James M. Naessens, ScD, MPH

Study Design and Setting

We conducted a retrospective observational cohort study of all patients undergoing elective primary TKA from January 1, 1996, to December 31, 2004, comparing outcomes of patients assigned to SOS units with those of patients assigned to NON units. Patients were admitted to Rochester Methodist Hospital, Mayo Clinic, a tertiary‐care primary surgical teaching hospital that has 794 beds and more than 15,000 admissions annually. There were 13 faculty orthopedic surgeons performing elective nontraumatic lower‐extremity joint procedures during the study period, each with orthopedic residents rotating as part of the patient care team.

Study Population

All patients at Mayo Clinic who had undergone a joint replacement were followed prospectively, and data were collected using standardized forms and protocols, the methodologies of which have been described previously.22 Follow‐up was greater than 95% complete. Using the joint registry, patients who had undergone a TKA were identified (n = 9798). Postoperative patients initially transferred from the postanesthesia care unit to a general care floor were included. We excluded patients who required urgent, revision, or bilateral arthroplasties; who had been treated at or transferred from another institution; and whose primary surgical indication was trauma or septic arthritis. Subjects admitted to the hospital on the day prior to the procedure and subjects initially transferred directly from the postanesthesia care unit to the intensive care unit (ICU) were excluded, including patients requiring immediate postoperative cardiac monitoring. All primary surgical interventions were performed between Monday and Friday. The study authors identified 5883 eligible patients.

Patient clinical and demographic data including surgical indication; age; sex; height and weight at surgery; and dates of admission, surgery, death, discharge, and last follow‐up were abstracted from the registry. Type of anesthesia (general, regional, combined), American Society of Anesthesiologists (ASA) physical status class, and date and time of ICU admission and discharge were abstracted from individual departmental databases. The Decision Support System (DSS) administrative database (Eclipsys, Boca Raton, FL) was utilized to abstract relevant clinical variables, including major comorbid conditions such as cancer, cerebrovascular disease, chronic pulmonary disease, congestive heart failure, dementia, diabetes, hemiplegia, HIV/AIDS, metastatic solid tumors, myocardial infarction, peripheral vascular disease, renal disease, rheumatologic disease, and ulcers. A composite Charlson comorbidity score was computed as previously described.23, 24 Administrative variables regarding patient encounters including inpatient stay variableslength of stay, costs, patient location, nursing care units, admission times, discharge disposition and datewere also obtained from the DSS database.

Variables and Definitions

Length of stay was defined as the number of days from time of admission for the surgical episode to time of discharge. All costs were based on a provider perspective using standardized 2005 costs based on inflation‐adjusted estimates as previously described.3, 25, 26 We assessed resource utilization among patients who received care on an SOS unit by determining length of stay and total, hospital, and physician costs for the specified surgical episode. We also assessed blood bank, ICU, laboratory, pharmacy, physical therapy, occupational therapy, respiratory therapy, radiology, and room‐and‐board costs. Blood bank costs consisted of the costs of storing, processing, and administering the transfusion. Surgical procedure, anesthesia, and preoperative service costs were excluded from our cost analyses, as our aim was to examine hospital flow and resource utilization from time of transfer from the postanesthesia care unit to hospital discharge in order to specifically examine the impact of an SOS unit. We compared unexpected ICU admissions and stays and the resources utilized of patients in these 2 groups.

State and federal death registries confirmed patient expiration and primary cause of death. In‐hospital mortality was defined as death during the same hospital admission as the indexed surgical episode. Thirty‐day mortality was defined as death occurring within 30 days of the surgical procedure. Readmission at 30 days was defined as any admission to our institutions within a 30‐day period whose purpose was possibly related to the initial surgical episode and not a result of an elective admission. A priori we were aware of the small number of these events in the elective joint population. Therefore, we combined inpatient 30‐day mortality, 30‐day reoperation, and 30‐day readmission rates as a composite endpoint.

Specialized Orthopedic Surgery Units

An SOS unit was defined as a general care nursing unit where patients receive all their postoperative care after elective TKA. Such a unit has a multidisciplinary staff that has orthopedic expertise. The differences between an SOS unit and a NON unit are described in Table 1. Bed availability at the time of discharge from the postanesthesia care unit was the exclusive factor for admission to this unit. Bed availability was dependent on staff availability or whether there was an excess number of operative cases. The number and severity of patient medical comorbidities or complications, the time of discharge from the postanesthesia care unit, and patient room preference had no impact on which unit patients were admitted to. Patients were allocated to the SOS group or the NON unit group according to their physical location the evening of admission. Monitored beds at this facility are solely located in the ICU, and neither SOS nor NON units have this capability. Any patient requiring a monitored bed at any time, regardless of the reason, would be transferred directly to the ICU. Daily rounds were performed on either unit by the primary orthopedic team. The need for either medical or pain service consultation was at the discretion of the primary orthopedic team and not dependent on the patient's physical location.

