Sustaining Improvement in Hand Hygiene and Health Care Associated Infections

Transcription

1 Improvement from Front Office to Front Line Sustaining Improvement in Hand Hygiene and Health Care Associated Infections The data from this study suggest that procedure-specific processes to reduce infections related to devices such as central lines or ventilators or others, such as urinary catheters, may not achieve extremely low rates of infection unless they are accompanied by high rates of hand hygiene compliance. Using the Targeted Solutions Tool to Improve Hand Hygiene Compliance Is Associated with Decreased Health Care Associated Infections (p. 15) Features Infection Prevention and Control Using the Targeted Solutions Tool to Improve Hand Hygiene Compliance Is Associated with Decreased Health Care Associated Infections Performance Improvement The Costs of Participating in a Diabetes Quality Improvement Collaborative: Variation Among Five Clinics Risk Assessment and Event Analysis A Tool for the Concise Analysis of Patient Safety Incidents Care Processes Factors Associated with Inpatient Thoracentesis Procedure Quality at University Hospitals Departments To Our Readers Tool Tutorial Post Event Debriefs: A Commitment to Learning How to Better Care for Patients and Staff

2 Care Processes Factors Associated with Inpatient Thoracentesis Procedure Quality at University Hospitals Sarah E. Kozmic, BS; Diane B. Wayne, MD; Joe Feinglass, PhD; Samuel F. Hohmann, PhD; Jeffrey H. Barsuk, MD, MS Approximately 178,000 thoracentesis procedures are performed in the United States each year. 1 The most common complication is pneumothorax, which occurs after 1.3% to 18% of procedures. 2 4 Because up to 34% of procedure-associated pneumothoraces require a chest tube, these events may result in hospital admission or add to existing hospital costs and length of stay (LOS). 5 Previous studies show that when experienced performers use ultrasonography and avoid unnecessary procedures, the rate of thoracentesis-associated iatrogenic pneumothorax is reduced. 4,6 However, many clinicians lack the experience and knowledge to safely perform thoracentesis procedures. 7 In the United States, thoracentesis procedures are increasingly referred to interventional radiology (IR) rather than performed at the bedside. 8,9 Among Medicare beneficiaries, thoracentesis procedures performed by radiologists increased by 358% from 14,531 to 66,602 from 1993 to The reason for this shift is likely multifactorial. The American Board of Internal Medicine (ABIM) no longer requires that internal medicine residents demonstrate competence in thoracentesis procedures, 10 and some physicians may lack the skills needed to safely perform them. Time constraints, inadequate reimbursement, and low physician confidence regarding their abilities are likely additional factors contributing to more frequent IR referrals. 11,12 Even among pulmonary and critical care medicine physicians, the rate of IR referral for thoracentesis procedures increased from 10% to 52% from 1993 to This trend is concerning, because bedside procedures are safe 1,13,14 and less costly than IR procedures. 13,15,16 Although national trends have been previously studied in Medicare patients, little is known about thoracentesis practice patterns at university hospitals. In addition, the impact of specialty on postthoracentesis clinical outcomes is unknown. Therefore, this study had two aims: (1) to describe which specialties performed inpatient thoracentesis procedures at University HealthSystem Consortium (UHC) hospitals, and (2) to evaluate the extent to which the rates of iatrogenic pneumothorax, total hospital costs, and LOS varied by provider specialty. Article-at-a-Glance Background: Physicians increasingly refer thoracentesis procedures to interventional radiology (IR) rather than performing them at the bedside. Factors associated with thoracentesis procedures at university hospitals were studied to determine clinical outcomes by provider specialty. Methods: An administrative database review was performed of patients who underwent an inpatient thoracentesis procedure in hospitals participating in the University HealthSystem Consortium (UHC) Database from January 2010 through September The incidence of iatrogenic pneumothorax, mean total hospital costs, and mean length of stay (LOS) were compared by clinical specialty. Results: There were 113,860 admissions with 132,472 thoracentesis procedures performed on 99,509 patients at 234 UHC hospitals. IR performed 43,783 (33%) thoracentesis procedures; medicine, 22,243 (17%); and pulmonary, 26,887 (20%). The incidence of iatrogenic pneumothorax was 2.8% for IR, 2.9% for medicine, and 3.1% for pulmonary. Medicine and pulmonary had equivalent risk of iatrogenic