HIP FRACTURES IN THE ELderly

ORIGINAL INVESTIGATION Effects of a Hospitalist Model on Elderly Patients With Hip Fracture Michael P. Phy, DO; David J. Vanness, PhD; L. Joseph Melton III, MD; Kirsten Hall Long, PhD; Cathy D. Schleck; Dirk R. Larson, MS; Paul M. Huddleston, MD; Jeanne M. Huddleston, MD Background: Hospitalists increased role in perioperative medicine allows for examination of their effects on surgical patients. This study examined the effects of a hospitalist service created to medically manage elderly patients with hip fracture. Methods: During a 2-year historical cohort study of 466 patients 65 years or older admitted for surgical repair of hip fracture, we examined outcomes 1 year prior to and subsequent to the change from the standard to the hospitalist model. Results: The mean (SD) time to surgery (38 [47] vs 25 [53] hours; P.001), time from surgery to dismissal (9 [8] vs 7 [5] days; P=.04), and length of stay (10.6 [9] vs 8.4 [6] days; P.001) were shorter in the hospitalist group. Predictors of shorter time to surgery were care by the hospitalist group (P=.002), older age (P=.01), and fall as the mechanism of fracture (P.001), while American Society of Anesthesia scores of 3 and 4 were associated with increased time to surgery (P.001). Receiving care by the hospitalist group (P.001) and diagnosis of delirium (P.001) were associated with increased chance of earlier dismissal, while admission to the intensive care unit decreased this chance (P.001). Diagnosis of delirium was more frequent in the hospitalist group (74 [32.2%] of 230 vs 42 [17.8%] of 236; P.001). There were no differences in inpatient deaths or 30-day readmission rates. Conclusion: In elderly patients with hip fracture, a hospitalist model decreased time to surgery, time from surgery to dismissal, and length of stay without adversely affecting inpatient deaths or 30-day readmission rates. Arch Intern Med. 2005;165:796-801 Author Affiliations: Hospital Internal Medicine, Division of General Internal Medicine (Drs Phy and J. M. Huddleston), Division of Epidemiology (Dr Melton), Division of Health Care Policy and Research (Drs Long and J. M. Huddleston), Division of Biostatistics (Ms Schleck and Mr Larson), and the Department of Orthopedic Surgery (Dr P. M. Huddleston), Mayo Clinic College of Medicine, Rochester, Minn; and the Department of Population Health Sciences, University of Wisconsin Medical School, Madison (Dr Vanness). Financial Disclosure: Dr Vanness receives an honorarium from the Mayo Foundation, Rochester. HIP FRACTURES IN THE ELderly are a frequent, morbid, and expensive medical problem. 1 In 1998, over 320000 patients with hip fracture were admitted to US hospitals; persons over the age of 65 years accounted for 90% of those hospitalizations. 2 Magaziner and colleagues 3 estimate that 1 year after hip fracture the mortality rate is as high as 24%, only 40% of patients can independently perform their activities of daily living, and only 54% can walk unaided. In addition to the morbidity, the estimated cost of caring for hip fractures occurring in the United States each year exceeds $11 billion in 2002 dollars. 4 Because of the projected increase in the numbers of elderly Americans, the number of hip fractures is expected to exceed 500000 annually by the year 2040. 5 The American Academy of Orthopedic Surgeons declared that poor coordination among providers is the greatest factor compromising quality of care for patients with hip fracture. 6 A purported strength of the hospitalist model is that physicians who specialize in inpatient medicine may be more adept at coordinating complicated inpatient episodes of care. 7 Recent studies 8-10 of hospitalist models suggest promising results with respect to mortality and length of stay for medical patients. Currently, hospitalists are delivering more perioperative care, 11 but their usefulness in improving outcomes for surgical patients has been evaluated by only a single study. 12 Beginning July 1, 2001, patients 65 years and older having surgical repair of hip fracture at our institution were medically managed by hospitalists. Time to surgery, time from surgery to dismissal, length of stay, and inpatient complications were compared before and after this practice change. We hypothesized that patients cared for by hospitalists would have decreased time to surgery, time from surgery to dismissal, and length of stay. METHODS PATIENT SELECTION AND INTERVENTION To examine the effects of this model, we studied consecutive patients admitted for surgical 796