All data were subsequently combined into a single database to facilitate data analysis. We further excluded 44 patients because no cost information was available, 9 patients who had multiple joint replacements performed during the specified surgical hospitalization, 69 patients because they had not authorized their medical records to be used for the purposes of research; 163 patients admitted directly to the ICU, 63 patients admitted the day prior to surgery, and 1 patient whose billing data suggested an outpatient encounter. A final patient cohort of 5534 patients was in the analysis. With the observed sample size and the overall variability, our study had 80% power to detect a difference between the 2 groups as small as 0.22 days in length of stay and $761 in hospital costs. The study was approved by our institutional review board. All study patients had authorized the use of their medical records for the purposes of research. Funding was obtained through an intramurally sponsored Small Grants Program by the Division of General Internal Medicine, which had no impact on the design of the study, reporting, or decision to submit an article on the study for publication.

Statistical Analysis

The statistical analysis compared baseline health and demographic characteristics of the patients cared for on SOS units with those cared for on NON units using chi‐square tests for nominal factors and the 2‐sample Wilcoxon rank sum tests for continuous variables. We used the chi‐square test to test for unadjusted differences in sex, patient residence (local or referred), race, individual Charlson comorbid conditions, anesthesia type, admitting diagnosis, 30‐day readmission rate, and discharge location. The 2‐sample Wilcoxon rank sum test assessed unadjusted differences in length of stay, costs, age, ICU days of stay, number of reoperations, total Charlson score and ASA class. Thirty‐day mortality rates were tested using the Fisher exact test.

Differences between patients in SOS and NON units in length of stay (LOS) and costs were the study's primary outcomes. We adjusted for baseline and surgical covariates using generalized linear regression models for these outcomes. The effect of the nursing unit was based on regression coefficients for age, sex, ASA class, anesthesia type, Charlson comorbidities, and surgical year. Age was analyzed using 5 categories: <55; 55‐64; 70‐74; 54‐69, and >75 years, with 65‐69 years used as the reference group. Each Charlson comorbid condition was treated as an indicator variable. Indicator variables were also assigned to surgical year, with 2004 used as the reference. These variables were subsequently entered into the model to calculate the differences between patients on an SOS unit and those on a NON unit.

Our secondary outcomes included ICU utilization and 30‐day outcomes of mortality, reoperations, and readmissions. We then assessed the effect of treatment on the SOS unit using the entire cohort (n = 5534) for unplanned postoperative ICU stay (yes or no) and on our combined endpoint after adjusting for the variables listed previously, using logistic regression models. A P value < 0.05 was considered statistically significant. All analyses were performed using statistical software (SAS, version 9.1; SAS Institute Inc, Cary, NC).

Thirty‐Day Outcomes

We elected to combine 30‐day mortality, reoperations, and readmissions pertaining to the joint procedure as a composite endpoint and found no differences in outcomes between groups. These results suggest that these longer‐term patient‐specific outcomes are likely not related to the specialty nursing care. We used a 30‐day endpoint assuming that a longer period may have led to the inclusion of deaths that were not directly attributable to the surgical intervention. In addition, a previous study advocated using 30 days as an endpoint for follow‐up, as it adequately accounts for adverse events.35 Our institution is also a referral center; hence, we would likely be unable to capture all events if we were to use the standard 90‐day period used for payment for this procedure, as these data are not canvassed by the joint registry.

Discharge Disposition

NON unit patients tended to have a higher degree of home health arranged at discharge. The NON unit nursing staff cares for other nonorthopedic surgical patients daily and may transfer their patterns of care utilization to the orthopedic patients despite different postoperative needs. In addition, if NON unit nursing staff members care for TKA patients only intermittently, they may not have as clear a working understanding of the particular postoperative requirements of TKA patients and consequently request unnecessary home health services and general community resources. Alternatively, patients cared for on NON units may actually have needed more assistance and more services on discharge. Although purely speculative, patients cared for by dedicated orthopedic surgery staff may develop added confidence from the experience of the allied care staff and feel less of a need for postdismissal services.