pneumothorax compared to IR after controlling for clinical covariates. Admissions with medicine and pulmonary procedures were associated with significantly lower costs compared to IR admissions (p < 0.001) after controlling for clinical covariates. Admissions with IR procedures had a mean LOS of 14.1 days; medicine, 13.2 days; and pulmonary, 15.9 days. Admissions with medicine and pulmonary procedures were associated with fewer hospital days when compared to IR in the controlled model (p < 0.001). Conclusion: Admissions with medicine and pulmonary bedside thoracentesis procedures are as safe and less costly than IR procedures. Shifting IR thoracentesis procedures to the bedside might be a potential way to reduce hospital costs while still ensuring high-quality patient care, provided that portable ultrasound is used. 34

3 Methods Design We analyzed data from all patients with inpatient thoracentesis procedures in UHC Database hospitals from January 2010 through September UHC is an alliance of 120 nonprofit academic medical centers and their 290 affiliated hospitals. 17 The UHC Database contains clinical, operational, financial, and patient safety data from affiliated hospitals. The Northwestern University Institutional Review Board approved this study. We compared demographic and clinical information for all patients who underwent thoracentesis procedures analyzed by the clinical specialty of the physician performing the procedure. We also compared the incidence of iatrogenic pneumothorax, total hospital costs, and LOS by clinical specialty. Procedure We analyzed all patients older than the age of 18 years with thoracentesis procedures (International Classification of Diseases Ninth Revision [ICD-9] procedure code 34.91) during the study period. We collected patient demographic and clinical characteristics for each admission, including age, sex, procedure provider specialty, and whether there was an ICU stay. We calculated a Charlson score to assess severity of illness on admission. 18,19 The Charlson score is based on 19 chronic disease comorbidities derived from ICD-9 codes and predicts one-year mortality for hospitalized patients. We also obtained the All Patient Refined Diagnosis Related Group (APR-DRG) Severity of Illness (SOI) code for all admissions from the UHC Database. The APR-DRG admission SOI code is calculated based on ICD- 9 codes present on admission and rates severity of illness from 1 to 4 (1 = minor, 2 = moderate, 3 = major, 4 = extreme). 20,21 The APR-DRG SOI code was designed to control for severity of illness in cost models. ICD-9 codes were also used to identify patients with coagulation disorders, blood loss anemia, hyponatremia, hypotension, and thrombocytopenia. In addition, we used ICD-9 procedure codes to identify patients who had a central venous catheter (CVC) insertion, pacemaker insertion, positive pressure ventilation (invasive and noninvasive), pericardiocentesis, or endoscopic or percutaneous lung biopsy in addition to a thoracentesis. We queried the UHC Database for clinical outcomes, including iatrogenic pneumothorax (ICD-9 code 512.1), total hospital cost, and LOS. We used codes for present on admission to exclude preexisting pneumothorax diagnosis codes. UHC generates direct cost estimates by applying Medicare Cost Report ratios of cost to charges, with the labor cost further adjusted by the respective area wage index. We excluded admissions with missing data from our cost analyses. Admissions were divided into six categories on the basis of the specialty performing the thoracentesis: interventional radiology, including interventional or diagnostic radiology (IR); medicine, including family medicine, general medicine, and hospital medicine; pulmonary, including pulmonary and critical care; subspecialty medicine, including infectious disease, cardiology, nephrology, hematology/oncology, endocrinology, gastroenterology/hepatology, and geriatrics; surgery, including general surgery and cardiothoracic surgery; and all other, including other and unidentified specialties. We described patient characteristics categorized by specialty group when it was the only procedure provider. Patients who underwent procedures by more than one specialty group within the same admission were classified as hybrid admissions. Statistical Analysis To test the significance of differences between patient characteristics and outcomes across specialties, we used chi-square tests for categorical variables and analysis of variance (ANOVA) or the Kruskal-Wallis rank test for continuous variables. We used random-effects Poisson regression, which adjusts standard errors for multiple procedures for individual patients and produces an incidence rate ratio (IRR) that is closer to relative risk than the odds ratio, to model the likelihood of iatrogenic pneumothorax by specialty. 