repair of hip fracture 1 year prior to and subsequent to the practice change. We used the Mayo Clinic surgical index to identify patients admitted between July 1, 2000, and June 30, 2002, with operating room procedure codes (International Classification of Diseases, Ninth Revision 13 ) matching at least 1 of 11 hip surgery codes. This list was cross-referenced with indication for surgery to identify the primary operative diagnoses of hip fracture. Patients were ineligible if they were transferred to the study hospital 72 or more hours after being admitted to a different facility. We identified a total of 466 patients who met the eligibility criteria. Patients admitted between July 1, 2000, and June 30, 2001, were included in the standard group (236 [51%] of 466). Patients admitted between July 1, 2001, and June 30, 2002, were included in the hospitalist group (230 [49%] of 466). In the standard group, patients with hip fracture were triaged by the emergency department physician to either a teaching orthopedic surgery service or a teaching medical service (creating a combination of 23 different admitting services geographically based on 10 different patient care units). This decision was based on the presence of significant concurrent medical problems. Throughout the hospitalization, laboratory evaluations, other tests, and consultations were ordered at the discretion of the admitting service staff. The surgery team provided all routine surgical care. In the hospitalist group, patients with hip fracture were admitted by the teaching orthopedic surgery service and comanaged by a hospitalist service. The hospitalist service was staffed at any given time by 1 physician, 2 allied health practitioners (nurse practitioners or physician assistants), and no residents. During the study period, 12 hospitalists and 12 allied health practitioners participated in patient care. The hospitalist performed the preoperative examination in the emergency department or when the patient arrived on the surgical floor. Medical conditions that warranted further investigation, including subspecialty consultations, were evaluated at the discretion of the hospitalist team. The hospitalist team managed all medical needs of the patient, including writing daily notes and medical orders and obtaining any other diagnostic studies that were indicated. The hospitalist team edited the medical components of the electronic dismissal summary and communicated with the patient s referring medical physicians. This model is similar to one that we previously studied, and details have been published elsewhere. 12 The hospitalist group used a census cap and did not take new patients when the census reached 20 patients. If this service was capped, a patient was triaged to the surgical service with recommendations to obtain medical consultation or to a teaching medical service. In either case, the patients were not cared for by the hospitalist group. Of the 230 patients who met eligibility requirements, 23 (10%) were not cared for by the hospitalist service. Although these 23 patients were treated by primary orthopedic surgery services or other medical services, they were included in the hospitalist group in an intent-to-treat approach. Each patient included in the study provided authorization to use their medical records for research, 14 and the study was approved in advance by the Mayo Institutional Review Board. DATA COLLECTION AND MANAGEMENT Medical records were abstracted for patient demographic data, mechanism and type of fracture, date and time of admission and surgery, American Society of Anesthesia (ASA) classification, 15 comorbidities, admission clinical data, medications, inpatient complications, and readmission rates. All data were manually abstracted by study nurses using a single case report form. The primary investigator (M.P.P.) audited approximately 10% of records for accuracy purposes and adjudicated questions concerning documentation of inpatient complications. When necessary, the investigator was blinded to the patient s cohort status. Inpatient complications were based on objective