Role of Hospitalists in Specialized Care Pathways

Hospitalists are known to improve efficiency without reducing patient satisfaction. Their role has been demonstrated in different patient populations.1, 2, 3638 In a study of hip fracture patients, a hospitalist care model demonstrated a reduction in length of stay and time to surgery, without compromising long‐term outcomes.4, 39 Utilizing a hospitalist/midlevel care provider team approach to reduce LOS in units with a static number of beds can possibly increase bed turnover and prevent triaging of patients onto NON units. This is but one example of how a medical‐surgical partnership can improve outcomes. However, in an era where cost‐effective and regulatory practices require optimal resource allocation, hospitalists are in a key position to foster quality improvement projects, promote patient safety measures, and enhance systems care delivery. Becoming involved in designing specialized clinical units, with an emphasis on a multidisciplinary care approach, and developing their relationships with hospital administrators and nursing staff should be among their priorities. The Society of Hospital Medicine has also been committed to the care of the elderly through its core competencies40 and the orthopedic population that will benefit from such process changes and care pathways. Hospital innovations such as the implementation of SOS‐type units not only for other medical‐surgical partnerships but also for site‐based units caring for geriatric patients can be top priorities for hospitalists.

Strengths and Applicability

Our results are important in that they can likely be applied to both large tertiary‐care centers and smaller community‐based centers that perform specialized orthopedic surgeries. Nurses on specialized orthopedic units are very familiar with this postoperative population and have developed expertise in the care of these patients. These experienced nurses can likely be found on orthopedic units in tertiary‐care centers or surgical units in smaller facilities. Furthermore, our results support the benefits of interdisciplinary advanced teamwork. When an interdisciplinary group of health care providers works together on a daily basis, certain habits and patterns inevitably develop that often are unplanned and may be difficult to measure. This enhanced patient flow may not occur if these patients are cared for by providers unfamiliar with each other's work patterns. The importance of optimized teamwork is not hospital‐size dependent. Only primary elective knee arthroplasties were included to minimize confounding bias by bilateral or revision surgeries or indications such as septic arthritis, which are known to lead to increased length of stay, costs and complications.41

Limitations

Our study has the limitations of its retrospective nonrandomized study design, and only a prospective, randomized investigation could definitively address our aims. By excluding sicker patients, such as those referred with complicated health issues or high‐risk patients who required admission in advance of the proposed surgery for monitoring of perioperative anticoagulation issues, our estimates of possible differences between our comparison groups may have been conservative. We are unaware of how these sicker patients would fare on either nursing unit. Furthermore, what occurs in the hospital setting may not only have an impact on the hospital stay but may also influence long‐term outcomes. This is impossible to assess with analysis of administrative databases.

We relied on the complete and accurate recording of data from various databases, depending on the validity of data entry and collection. With a large cohort of patients, any errors in documentation or abstraction would be expected to be similar in both groups. Furthermore, confounding variables such as patient comorbidities are extracted from administrative data sets whose personnel might not be as familiar with the medical aspects of patient care. We used linear and logistic regression analyses to account for known differences in baseline characteristics despite the sample sizes being proportionally larger in the SOS group. Although we attribute the shortened length of stay in the SOS group to the interdisciplinary team approach, we were unable to determine to what extent this was a result of nursing staff or discharge planning. By using administrative databases, we were unable to abstract the consensus time and date of discharge, when all hospital staff deemed the patient ready for discharge, and hence relied on the actual time of discharge, which can be heavily reliant on availability at skilled nursing facilities. In addition, it was unknown whether patients discharged from SOS units were, by matter of protocol, discharged earlier in the day. Nevertheless, this small difference in length of stay can improve patient flow by opening up postoperative patient beds. Furthermore, such data sets are unable to provide information on patient satisfaction or quality‐of‐life measures, both of which are important determinants in specialized care pathways.42 The patient population served by our institution is generally ethnically homogeneous, thereby limiting potential generalizations to tertiary‐care centers or geographical areas with a population similar to ours. Our study also was not intended as a formal cost‐effectiveness analysis; hence, the impact of possible startup costs to begin a similar nursing unit was not explored. Although differences in practice management can be considered a limitation of not only operative but also perioperative care, we neither expected nor encountered any significant or drastic alterations during the study period, and year of surgery was adjusted for in our analysis. However, prospective randomized controlled studies testing specific clinical pathways and practice‐related innovations are needed to better examine these outcomes.

