (48%), mask (46%), or cap (15%). Residents in the intervention group were more likely to report use of a mask, and there was a trend toward increased use of large drapes (Table 2). No patient characteristics predicted femoral venous insertion (data not shown).

Complications

The most frequent complication was arterial puncture (n = 4); all four occurred during femoral venous insertion attempts. Compared to subclavian or internal jugular venous placement, there was a trend toward more mechanical complications among femoral catheters (Table 3). One episode of clinical sepsis occurred, in an intervention‐group patient who had femoral and internal jugular catheters, and no pneumothoraxes or episodes of venous thromboembolism occurred (Table 3). The overall incidence of bloodstream infection was 2.7 per 1000 central‐line days; there was no difference between the intervention and control groups (9.2 versus 0 per 1000 central‐line days; P = .29).

Survey Responses

Before the educational session, many residents did not recognize that femoral venous catheter insertions had a higher risk of arterial puncture or venous thrombosis (Table 4); by the final survey, residents were more likely to recognize the higher risk of these complications during femoral venous insertions. Most residents recognized the higher risk of infectious complications at the femoral site (Table 4).

Residents overwhelmingly responded that the lecture was useful (95%), that mannequins provided a valuable skill‐building exercise (90%), and that the session should be incorporated into the training program (95%). Immediately after the session, residents had increased confidence about inserting a central venous catheter at any venous site, especially for internal jugular or subclavian insertions. By the final survey, the confidence of residents about inserting catheters in the internal jugular or subclavian veins had returned to baseline but had increased for femoral‐site insertions (Table 4).

Most residents in the intervention group agreed that the educational session motivated them to remove unnecessary catheters, improve insertion‐related infection‐control practices, and place the catheter in an internal jugular or subclavian vein; some agreed because of the educational session, they were less likely to place a central venous catheter. Some reported successfully inserting a central venous catheter in the subclavian or internal jugular vein for the first time (Table 4).