22,23 Differences between specialties were adjusted for patient age, gender, obesity, ICU stay, Charlson score, APR-DRG SOI code, number of thoracentesis procedures performed during the admission, coagulation disorders, blood loss anemia, hyponatremia, hypotension, thrombocytopenia, and other procedures that are associated with iatrogenic pneumothorax (CVC insertion, pacemaker insertion, positive pressure ventilation, pericardiocentesis, and endoscopic or percutaneous lung biopsy). We used random-effects multiple linear regression to model cost and LOS by medical specialty adjusting for patient age, gender, obesity, ICU stay, Charlson score, APR-DRG SOI code, number of thoracentesis procedures performed during the hospital stay, coagulation disorders, blood loss anemia, hyponatremia, hypotension, thrombocytopenia, CVC insertion, pacemaker insertion, positive pressure ventilation, pericardiocentesis, and endoscopic or percutaneous lung biopsy. We restricted samples used in all regression models to admissions at hospitals that performed both IR and medicine procedures. We also estimated random-effects models using log transformations of cost and LOS to test the sensitivity of specialty differences because hospital cost and LOS data were not 35

4 normally distributed. We only present linear models of cost and LOS in the results to improve interpretation and because these models yielded nearly identical findings. We used random-effects multiple linear regression to model cost by medical specialty adjusting for LOS in addition to the other covariates listed. We performed this analysis to determine the differences in costs by specialty beyond costs attributable to increased LOS. We present the results of the cost, LOS, and LOS adjusted cost models. In addition, we performed two sensitivity analyses. First, we excluded admissions with other procedures that are associated with iatrogenic pneumothorax and may also increase LOS and cost (CVC or pacemaker insertion, positive pressure ventilation, pericardiocentesis, and endoscopic or percutaneous lung biopsy) in the regression models of iatrogenic pneumothorax, cost and LOS. We only present the full analysis because the results were virtually identical (with IRRs within hundredths difference). Next, we estimated a logistic regression model of the likelihood of IR procedures to determine the extent of selective referral for those patients, using the same demographic and clinical covariates described above as independent variables. We tested whether inclusion of predicted probabilities for IR referral derived from this model (a propensity score) 24 affected our models of pneumothorax, costs, and LOS. We found that this referral model was very weakly predictive (c-statistic = 0.59) and that inclusion of IR propensity scores had virtually no effect on results. We performed all statistical analyses using Stata version 12 (Statacorp, LP; College Station, Texas). Results Thoracentesis Procedures Our UHC Database query provided information on 132,472 thoracentesis procedures (113,860 admissions) performed on 99,509 patients at 234 hospitals. Table 1 (right) presents the frequencies of each specialty group performing thoracentesis procedures. IR performed 33% of the procedures. Medicine, pulmonary, and subspecialty medicine performed 17%, 20%, and 8% of the procedures, respectively. There were 12,167 (9%) procedures with unknown specialty. Demographic and Clinical Data Table 2 (page 37) presents demographic and clinical data. Patient characteristics were significantly different across specialty groups. Admissions with pulmonary or surgery procedures were more likely to have an ICU stay (51%, 64%, respectively) compared to IR (37%, both p < 0.001). However, medicine admissions had similar ICU stays (36%) compared to IR. Admissions with pulmonary procedures had a higher severity of illness shown Table 1. Thoracentesis Procedures (N = 132,472) Performed at UHC Hospitals, January 2010 September 2013, by Specialty Specialty No. % Interventional radiology 43, Medicine 22, Family medicine 1, General medicine 18, Hospital medicine 3, Pulmonary 26, Subspecialty medicine 9, Infectious disease Cardiology 3, Nephrology Hematology/oncology 3, Endocrinology Gastroenterology 1, Hepatology Geriatrics Surgery 6, General surgery 2, Cardiothoracic surgery 3, All other 23, Other specialties 10, Specialty unknown 12, by a Charlson score of 3.8 compared to admissions with IR procedures at 3.6 (p < 0.001), while admissions with surgery procedures had a lower mean Charlson score (2.9) compared to IR (p < 0.001). Medicine patients had similar Charlson scores to IR. Iatrogenic pneumothoraces occurred in 2,618 (3.1%) admissions. The incidence of iatrogenic