Study Design and Setting

We conducted a retrospective observational cohort study of all patients undergoing elective primary TKA from January 1, 1996, to December 31, 2004, comparing outcomes of patients assigned to SOS units with those of patients assigned to NON units. Patients were admitted to Rochester Methodist Hospital, Mayo Clinic, a tertiary‐care primary surgical teaching hospital that has 794 beds and more than 15,000 admissions annually. There were 13 faculty orthopedic surgeons performing elective nontraumatic lower‐extremity joint procedures during the study period, each with orthopedic residents rotating as part of the patient care team.

Study Population

All patients at Mayo Clinic who had undergone a joint replacement were followed prospectively, and data were collected using standardized forms and protocols, the methodologies of which have been described previously.22 Follow‐up was greater than 95% complete. Using the joint registry, patients who had undergone a TKA were identified (n = 9798). Postoperative patients initially transferred from the postanesthesia care unit to a general care floor were included. We excluded patients who required urgent, revision, or bilateral arthroplasties; who had been treated at or transferred from another institution; and whose primary surgical indication was trauma or septic arthritis. Subjects admitted to the hospital on the day prior to the procedure and subjects initially transferred directly from the postanesthesia care unit to the intensive care unit (ICU) were excluded, including patients requiring immediate postoperative cardiac monitoring. All primary surgical interventions were performed between Monday and Friday. The study authors identified 5883 eligible patients.

Patient clinical and demographic data including surgical indication; age; sex; height and weight at surgery; and dates of admission, surgery, death, discharge, and last follow‐up were abstracted from the registry. Type of anesthesia (general, regional, combined), American Society of Anesthesiologists (ASA) physical status class, and date and time of ICU admission and discharge were abstracted from individual departmental databases. The Decision Support System (DSS) administrative database (Eclipsys, Boca Raton, FL) was utilized to abstract relevant clinical variables, including major comorbid conditions such as cancer, cerebrovascular disease, chronic pulmonary disease, congestive heart failure, dementia, diabetes, hemiplegia, HIV/AIDS, metastatic solid tumors, myocardial infarction, peripheral vascular disease, renal disease, rheumatologic disease, and ulcers. A composite Charlson comorbidity score was computed as previously described.23, 24 Administrative variables regarding patient encounters including inpatient stay variableslength of stay, costs, patient location, nursing care units, admission times, discharge disposition and datewere also obtained from the DSS database.

Variables and Definitions

Length of stay was defined as the number of days from time of admission for the surgical episode to time of discharge. All costs were based on a provider perspective using standardized 2005 costs based on inflation‐adjusted estimates as previously described.3, 25, 26 We assessed resource utilization among patients who received care on an SOS unit by determining length of stay and total, hospital, and physician costs for the specified surgical episode. We also assessed blood bank, ICU, laboratory, pharmacy, physical therapy, occupational therapy, respiratory therapy, radiology, and room‐and‐board costs. Blood bank costs consisted of the costs of storing, processing, and administering the transfusion. Surgical procedure, anesthesia, and preoperative service costs were excluded from our cost analyses, as our aim was to examine hospital flow and resource utilization from time of transfer from the postanesthesia care unit to hospital discharge in order to specifically examine the impact of an SOS unit. We compared unexpected ICU admissions and stays and the resources utilized of patients in these 2 groups.

State and federal death registries confirmed patient expiration and primary cause of death. In‐hospital mortality was defined as death during the same hospital admission as the indexed surgical episode. Thirty‐day mortality was defined as death occurring within 30 days of the surgical procedure. Readmission at 30 days was defined as any admission to our institutions within a 30‐day period whose purpose was possibly related to the initial surgical episode and not a result of an elective admission. A priori we were aware of the small number of these events in the elective joint population. Therefore, we combined inpatient 30‐day mortality, 30‐day reoperation, and 30‐day readmission rates as a composite endpoint.