criteria or when documented in the clinical record. Only patients readmitted to the study hospital within 30 days of dismissal were counted as having a 30-day readmission. Time to surgery was defined as the time in hours from admission to the ward to the time the surgery began. Time from surgery to dismissal was defined as the time in days from the start of surgery to the time of dismissal. Length of stay was defined as the time in days from admission to the time of dismissal. STATISTICAL ANALYSIS We examined differences in baseline health and demographic characteristics of the patients in the standard and hospitalist groups using 2 test for discrete nominal variables and 2-sample t or rank sum test for continuous variables. We tested for unadjusted differences in time to surgery, time from surgery to dismissal, and length of stay using 2-sample t or rank sum test and 2 test for unadjusted differences in inpatient mortality, complications, and 30-day readmission rates. We assessed the effect of the hospitalist group (yes or no) on the entire cohort (N=466) for the outcomes of time to surgery and surgery to dismissal after adjusting for a priori variables that might have influenced these outcomes. These were entered into univariate linear regression and proportional hazard regression models, respectively. Significant variables from these regression analyses were included as candidate variables in stepwise and backward selection multivariable models. The selection models were validated using the bootstrap method. 16 To acquire more detailed information on the hospitalist and standard groups, we examined the effects of the chosen variables on time to surgery and surgery to dismissal within the 2 groups using similar multivariable models. All analyses were performed using statistical software (SAS, version 8.2; SAS Institute Inc, Cary, NC). RESULTS PATIENT CHARACTERISTICS Comparison of baseline characteristics between the 2 groups is shown in Table 1. Admission signs and symptoms did not significantly differ between the 2 groups, except for admission hypoxia, which was more common in the hospitalist group (26 [11.3%] of 230 vs 13 [5.5%] of 236; P=.02). LENGTH OF STAY, 30-DAY READMISSION RATE, TIME TO SURGERY, AND TIME FROM SURGERY TO DISMISSAL The mean overall length of stay was 2.2 days shorter in the hospitalist group (8.4 vs 10.6 days; P.001). Despite the shorter length of stay, we found no statistical significance in 30-day readmission rates (20 [8.7%] of 230 in the hospitalist group vs 25 [10.6%] of 236 in the standard group; P=.49; Table 2). The mean time to surgery (25 vs 38 hours; P.001) and time from surgery to dismissal (7 vs 9 days; P=.04) were significantly shorter 797

Table 1. Characteristics of Standard and Hospitalist s* Characteristic Standard (N = 236) Hospitalist (N = 230) P Value Age, mean, y 82 83.34 Female 171 (72.5) 163 (70.9).70 Male 65 (27.5) 67 (29.1) Fracture type Intertrochanteric 118 (50.0) 112 (48.7) Femoral neck 118 (50.0) 118 (51.3).78 Mechanism of fracture Fall 219 (92.8) 212 (92.2) Trauma 1 (0.4) 3 (1.3) Pathologic 7 (3.0) 6 (2.6).82 Unknown 9 (3.8) 7 (3.9) ASA score 15 1-2 33 (14.0) 23 (10.0) 3 166 (70.3) 166 (72.2).38 4 37 (15.7) 41 (17.8) Comorbidity Diabetes 45 (19.1) 46 (20.0).80 CHF 41 (17.4) 49 (21.3).28 CAD 69 (29.2) 77 (33.5).32 Dementia 54 (22.9) 62 (27.0).31 COPD 36 (15.3) 38 (16.5).71 Renal insufficiency 17 (7.2) 17 (7.4).94 CVA or TIA 36 (15.3) 50 (21.7).07 Residence at admission Home 149 (63.1) 138 (60.0) Assisted living 32 (13.6) 42 (18.3).38 Nursing home 55 (23.3) 50 (21.7) Ambulatory status Independent 114 (48.3) 89 (38.7) Assistive device 99 (42.0) 115 (50.0) Personal help 9 (3.8) 16 (7.0).14 Transfer to bed or chair 9 (3.8) 7 (3.0) Nonambulatory 5 (2.1) 3 (1.3) Symptoms at admission Pulmonary edema 37 (15.7) 29 (12.6).34 Hypoxia 13 (5.5) 26 (11.3).02 Hypotension 4 (1.7) 3 (1.3).99 Tachycardia 19 (8.1) 25 (10.9).30 Medicines at admission Warfarin (Coumadin) 26 (11.0) 32 (13.9).34 Acetylsalicylic acid (aspirin) 98 (41.5) 103 (44.8).48 Day of week at admission Monday to Thursday 139 (58.9) 135 (58.7) Friday to Sunday 97 (41.1) 95 (41.3).96 Time of day at admission 6 AM to 6 PM 105 (44.5) 101 (43.9) 6 PM to 6 AM 131 (55.5) 129 (56.1).90 Dismissal location Home with relatives or home 24 (10.5) 13 (5.9) health care Nursing