pneumothorax was lower for IR, medicine, and subspecialty medicine procedures (2.8%, 2.9%, and 2.9%, respectively, p < 0.001). Higher rates of iatrogenic pneumothorax occurred for admissions with surgery procedures (5.4%, p < 0.001). Cost data were available for 94.9% of admissions, resulting in a mean total cost of $49,366. Mean costs for admissions with IR, medicine, and subspecialty medicine procedures were significantly lower ($43,984, $48,745, and $38,119, respectively, p < 0.001). Admissions with pulmonary or surgery procedures were significantly higher ($51,669 and $80,406, respectively, p < 0.001). Admissions with IR procedures had a mean LOS of 14.1 days compared to 13.2 days for medicine procedures (p < 0.001). LOS for admissions with pulmonary procedures was significantly higher at 15.9 days (p < 0.001). 36

5 Table 2. Demographic and Clinical Data by Specialty Performing the Thoracentesis Procedure; All Admissions, January 2010 September 2013* Age Group (years) All Admissions N = 113,860 Interventional Radiology n = 36,551 Medicine n = 18,201 Pulmonary n = 22,070 Subspecialty Medicine n = 8,301 Surgery n = 5,532 All Other Specialties n = 19,818 Hybrid (more than one specialty) n = 3, ,347 (16.1) 5,700 (15.6) 3,033 (16.7) 3,351 (15.2) 1,384 (16.7) 936 (16.9) 3,334 (16.8) 609 (18.0) ,954 (19.3) 6,888 (18.8) 3,524 (19.4) 4,003 (18.1) 1,794 (21.6) 1,023 (18.5) 3,896 (19.7) 826 (24.4) ,316 (24.0) 8,524 (23.3) 4,036 (22.2) 5,399 (24.5) 2,143 (25.8) 1,294 (23.4) 5,073 (25.6) 847 (25.0) ,115 (20.3) 7,507 (20.5) 3,423 (18.8) 4,631 (21.0) 1,651 (19.9) 1,229 (22.2) 4,083 (20.6) 591 (17.4) ,128 (20.3) 7,932 (21.7) 4,185 (23.0) 4,686 (21.2) 1,329 (16.0) 1,050 (19.0) 3,432 (17.3) 514 (15.2) Male 58,373 (51.3) 18,467 (50.5) 9,379 (51.5) 11,519 (52.2) 4,440 (53.5) 3,008 (54.4) 9,794 (49.4) 1,766 (52.1) Obesity, BMI 40+ 8,432 (7.4) 3,025 (8.3) 1,224 (6.7) 1,615 (7.3) 526 (6.3) 453 (8.2) 1,296 (6.5) 293 (8.7) ICU stay 48,294 (42.4) 13,360 (36.6) 6,554 (36.0) 11,266 (51.0) 2,810 (33.9) 3,543 (64.0) 8,819 (44.5) 1,942 (57.3) Coagulation disorders 10,795 (9.5) 3,157 (8.6) 1,538 (8.5) 2,278 (10.3) 654 (7.9) 598 (10.8) 2,082 (10.5) 488 (14.4) Blood loss anemia 2,007 (1.8) 666 (1.8) 321 (1.8) 402 (1.8) 104 (1.3) 77 (1.4) 359 (1.8) 78 (2.3) Hyponatremia 18,710 (16.4) 5,856 (16.0) 3,086 (17.0) 3,629 (16.4) 1,235 (14.9) 805 (14.6) 3,319 (16.7) 780 (23.0) Hypotension 9,930 (8.7) 3,093 (8.5) 1,374 (7.5) 2,017 (9.1) 740 (8.9) 512 (9.3) 1,834 (9.3) 360 (10.6) Thrombocytopenia 15,904 (14.0) 4,687 (12.8) 2,391 (13.1) 3,289 (14.9) 1,135 (13.7) 768 (13.9) 3,004 (15.2) 630 (18.6) Mean Charlson 3.7 (3.0) 3.6 (3.0) 3.7 (2.9) 3.8 (3.0) 4.1 (3.0) 2.9 (2.7) 3.8 (3.0) 4.0 (3.0) score (SD) APR-DRG SOI 1 1,572 (1.4) 510 (1.4) 146 (0.8) 204 (0.9) 52 (0.6) 213 (3.9) 420 (2.1) 27 (0.8) 2 18,807 (16.7) 6,384 (17.7) 3,015 (16.8) 2,918 (13.3) 1,292 (15.7) 1,292 (23.5) 3,520 (17.9) 386 (11.4) 3 53,825 (47.7) 18,105 (50.1) 8,529 (47.4) 9,604 (43.8) 4,193 (51.0) 2,523 (45.9) 9,437 (48.0) 1,434 (42.4) 4 38,640 (34.2) 11,134 (30.8) 6,309 (35.1) 9,226 (42.0) 2,678 (32.6) 1,472 (26.8) 6,284 (32.0) 1,537 (45.4) Mean (SD) number of thoracentesis procedures per admission 1.2 (0.5) 1.1 (0.4) 1.1 (0.5) 1.1 (0.4) 1.1 (0.4) 1.1 (0.4) 1.1 (0.4) 2.4 (0.9) Central venous catheter insertion 14,338 (12.6) 4,336 (11.9) 1,964 (10.8) 3,260 (14.8) 757 (9.1) 763 (13.8) 2,573 (13.0) 685 (20.2) Pacemaker insertion 1,099 (1.0) 291 (0.8) 115 (0.6) 191 (0.9) 99 (1.2) 116 (2.1) 243 (1.2) 44 (1.3) Noninvasive 7,574 (6.7) 1,838 (5.0) 1,265 (7.0) 1,838 (8.3) 479 (5.8) 412 (7.4) 1,417 (7.2) 325 (9.6) ventilation Invasive ventilation 22,143 (19.4) 5,354 (14.6) 2,750 (15.1) 6,122 (27.7) 1,131 (13.6) 1,526 (27.6) 4,252 (21.5) 1,008 (29.8) Pericardiocentesis 1,378 (1.2) 330 (0.9) 205 (1.1) 259 (1.2) 213 (2.6) 70 (1.3) 252 (1.3) 49 (1.4) Endoscopic lung biopsy 11,198 (9.8) 2,665 (7.3) 1,486 (8.2) 3,417 (15.5) 480 (5.8) 534 (9.7) 2,111 (10.7) 505 (14.9) Percutaneous lung biopsy 933 (0.8) 416 (1.1) 154 (0.8) 135 (0.6) 46 (0.6) 18 (0.3) 128 (0.6) 36 (1.1) BMI, body mass index; SD, standard deviation; APR-DRG SOI, All Patient Refined Diagnosis Related Group Severity of Illness: 1 = minor, 2 = moderate, 3 = major, 4 = extreme. * All comparisons across all specialties; p < 0.001; chi-square tests for categorical variables and analysis of variance or the Kruskal-Wallis rank test for continuous variables. 37