Specialized Orthopedic Surgery Units

An SOS unit was defined as a general care nursing unit where patients receive all their postoperative care after elective TKA. Such a unit has a multidisciplinary staff that has orthopedic expertise. The differences between an SOS unit and a NON unit are described in Table 1. Bed availability at the time of discharge from the postanesthesia care unit was the exclusive factor for admission to this unit. Bed availability was dependent on staff availability or whether there was an excess number of operative cases. The number and severity of patient medical comorbidities or complications, the time of discharge from the postanesthesia care unit, and patient room preference had no impact on which unit patients were admitted to. Patients were allocated to the SOS group or the NON unit group according to their physical location the evening of admission. Monitored beds at this facility are solely located in the ICU, and neither SOS nor NON units have this capability. Any patient requiring a monitored bed at any time, regardless of the reason, would be transferred directly to the ICU. Daily rounds were performed on either unit by the primary orthopedic team. The need for either medical or pain service consultation was at the discretion of the primary orthopedic team and not dependent on the patient's physical location.

All data were subsequently combined into a single database to facilitate data analysis. We further excluded 44 patients because no cost information was available, 9 patients who had multiple joint replacements performed during the specified surgical hospitalization, 69 patients because they had not authorized their medical records to be used for the purposes of research; 163 patients admitted directly to the ICU, 63 patients admitted the day prior to surgery, and 1 patient whose billing data suggested an outpatient encounter. A final patient cohort of 5534 patients was in the analysis. With the observed sample size and the overall variability, our study had 80% power to detect a difference between the 2 groups as small as 0.22 days in length of stay and $761 in hospital costs. The study was approved by our institutional review board. All study patients had authorized the use of their medical records for the purposes of research. Funding was obtained through an intramurally sponsored Small Grants Program by the Division of General Internal Medicine, which had no impact on the design of the study, reporting, or decision to submit an article on the study for publication.

Statistical Analysis

The statistical analysis compared baseline health and demographic characteristics of the patients cared for on SOS units with those cared for on NON units using chi‐square tests for nominal factors and the 2‐sample Wilcoxon rank sum tests for continuous variables. We used the chi‐square test to test for unadjusted differences in sex, patient residence (local or referred), race, individual Charlson comorbid conditions, anesthesia type, admitting diagnosis, 30‐day readmission rate, and discharge location. The 2‐sample Wilcoxon rank sum test assessed unadjusted differences in length of stay, costs, age, ICU days of stay, number of reoperations, total Charlson score and ASA class. Thirty‐day mortality rates were tested using the Fisher exact test.

Differences between patients in SOS and NON units in length of stay (LOS) and costs were the study's primary outcomes. We adjusted for baseline and surgical covariates using generalized linear regression models for these outcomes. The effect of the nursing unit was based on regression coefficients for age, sex, ASA class, anesthesia type, Charlson comorbidities, and surgical year. Age was analyzed using 5 categories: <55; 55‐64; 70‐74; 54‐69, and >75 years, with 65‐69 years used as the reference group. Each Charlson comorbid condition was treated as an indicator variable. Indicator variables were also assigned to surgical year, with 2004 used as the reference. These variables were subsequently entered into the model to calculate the differences between patients on an SOS unit and those on a NON unit.

Our secondary outcomes included ICU utilization and 30‐day outcomes of mortality, reoperations, and readmissions. We then assessed the effect of treatment on the SOS unit using the entire cohort (n = 5534) for unplanned postoperative ICU stay (yes or no) and on our combined endpoint after adjusting for the variables listed previously, using logistic regression models. A P value < 0.05 was considered statistically significant. All analyses were performed using statistical software (SAS, version 9.1; SAS Institute Inc, Cary, NC).

Thirty‐Day Outcomes

We elected to combine 30‐day mortality, reoperations, and readmissions pertaining to the joint procedure as a composite endpoint and found no differences in outcomes between groups. These results suggest that these longer‐term patient‐specific outcomes are likely not related to the specialty nursing care. We used a 30‐day endpoint assuming that a longer period may have led to the inclusion of deaths that were not directly attributable to the surgical intervention. In addition, a previous study advocated using 30 days as an endpoint for follow‐up, as it adequately accounts for adverse events.35 Our institution is also a referral center; hence, we would likely be unable to capture all events if we were to use the standard 90‐day period used for payment for this procedure, as these data are not canvassed by the joint registry.

Discharge Disposition

NON unit patients tended to have a higher degree of home health arranged at discharge. The NON unit nursing staff cares for other nonorthopedic surgical patients daily and may transfer their patterns of care utilization to the orthopedic patients despite different postoperative needs. In addition, if NON unit nursing staff members care for TKA patients only intermittently, they may not have as clear a working understanding of the particular postoperative requirements of TKA patients and consequently request unnecessary home health services and general community resources. Alternatively, patients cared for on NON units may actually have needed more assistance and more services on discharge. Although purely speculative, patients cared for by dedicated orthopedic surgery staff may develop added confidence from the experience of the allied care staff and feel less of a need for postdismissal services.