home 196 (86.0) 192 (87.3).07 Another hospital or hospice 8 (3.5) 15 (6.8) Abbreviations: ASA, American Society of Anesthesia; CAD, coronary artery disease; CHF, congestive heart failure; COPD, chronic obstructive pulmonary disease; CVA, cardiovascular accident; TIA, transient ischemic attack. *Data are presented as number (percentage) unless indicated otherwise. We excluded 18 inpatient deaths from this analysis. in the hospitalist group. In addition, more patients in the hospitalist group went to surgery within 24 hours after admission (167 [72.6%] of 230 vs 112 [47.5%] of 236; P.001). Table 2. Time to Surgery, Surgery to Dismissal, and Length of Stay for the Standard and Hospitalist s* Variable Standard TIME TO SURGERY Hospitalist After adjustment, ASA scores 3 and 4 were associated with a 26.2- and 45.3-hour increase in time to surgery, respectively (P.001), while older age (P=.01) and fall as the mechanism of fracture (P.001) were associated with shorter time to surgery for the entire cohort (N=466). Patients in the hospitalist group went to surgery an average of 13.8 hours earlier than patients in the standard group (P=.002; Table 3). Adjusting for time of admission had no significant impact on this effect. When the time to surgery was analyzed within each group, we found that, in the standard group (n=236), patients admitted on Friday through Sunday went to surgery 19.1 hours later than those admitted Monday through Thursday (P.001). An ASA score of 3 or 4 in the standard group increased the time to surgery by 19.0 (P=.03) and 46.5 (P.001) hours, respectively. The presence of hypotension or hypoxia on admission in the standard group similarly increased the time to surgery by 47.6 (P=.03) and 44.2 (P.001) hours, respectively. In the hospitalist group, no variable was significantly associated with increased time to surgery, while fall as the mechanism of fracture was associated with a 47.4-hour decrease in time to surgery (P.001). SURGERY TO DISMISSAL P Value Time to surgery, h 38 (47) [26] 25 (53) [16].001 Surgery to discharge, d 9 (8) [6] 7 (5) [6].04 Length of stay, d 10.6 (9) [8] 8.4 (6) [7].001 No. (%) with 30-d readmission 25 (10.6) 20 (8.7).49 *Data are presented as mean (SD) [median] unless indicated otherwise. Table 3. Multivariate Linear Regression Analysis of Time to Surgery in Entire Cohort Variable Estimated Regression Coefficient, h P Value Age 0.7.01 ASA score 15 3 26.2.001 4 45.3.001 Fall as mechanism of fracture 30.1.001 Management by hospitalist service 13.8.002 Abbreviation: ASA, American Society of Anesthesia. Proportional hazard regression analysis for the entire cohort (N=466) identified that admission to the intensive care unit (ICU) was associated with a decreased chance of earlier dismissal (P.001), while delirium was associated with an increased chance of earlier dismissal (P.001; Table 4). After adjusting for these variables, 798

Table 4. Proportional Hazards Regression of Time From Surgery to Dismissal Variable patients managed by the hospitalist group had an increased chance of earlier dismissal (P.001). INPATIENT COMPLICATIONS There were no differences in inpatient deaths between the 2 groups (P=.59). There were no differences regarding inpatient complications, except that the diagnosis of delirium was made more frequently in the hospitalist group (P.001; Table 5). COMMENT Hazard Ratio (95% Confidence Interval) P Value Admission to the intensive care unit 0.7 (0.6-0.8).001 Delirium 1.4 (1.2-1.7).001 Management by hospitalist service 1.2 (1.1-1.4).001 Previous studies 8-10,17,18 have demonstrated that hospitalist models reduce the length of stay for medical patients. The results in our study parallel the results of previous studies in a surgical patient population. Elderly patients admitted for surgical repair of hip fracture went to surgery faster, were dismissed sooner after surgery, and had decreased overall length of stay after the implementation of a hospitalist model. Receiving care by the hospitalist group was an independent predictor of decreased time to surgery and an increased chance of earlier dismissal after surgery. There were no significant differences in 30-day readmission rates, inpatient deaths, or complications, except for delirium, which was diagnosed more often in the hospitalist group. Improved coordination of care by