6 Likelihood of Iatrogenic Pneumothorax Appendix 1 (available in online article) presents results of random-effects Poisson regression for the likelihood of iatro genic pneumothorax. When compared to IR, procedures per formed by medicine had approximately the same risk of iatrogenic pneumothorax (IRR 1.07; 95% confidence interval [CI], ). Pulmonary (IRR 1.00; 95% CI, ) and subspecialty medicine (IRR 1.05; 95% CI, ) also had equivalent risk of iatrogenic pneumothorax as IR, whereas thoracentesis procedures performed by surgery specialties resulted in a significantly higher risk of iatrogenic pneumothorax (IRR 1.66; 95% CI, ). Patients who also had a CVC insertion had a lower risk of iatrogenic pneumothorax (IRR 0.86; 95% CI, ), while patients who had a pacemaker insertion had a 46% higher risk (IRR 1.46; 95% CI, ), and patients undergoing percutaneous lung biopsy had a 4.5-times higher risk (IRR 4.46; 95% CI ). Regression Models for Total Cost and Length of Stay Appendix 2 (available in online article) displays results of the regression models for total cost and LOS. Admissions with medicine, pulmonary, and subspecialty medicine procedures had significantly lower costs than IR admissions (p < 0.001), while admissions with surgery procedures were associated with $10,635 in additional costs compared to IR procedures (p < 0.001). Admissions with medicine procedures were also associated with a three-day decrease in LOS compared to IR admissions (p < 0.001). Admissions with pulmonary and subspecialty medicine procedures were associated with decreased LOS compared to IR admissions (3.5 and 2.8, respectively, p < 0.001). Each additional hospital day was associated with an additional cost of $3,392 (p < ). In the LOS adjusted cost model, admissions with medicine procedures were associated with a cost savings of $5,725 compared to admissions with IR procedures (p < 0.001). Admissions with pulmonary and subspecialty medicine procedures were also associated with lower costs than IR ($4,703, p < and $2,055, p = 0.002, respectively), while admissions with surgery and all other specialty procedures were associated with higher LOS adjusted costs (p < 0.001). Discussion This study shows that a substantial number of inpatient thoracentesis procedures are performed at the bedside by internal medicine trained clinicians at university hospitals. It is possible that because of the limitations of the classification system, some procedure performers were misclassified. For example, trauma surgeons may have been listed under critical care, and critical care physicians, under pulmonary. However, this methodology was used in another study evaluating Medicare patients, which, including ambulatory procedures, revealed that internal medicine trained clinicians performed only approximately 30% of thoracentesis procedures, with approximately 23% performed by pulmonary/critical care. 8 Although the ABIM no longer requires that internal medicine residents demonstrate competency in thoracentesis procedures, 10 our findings show that internal medicine trained clinicians frequently perform them at university hospitals. Because we used an administrative database, it was not possible to determine if the clinicians who performed the thoracentesis procedures in this study were residents, fellows, or attending physicians. However, it is likely that many of these bedside thoracentesis procedures were performed by internal medicine residents and pulmonary fellows. This is concerning, because internal medicine trainees report variable confidence, experience, expertise, and supervision regarding invasive procedures. 12, Earlier research also demonstrates that graduating residents and fellows are not always competent to perform procedures such as thoracentesis, 29 temporary hemodialysis catheter insertion, 30 and lumbar puncture. 31 Simulation-based education boosts thoracentesis clinical skills 7,29 and improves patient care quality in invasive procedures including advanced cardiac life support, 32,33 paracentesis, 13 lumbar puncture, 34 and central venous catheter insertion To rigorously assess clinical skills, we recommend that all clinicians who perform bedside thoracentesis procedures undergo simulation-based education before performing the procedure on actual patients. Another finding of this study is an equivalent incidence of iatrogenic pneumothorax after bedside and IR thoracentesis procedures. This safety profile is similar to findings of earlier research 1,13,23 and provides additional rationale to perform bedside thoracentesis procedures. Although the use of ultrasound is associated with a decreased risk of pneumothorax, we were unable to obtain information on ultrasound use from the UHC Data base. In addition, in our own institution administrative data underestimate the use of ultrasound for procedures because clinicians do not typically save the ultrasound image and therefore cannot bill for it. For this reason, administrative data could be inaccurate even if they were available. The higher rate of iatrogenic pneumothorax seen in procedures performed by surgery may be the result of several factors. One possibility is the performance of other invasive chest procedures not identified in our models. Earlier studies found increased costs and LOS associated 38