Role of Hospitalists in Specialized Care Pathways

Hospitalists are known to improve efficiency without reducing patient satisfaction. Their role has been demonstrated in different patient populations.1, 2, 3638 In a study of hip fracture patients, a hospitalist care model demonstrated a reduction in length of stay and time to surgery, without compromising long‐term outcomes.4, 39 Utilizing a hospitalist/midlevel care provider team approach to reduce LOS in units with a static number of beds can possibly increase bed turnover and prevent triaging of patients onto NON units. This is but one example of how a medical‐surgical partnership can improve outcomes. However, in an era where cost‐effective and regulatory practices require optimal resource allocation, hospitalists are in a key position to foster quality improvement projects, promote patient safety measures, and enhance systems care delivery. Becoming involved in designing specialized clinical units, with an emphasis on a multidisciplinary care approach, and developing their relationships with hospital administrators and nursing staff should be among their priorities. The Society of Hospital Medicine has also been committed to the care of the elderly through its core competencies40 and the orthopedic population that will benefit from such process changes and care pathways. Hospital innovations such as the implementation of SOS‐type units not only for other medical‐surgical partnerships but also for site‐based units caring for geriatric patients can be top priorities for hospitalists.

Strengths and Applicability

Our results are important in that they can likely be applied to both large tertiary‐care centers and smaller community‐based centers that perform specialized orthopedic surgeries. Nurses on specialized orthopedic units are very familiar with this postoperative population and have developed expertise in the care of these patients. These experienced nurses can likely be found on orthopedic units in tertiary‐care centers or surgical units in smaller facilities. Furthermore, our results support the benefits of interdisciplinary advanced teamwork. When an interdisciplinary group of health care providers works together on a daily basis, certain habits and patterns inevitably develop that often are unplanned and may be difficult to measure. This enhanced patient flow may not occur if these patients are cared for by providers unfamiliar with each other's work patterns. The importance of optimized teamwork is not hospital‐size dependent. Only primary elective knee arthroplasties were included to minimize confounding bias by bilateral or revision surgeries or indications such as septic arthritis, which are known to lead to increased length of stay, costs and complications.41

Limitations

Our study has the limitations of its retrospective nonrandomized study design, and only a prospective, randomized investigation could definitively address our aims. By excluding sicker patients, such as those referred with complicated health issues or high‐risk patients who required admission in advance of the proposed surgery for monitoring of perioperative anticoagulation issues, our estimates of possible differences between our comparison groups may have been conservative. We are unaware of how these sicker patients would fare on either nursing unit. Furthermore, what occurs in the hospital setting may not only have an impact on the hospital stay but may also influence long‐term outcomes. This is impossible to assess with analysis of administrative databases.

We relied on the complete and accurate recording of data from various databases, depending on the validity of data entry and collection. With a large cohort of patients, any errors in documentation or abstraction would be expected to be similar in both groups. Furthermore, confounding variables such as patient comorbidities are extracted from administrative data sets whose personnel might not be as familiar with the medical aspects of patient care. We used linear and logistic regression analyses to account for known differences in baseline characteristics despite the sample sizes being proportionally larger in the SOS group. Although we attribute the shortened length of stay in the SOS group to the interdisciplinary team approach, we were unable to determine to what extent this was a result of nursing staff or discharge planning. By using administrative databases, we were unable to abstract the consensus time and date of discharge, when all hospital staff deemed the patient ready for discharge, and hence relied on the actual time of discharge, which can be heavily reliant on availability at skilled nursing facilities. In addition, it was unknown whether patients discharged from SOS units were, by matter of protocol, discharged earlier in the day. Nevertheless, this small difference in length of stay can improve patient flow by opening up postoperative patient beds. Furthermore, such data sets are unable to provide information on patient satisfaction or quality‐of‐life measures, both of which are important determinants in specialized care pathways.42 The patient population served by our institution is generally ethnically homogeneous, thereby limiting potential generalizations to tertiary‐care centers or geographical areas with a population similar to ours. Our study also was not intended as a formal cost‐effectiveness analysis; hence, the impact of possible startup costs to begin a similar nursing unit was not explored. Although differences in practice management can be considered a limitation of not only operative but also perioperative care, we neither expected nor encountered any significant or drastic alterations during the study period, and year of surgery was adjusted for in our analysis. However, prospective randomized controlled studies testing specific clinical pathways and practice‐related innovations are needed to better examine these outcomes.