the hospitalist group and decreased patient care variability may explain part of these results. Variability in the admission process was simplified when the possibility of having a patient triaged to 1 of 23 inpatient services and multiple patient care units was eliminated. Concentrating these patients to 1 service (with a core group of medical personnel) and geographically placing them in 2 patient care units decreased variability and improved coordination of care by the hospitalist team, thus increasing efficiency. Meltzer and colleagues 9 reported that, as hospitalists gained disease-specific experience with certain medical conditions, length of stay and mortality rates decreased. The positive effects of the hospitalist model in this study may be partly due to the experience the hospitalists gained by focusing their practice on one type of patient population. As hospitalists treated more patients with hip fracture, their accumulated experience helped them determine when a patient s medical condition warranted going to surgery or being dismissed sooner. Table 5. Inpatient Complications* Complication Standard (n = 236) Hospitalist (n = 230) P Value Major Death 8 (3.4) 10 (4.4).59 Respiratory failure 8 (3.4) 9 (3.9).76 Pulmonary edema 2 (0.9) 1 (0.4).99 MI 12 (5.1) 12 (5.2).95 Renal failure 7 (3.0) 5 (2.2).59 Intermediate Pneumonia 29 (12.3) 33 (14.4).51 CHF 13 (5.5) 21 (9.2).13 Unstable angina 2 (0.9) 8 (3.5).06 Atrial fibrillation 17 (7.2) 20 (8.7).55 Acute central nervous 2 (0.9) 5 (2.2).28 system event (eg, TIA or CVA) Delirium 42 (17.8) 74 (32.2).001 DVT 2 (0.9) 3 (1.3).63 Wound infection 3 (1.3) 7 (3.0).22 Minor Urinary tract infection 40 (17.0) 47 (20.4).33 Falls 10 (4.2) 10 (4.4).95 Cellulitis 2 (0.9) 1 (0.4).99 Fracture 1 (0.4) 1 (0.4).99 New cancer 3 (1.3) 3 (1.3).99 Abbreviations: CHF, congestive heart failure; CVA, cardiovascular accident; DVT, deep venous thrombosis; MI, myocardial infarction; TIA, transient ischemic attack. *Data are presented as number (percentage) of patients unless indicated otherwise. Changing the medical management of these patients from a resident teaching service to a nonresident service may have facilitated quicker decision making regarding time to surgery and dismissal by physicians. The hospitalist and standard groups used similar methods regarding patient evaluations and care; however, we were not able to collect data regarding the total time it took to complete specific processes of care (ie, preoperative evaluation). This information may have been helpful in determining where the actual decrease in time to surgery or earlier dismissal between the 2 groups occurred. When the entire cohort was examined, neither day of the week nor time of admission was significantly associated with time to surgery. However, individual analysis of the standard and hospitalist groups found that a weekend admission in the standard group was associated with a 19.1-hour increase in time to surgery. This variable had no effect on time to surgery within the hospitalist group, suggesting that a fundamental difference existed in the coordination of care of patients admitted over the weekend prior to the implementation of the hospitalist model. Although patients in the hospitalist group went to surgery sooner, the hospitalist service did not influence the priority of surgical scheduling. These patients were evaluated and treated by the orthopedic trauma service. Surgical scheduling was based on the availability of the surgeon and operating room and priority of other pending surgical cases. Stay in the ICU was associated with decreased chance of earlier dismissal. This is not surprising since a patient needing this level of care is probably sicker and might 799