7 with IR procedures compared to bedside procedures. 13,15,16 This study demonstrates similar findings with thoracentesis procedures because admissions with IR thoracenteses were associated with additional cost and LOS compared to admissions with bedside procedures. Scheduling delays associated with IR procedures likely contribute to higher LOS and increased costs. However, as shown in the LOS adjusted model, not all of the increased cost of admissions with IR procedures can be attributed to increased LOS. Although we are limited to the administrative data available for this study, further study is needed to identify differences in procedure costs by provider specialty. IR procedures also require space, transport, and nursing, which contribute to additional costs. In the current economic climate, hospitals must identify strategies to reduce LOS and cost. However, the large cost and LOS differences seen in IR patients may be due to unknown patient characteristics, and further study is needed to evaluate these differences. This study shows that shifting IR thoracentesis procedures to the bedside might be a potential way to reduce hospital costs while still ensuring high-quality patient care, provided portable ultrasound is used. In addition, a recent study showed that patients often prefer the bedside location for thoracentesis procedures. 39 Limitations This study has several limitations. First, our analysis was limited to the information available from ICD-9 codes in the UHC Database. ICD-9 codes tend to be very specific but lack sensitivity for certain diagnoses The observational nature of this study limits our ability to make cause-and-effect inferences, and we may have overlooked important patient data reflecting severity of illness at admission or specialty selection factors. However, we did attempt to control for severity of illness by using Charlson score and APR-DRG SOI codes, as well as if patients had documented clinical covariates thought to influence thoracentesis outcomes, which, like all severity of illness measures, have their own limitations. Second, we were unable to confirm that iatrogenic pneumothoraces were caused exclusively by thoracentesis procedures. However, we conducted two analyses, one controlling for other common procedures that cause pneumothorax, and one excluding them. Third, given our inability to perform medical records reviews, we were unable to include important clinical outcomes such as inpatient mortality and 30-day hospital readmission resulting from thoracentesis procedures. In addition, coding for hemothorax did not allow enough specificity to link this complication to thoracentesis procedures. Fourth, we did not fully examine selection bias in specialty care. However, our propensity score model revealed that IR referral was highly discretionary, and the odds ratios in the IR referral propensity model were so small (data not shown) that they did not affect the overall results when added to the regression models. Finally, we were unable to explore additional variables such as referral patterns, operator experience, and diagnostic versus therapeutic procedures. We encourage individual institutions to pursue the impact of these variables in future studies. Conclusions Although IR referral rates are increasing nationally, providers continue to perform a substantial number of bedside thoracentesis procedures at university hospitals. These bedside procedures are as safe as IR procedures and are less costly. Bedside thoracentesis procedures benefit patients and society by reducing costs while maintaining quality. Hospitals should consider mechanisms to promote the performance of bedside thoracentesis procedures. J Sarah E. Kozmic, BS, formerly Clinical Research Associate, Northwestern University Feinberg School of Medicine, Chicago, is Patient Centered Outcomes Research Coordinator, Advocate Lutheran General Hospital, Park Ridge, Illinois. Diane B. Wayne, MD, is Vice Dean, Education, and the Dr. John Sherman Appleman Professor of Medicine, Northwestern University Feinberg School of Medicine. Joe Feinglass, PhD, is Research Professor of Medicine, Northwestern University Feinberg School of Medicine. Samuel F. Hohmann, PhD, is Principal Consultant, Comparative Data & Informatics Research, University HealthSystem Consortium, Chicago. Jeffrey H. Barsuk, MD, MS, is Director of Simulation and Patient Safety and Associate Professor of Medicine, Northwestern University Feinberg School of Medicine. Please address correspondence to Jeffrey H. Barsuk, jbarsuk@northwestern.edu. Online Only Content See the online version of this article for Appendix 1. Random Effects Poisson Regression Results for the Effect of Physician Specialty on the Likelihood of Iatrogenic Pneumothorax for Patients with Thoracentesis Procedures (N = 101,915 Admissions from January 2010 to September 2013 at 185 Hospitals with Both Interventional Radiology and Medicine Specialties Performing Thoracentesis) Appendix 2. Random Effects Linear Regression Results for the Effect of Physician Specialty on the Associations with Inpatient Total Cost and Length of Stay for Patients with Thoracentesis Procedures (N = 101,915 Admissions from January 2010 to September 2013 at 185 Hospitals with Both Interventional Radiology and Medicine Specialties Performing Thoracentesis) 39