Study Design and Setting

We conducted a retrospective observational cohort study of all patients undergoing elective primary TKA from January 1, 1996, to December 31, 2004, comparing outcomes of patients assigned to SOS units with those of patients assigned to NON units. Patients were admitted to Rochester Methodist Hospital, Mayo Clinic, a tertiary‐care primary surgical teaching hospital that has 794 beds and more than 15,000 admissions annually. There were 13 faculty orthopedic surgeons performing elective nontraumatic lower‐extremity joint procedures during the study period, each with orthopedic residents rotating as part of the patient care team.

Study Population

All patients at Mayo Clinic who had undergone a joint replacement were followed prospectively, and data were collected using standardized forms and protocols, the methodologies of which have been described previously.22 Follow‐up was greater than 95% complete. Using the joint registry, patients who had undergone a TKA were identified (n = 9798). Postoperative patients initially transferred from the postanesthesia care unit to a general care floor were included. We excluded patients who required urgent, revision, or bilateral arthroplasties; who had been treated at or transferred from another institution; and whose primary surgical indication was trauma or septic arthritis. Subjects admitted to the hospital on the day prior to the procedure and subjects initially transferred directly from the postanesthesia care unit to the intensive care unit (ICU) were excluded, including patients requiring immediate postoperative cardiac monitoring. All primary surgical interventions were performed between Monday and Friday. The study authors identified 5883 eligible patients.

Patient clinical and demographic data including surgical indication; age; sex; height and weight at surgery; and dates of admission, surgery, death, discharge, and last follow‐up were abstracted from the registry. Type of anesthesia (general, regional, combined), American Society of Anesthesiologists (ASA) physical status class, and date and time of ICU admission and discharge were abstracted from individual departmental databases. The Decision Support System (DSS) administrative database (Eclipsys, Boca Raton, FL) was utilized to abstract relevant clinical variables, including major comorbid conditions such as cancer, cerebrovascular disease, chronic pulmonary disease, congestive heart failure, dementia, diabetes, hemiplegia, HIV/AIDS, metastatic solid tumors, myocardial infarction, peripheral vascular disease, renal disease, rheumatologic disease, and ulcers. A composite Charlson comorbidity score was computed as previously described.23, 24 Administrative variables regarding patient encounters including inpatient stay variableslength of stay, costs, patient location, nursing care units, admission times, discharge disposition and datewere also obtained from the DSS database.

Variables and Definitions

Length of stay was defined as the number of days from time of admission for the surgical episode to time of discharge. All costs were based on a provider perspective using standardized 2005 costs based on inflation‐adjusted estimates as previously described.3, 25, 26 We assessed resource utilization among patients who received care on an SOS unit by determining length of stay and total, hospital, and physician costs for the specified surgical episode. We also assessed blood bank, ICU, laboratory, pharmacy, physical therapy, occupational therapy, respiratory therapy, radiology, and room‐and‐board costs. Blood bank costs consisted of the costs of storing, processing, and administering the transfusion. Surgical procedure, anesthesia, and preoperative service costs were excluded from our cost analyses, as our aim was to examine hospital flow and resource utilization from time of transfer from the postanesthesia care unit to hospital discharge in order to specifically examine the impact of an SOS unit. We compared unexpected ICU admissions and stays and the resources utilized of patients in these 2 groups.

State and federal death registries confirmed patient expiration and primary cause of death. In‐hospital mortality was defined as death during the same hospital admission as the indexed surgical episode. Thirty‐day mortality was defined as death occurring within 30 days of the surgical procedure. Readmission at 30 days was defined as any admission to our institutions within a 30‐day period whose purpose was possibly related to the initial surgical episode and not a result of an elective admission. A priori we were aware of the small number of these events in the elective joint population. Therefore, we combined inpatient 30‐day mortality, 30‐day reoperation, and 30‐day readmission rates as a composite endpoint.