require longer stabilization before being dismissed. Fewer patients in the hospitalist group had an ICU admission compared with the standard group. Because fewer patients in the hospitalist group had an ICU stay, this could be another reason why patients in the hospitalist group had an increased chance of earlier dismissal after surgery. An unexpected finding was that the diagnosis of delirium was more frequent in the hospitalist group and that delirium was associated with an increased chance of earlier dismissal after surgery. Delirium was recorded only if a physician documented it in the medical record. Therefore, the larger proportion of patients diagnosed as having delirium in the hospitalist group may be explained by more conscientious recognition and recording of delirium. The observed proportion of delirium in the hospitalist group is supported by previous studies 19,20 reporting the frequency of delirium in elderly patients with medical conditions and hip fracture. Despite more reports of delirium in the hospitalist group, we found that regardless of group, delirium was significantly associated with an increased chance of earlier dismissal after surgery. When living status at admission and diagnosis of delirium were evaluated we found that patients living at home alone were less likely to have a diagnosis of delirium compared with patients living at home with family or in an assisted living center or nursing home. Because the patients in the latter group likely had a social system of posthospital care already available, they may have been more likely to be dismissed sooner than patients who required initial social service contacts and extended family support. Fundamental limitations of our study design should be noted. Although we used regression analysis to control for observable patient characteristics thought to influence outcomes, unobserved factors may have influenced these outcomes independent of the implementation of the hospitalist model. To minimize this possibility, we chose concurrent years of practice. During this period, we did not identify significant changes in the process of care for these patients. Despite the differences in inpatient outcomes, the type of service might not change long-term patient-oriented outcomes. Recent studies 8,9 suggest that there is a learning curve after adopting the hospitalist model and that the outcomes, such as length of stay, mortality, and costs, are not significantly different until the second year. Because this study did not capture all cost-related information, we did not answer the question of whether the shortened length of stay and any improvement in inpatient outcomes outweigh the cost of the hospitalist model. Future studies of this model would ideally extend the analysis of inpatient outcomes, incorporate long-term outcomes, and review the cost-related information. The comparison of a teaching standard model to a nonresident hospitalist model in our study may not extend to other institutions with nonteaching surgical services or teaching hospitalist services. Furthermore, our facility may have a different composition of local and referred patients than other facilities. To the extent that referral bias affects the outcomes, our results may not be generalizable. However, since nearly all the reported literature on patients undergoing surgery for hip fracture comes from referral centers, and patients with hip fracture are rarely referred to long-distance referral centers, our data should be comparable to other studies evaluating this patient population. This is a unique intervention at a single institution. We concede that only a few centers may dedicate a hospitalist service to manage one type of problem (hip fracture). However, if hospitals or academic centers are aiming to improve the management of patients with hip fracture at their centers, they might consider dedicating a core group of hospitalists to medically manage these patients. If this is not feasible or desirable, institutions may consider soliciting other motivated inpatient medical personnel to construct a team to incorporate and promote consistent management of these patients. Also, institutions might consider initially reviewing the admission process (how patients are allocated to services and the service s geographic location), weekend admissions (timing of appropriate medical consultation and following surgery), and social service expertise (reviewing efficiencies or delays) as ways to possibly improve outcomes. In conclusion, we observed that elderly patients undergoing surgical repair of hip fracture had a significantly shorter time to surgery, time from surgery to discharge, and overall length of stay after the implementation of a hospitalist model. Receiving care by the hospitalist group was an independent predictor of shorter time to surgery and an increased chance of earlier dismissal after surgery. These findings may be a