8 References 1. Daniels CE, Ryu JH. Improving the safety of thoracentesis. Curr Opin Pulm Med. 2011;17: Wrightson JM, et al. Pleural procedures and pleuroscopy. Respirology. 2009; 14: Raptopoulos V, et al. Factors affecting the development of pneumothorax associated with thoracentesis. AJR Am J Roentgenol. 1991;156: Colt HG, Brewer N, Barbur E. Evaluation of patient-related and procedure-related factors contributing to pneumothorax following thoracentesis. Chest. 1999;116: Gordon CE, et al. Pneumothorax following thoracentesis: A systematic review and meta-analysis. Arch Intern Med Feb 22;170: Wrightson JM, et al. Risk reduction in pleural procedures: Sonography, simulation and supervision. Curr Opin Pulm Med. 2010;16: Duncan DR, et al. Reducing iatrogenic risk in thoracentesis: Establishing best practice via experiential training in a zero-risk environment. Chest. 2009; 135: Duszak R Jr, Chatterjee AR, Schneider DA. National fluid shifts: Fifteen year trends in paracentesis and thoracentesis procedures. J Am Coll Radiol. 2010;7: Thakkar R, et al. Procedures performed by hospitalist and non-hospitalist general internists. J Gen Intern Med. 2010;25: American Board of Internal Medicine. Internal Medicine Policies. Accessed Dec 1, Wigton RS, Alguire P; American College of Physicians. The declining number and variety of procedures done by general internists: A resurvey of members of the American College of Physicians. Ann Intern Med Mar 6;146: Huang GC, et al. Beyond the comfort zone: Residents assess their comfort performing inpatient medical procedures. Am J Med. 2006;119:71.e e Barsuk JH, et al. Clinical outcomes after bedside and interventional radiology paracentesis procedures. Am J Med. 2013;126: Grogan DR, et al. Complications associated with thoracentesis: A prospective, randomized study comparing three different methods. Arch Intern Med. 1990;150: Barsuk JH, et al. Cost savings of performing paracentesis procedures at the bedside after simulation-based education. Simul Healthc. 2014;9: Barsuk JH, et al. Specialties performing paracentesis procedures at university hospitals: Implications for training and certification. J Hosp Med. 2014;9: UHC. Home page. Accessed Dec 1, Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. 1992;45: Romano PS, Roos LL, Jollis JG. Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: Differing perspectives. J Clin Epidemiol. 1993;46: M. 3M APR DRG Software. Accessed Dec 1, /Products-and-Services/Products-List-A-Z/APR-DRG-Software/. 21. Adrados M, et al. The effect of severity of illness on total joint arthroplasty costs across New York State hospitals: An analysis of 172,738 cases. J Arthroplasty. 2015;30: Zhang J, Yu KF. What s the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes. JAMA Nov 18;280: Zou G. A modified Poisson regression approach to prospective studies with binary data. Am J Epidemiol Apr 1;159: Rosenbaum PR. The role of known effects in observational studies. Biometrics. 1989;45: Lucas BP, et al. Impact of a bedside procedure service on general medicine inpatients: A firm-based trial. J Hosp Med. 2007;2: Farnan JM, et al. A systematic review: The effect of clinical supervision on patient and residency education outcomes. Acad Med. 2012;87: Haber LA, et al. Effects of increased overnight supervision on resident education, decision-making, and autonomy. J Hosp Med. 2012;7: Berns JS, O Neill WC. Performance of procedures by nephrologists and nephrology fellows at U.S. nephrology training programs. Clin J Am Soc Nephrol. 2008;3: Wayne DB, et al. Mastery learning of thoracentesis skills by internal medicine residents using simulation technology and deliberate practice. J Hosp Med. 2008;3: Barsuk JH, et al. Mastery learning of temporary hemodialysis catheter insertion skills by nephrology fellows using simulation technology and deliberate practice. Am J Kidney Dis. 2009;54: Barsuk JH, et al. Simulation-based education with mastery learning improves residents lumbar puncture skills. Neurology Jul 10;79: Didwania A, et al. Progress toward improving the quality of cardiac arrest medical team responses at an academic teaching hospital. J Grad Med Educ. 2011;3: Wayne DB, et al. Simulation-based education improves quality of care during cardiac arrest team responses at an academic teaching hospital: A casecontrol study. Chest. 