reflection of improved efficiency by hospitalists, accumulated disease-specific experience by hospitalists, and decreased variability in the admission process. Further studies that evaluate the long-term and financial outcomes of hospitalists performing perioperative care are warranted. Accepted for Publication: November 16, 2004. Correspondence: Jeanne M. Huddleston, MD, Mayo Clinic, 200 First St SW, Rochester, MN 55906 (huddleston.jeanne@mayo.edu). Previous Presentation: Presented as a poster at the Society of Hospital Medicine Seventh Annual Meeting; April 20, 2004; New Orleans, La. Acknowledgment: We acknowledge the work of Donna K. Lawson and her role in data collection and management. REFERENCES 1. Melton LJ III. Adverse outcomes of osteoporotic fractures in the general population. J Bone Miner Res. 2003;18:1139-1141. 2. Popovic JR, Kozak LJ. National hospital discharge survey: annual summary, 1998. Vital Health Stat 13. 2000;148:1-194. 3. Magaziner J, Simonsick EM, Kashner TM, Hebel JR, Kenzora JE. Predictors of functional recovery one year following hospital discharge for hip fracture: a prospective study. J Gerontol. 1990;45:M101-M107. 4. Ray NF, Chan JK, Thamer M, Melton LJ III. Medical expenditures for the treatment of osteoporotic fractures in the United States in 1995: report from the National Osteoporosis Foundation. J Bone Miner Res. 1997;12:24-35. 5. Cooper C, Campion G, Melton LJ III. Hip fractures in the elderly: a world-wide projection. Osteoporos Int. 1992;2:285-289. 6. Dorn B, Bowen J, Downes E, et al. National Consensus Conference on Improv- 800

ing the Continuum of Care for Patients With Hip Fracture. Washington, DC: American Academy of Orthopedic Surgeons; 2001. 7. Wachter RM, Goldman L. The emerging role of hospitalists in the American health care system. N Engl J Med. 1996;335:514-517. 8. Auerbach AD, Wachter RM, Katz P, Showstack J, Baron RB, Goldman L. Implementation of a voluntary hospitalist service at a community teaching hospital: improved clinical efficiency and patient outcomes. Ann Intern Med. 2002;137: 859-865. 9. Meltzer D, Manning WG, Morrison J, et al. Effects of physician experience on costs and outcomes on an academic general medicine service: results of a trial of hospitalists. Ann Intern Med. 2002;137:866-874. 10. Tenner PA, Dibrell H, Taylor RP. Improved survival with hospitalists in a pediatric intensive care unit. Crit Care Med. 2003;31:847-852. 11. Wachter RM. Documenting the value of a hospitalist program. Paper presented at: Society of Hospital Medicine Sixth Annual Meeting; April 1-2, 2003; San Diego, Calif. 12. Huddleston JM, Long KH, Naessens JM, et al; Hospitalist-Orthopedic Team Trial Investigators. Medical and surgical comanagement after elective hip and knee arthroplasty: a randomized, controlled trial. Ann Intern Med. 2004;141: 28-38. 13. World Health Organization. International Classification of Diseases, Ninth Revision (ICD-9). Geneva, Switzerland: World Health Organization; 1977. 14. Melton LJ III. The threat to medical-records research. N Engl J Med. 1997;337: 1466-1470. 15. Dripps R, Lamont A, Eckenhoff J. The role of anesthesia in surgical mortality. JAMA. 1961;178:261-266. 16. Sauerbrei W, Schumacher M. A bootstrap resampling procedure for model building: application to the Cox regression model. Stat Med. 1992;11:2093-2109. 17. Diamond HS, Goldberg E, Janosky JE. The effect of full-time faculty hospitalists on the efficiency of care at a community teaching hospital. Ann Intern Med. 1998; 129:197-203. 18. Wachter RM, Katz P, Showstack J, Bindman AB, Goldman L. Reorganizing an academic medical service: impact on cost, quality, patient satisfaction, and education. JAMA. 1998;279:1560-1565. 19. Gustafson Y, Berggren D, Brannstrom B, et al. Acute confusional states in elderly patients treated for femoral neck fracture. J Am Geriatr Soc. 1988;36: 525-530. 20. Rummans TA, Evans JM, Krahn LE, Fleming KC. Delirium in elderly patients: evaluation and management. Mayo Clin Proc. 1995;70:989-998. Announcement New Online Submission and Peer Review System. The Archives of Internal Medicine editorial office is now using an online manuscript submission and peer review system developed by ejournalpress that serves the needs of authors, reviewers, and editors. See http: //www.archinternmed.com for more detailed information. 801