2008;133: Kessler DO, et al. A randomized trial of simulation-based deliberate practice for infant lumbar puncture skills. Simul Healthc. 2011;6: Barsuk JH, et al. Use of simulation-based education to reduce catheterrelated bloodstream infections. Arch Intern Med Aug 10;169: Barsuk JH, et al. Dissemination of a simulation-based mastery learning intervention reduces central line-associated bloodstream infections. BMJ Qual Saf. 2014;23: Barsuk JH, et al. Use of simulation-based mastery learning to improve the quality of central venous catheter placement in a medical intensive care unit. J Hosp Med. 2009;4: Barsuk JH, et al. Simulation-based mastery learning reduces complications during central venous catheter insertion in a medical intensive care unit. Crit Care Med. 2009;37: Barsuk JH, et al. Are we providing patient-centered care? Preferences about paracentesis and thoracentesis procedures. Patient Experience Journal. 2014; 1(2): Birman-Deych E, et al. Accuracy of ICD-9-CM codes for identifying cardiovascular and stroke risk factors. Med Care. 2005;43: Goldstein LB. Accuracy of ICD-9-CM coding for the identification of patients with acute ischemic stroke: effect of modifier codes. Stroke. 1998;29: Hou JK, et al. Accuracy of diagnostic codes for identifying patients with ulcerative colitis and Crohn s disease in the Veterans Affairs health care system. Dig Dis Sci. 2014;59: Golinvaux NS, et al. Limitations of administrative databases in spine research: A study in obesity. Spine J Dec 1;14: Goto M, et al. Accuracy of administrative code data for the surveillance of healthcare-associated infections: A systematic review and meta-analysis. Clin Infect Dis. 2014;58: Marcus P, Braman SS. International classification of disease coding for obstructive lung disease: Does it reflect appropriate clinical documentation? Chest. 2010;138:

9 Online Only Content The Joint Commission Journal on Quality and Patient Safety Appendix 1. Random Effects Poisson Regression Results for the Effect of Physician Specialty on the Likelihood of Iatrogenic Pneumothorax for Patients with Thoracentesis Procedures (N = 101,915 Admissions from January 2010 to September 2013 at 185 Hospitals with Both Interventional Radiology and Medicine Specialties Performing Thoracentesis) Age Group (years) All Admissions (N = 101,915 Admissions) Incidence Rate Ratio Reference Lower 95% CI Upper Male Obesity, BMI ICU stay Charlson score APR-DRG SOI 1 Reference Specialty Category Interventional radiology Reference Medicine Pulmonary Subspecialty medicine Surgery All Admissions (N = 101,915 Admissions) Incidence 95% CI Rate Ratio Lower Upper All other specialties Hybrid No. of thoracentesis procedures per admission Coagulation disorders Blood loss anemia Hyponatremia Hypotension Thrombocytopenia Central venous catheter insertion Pacemaker insertion Positive pressure ventilation Pericardiocentesis Endoscopic lung biopsy Percutaneous lung biopsy CI, confidence interval; BMI, body mass index; APR-DRG SOI, All Patient Refined Diagnosis Related Group Severity of Illness: 1 = minor, 2 = moderate, 3 = major, 4 = extreme. AP1

10 Online Only Content The Joint Commission Journal on Quality and Patient Safety Appendix 2. Random Effects Linear Regression Results for the Effect of Physician Specialty on the Associations with Inpatient Total Cost and Length of Stay for Patients with Thoracentesis Procedures (N = 101,915 Admissions from January 2010 to September 2013 at 185 Hospitals with Both Interventional Radiology and Medicine Specialties Performing Thoracentesis) Age Group (years) Total Cost Length of Stay B SE P Value B SE P Value Reference Reference , < < , < < , < < , < < Male 4, < < Obesity, BMI Intensive care unit stay 30, < < Charlson score APR-DRG SOI 1 Reference Reference 2 4, , < , , < , , < < Specialty Category Interventional radiology Reference Reference Medicine 16, < < Pulmonary 16, < < Subspecialty medicine 11, , < < Surgery 10, , < All other specialties 1, < Hybrid 20, , < < Number of thoracentesis procedures per admission 22, < < Coagulation disorders 25, < < Blood loss anemia 7, , < < Hyponatremia 11, < < Hypotension 2, Thrombocytopenia 15, < < Central venous catheter insertion 31, < < Pacemaker insertion 55, , < < Positive pressure ventilation 46, < < Pericardiocentesis 3, , Endoscopic lung biopsy 40, < < Percutaneous lung biopsy 2, , < B, beta (estimate of the regression coefficient); SE, standard error; BMI, body mass index; APR-DRG SOI, All Patient Refined Diagnosis Related Group Severity of Illness: 1 = minor, 2 = moderate, 3 = major, 4 = extreme. AP2