Specialized Orthopedic Surgery Units

An SOS unit was defined as a general care nursing unit where patients receive all their postoperative care after elective TKA. Such a unit has a multidisciplinary staff that has orthopedic expertise. The differences between an SOS unit and a NON unit are described in Table 1. Bed availability at the time of discharge from the postanesthesia care unit was the exclusive factor for admission to this unit. Bed availability was dependent on staff availability or whether there was an excess number of operative cases. The number and severity of patient medical comorbidities or complications, the time of discharge from the postanesthesia care unit, and patient room preference had no impact on which unit patients were admitted to. Patients were allocated to the SOS group or the NON unit group according to their physical location the evening of admission. Monitored beds at this facility are solely located in the ICU, and neither SOS nor NON units have this capability. Any patient requiring a monitored bed at any time, regardless of the reason, would be transferred directly to the ICU. Daily rounds were performed on either unit by the primary orthopedic team. The need for either medical or pain service consultation was at the discretion of the primary orthopedic team and not dependent on the patient's physical location.

All data were subsequently combined into a single database to facilitate data analysis. We further excluded 44 patients because no cost information was available, 9 patients who had multiple joint replacements performed during the specified surgical hospitalization, 69 patients because they had not authorized their medical records to be used for the purposes of research; 163 patients admitted directly to the ICU, 63 patients admitted the day prior to surgery, and 1 patient whose billing data suggested an outpatient encounter. A final patient cohort of 5534 patients was in the analysis. With the observed sample size and the overall variability, our study had 80% power to detect a difference between the 2 groups as small as 0.22 days in length of stay and $761 in hospital costs. The study was approved by our institutional review board. All study patients had authorized the use of their medical records for the purposes of research. Funding was obtained through an intramurally sponsored Small Grants Program by the Division of General Internal Medicine, which had no impact on the design of the study, reporting, or decision to submit an article on the study for publication.

Statistical Analysis

The statistical analysis compared baseline health and demographic characteristics of the patients cared for on SOS units with those cared for on NON units using chi‐square tests for nominal factors and the 2‐sample Wilcoxon rank sum tests for continuous variables. We used the chi‐square test to test for unadjusted differences in sex, patient residence (local or referred), race, individual Charlson comorbid conditions, anesthesia type, admitting diagnosis, 30‐day readmission rate, and discharge location. The 2‐sample Wilcoxon rank sum test assessed unadjusted differences in length of stay, costs, age, ICU days of stay, number of reoperations, total Charlson score and ASA class. Thirty‐day mortality rates were tested using the Fisher exact test.

Differences between patients in SOS and NON units in length of stay (LOS) and costs were the study's primary outcomes. We adjusted for baseline and surgical covariates using generalized linear regression models for these outcomes. The effect of the nursing unit was based on regression coefficients for age, sex, ASA class, anesthesia type, Charlson comorbidities, and surgical year. Age was analyzed using 5 categories: <55; 55‐64; 70‐74; 54‐69, and >75 years, with 65‐69 years used as the reference group. Each Charlson comorbid condition was treated as an indicator variable. Indicator variables were also assigned to surgical year, with 2004 used as the reference. These variables were subsequently entered into the model to calculate the differences between patients on an SOS unit and those on a NON unit.

Our secondary outcomes included ICU utilization and 30‐day outcomes of mortality, reoperations, and readmissions. We then assessed the effect of treatment on the SOS unit using the entire cohort (n = 5534) for unplanned postoperative ICU stay (yes or no) and on our combined endpoint after adjusting for the variables listed previously, using logistic regression models. A P value < 0.05 was considered statistically significant. All analyses were performed using statistical software (SAS, version 9.1; SAS Institute Inc, Cary, NC).

Thirty‐Day Outcomes

We elected to combine 30‐day mortality, reoperations, and readmissions pertaining to the joint procedure as a composite endpoint and found no differences in outcomes between groups. These results suggest that these longer‐term patient‐specific outcomes are likely not related to the specialty nursing care. We used a 30‐day endpoint assuming that a longer period may have led to the inclusion of deaths that were not directly attributable to the surgical intervention. In addition, a previous study advocated using 30 days as an endpoint for follow‐up, as it adequately accounts for adverse events.35 Our institution is also a referral center; hence, we would likely be unable to capture all events if we were to use the standard 90‐day period used for payment for this procedure, as these data are not canvassed by the joint registry.

Discharge Disposition

NON unit patients tended to have a higher degree of home health arranged at discharge. The NON unit nursing staff cares for other nonorthopedic surgical patients daily and may transfer their patterns of care utilization to the orthopedic patients despite different postoperative needs. In addition, if NON unit nursing staff members care for TKA patients only intermittently, they may not have as clear a working understanding of the particular postoperative requirements of TKA patients and consequently request unnecessary home health services and general community resources. Alternatively, patients cared for on NON units may actually have needed more assistance and more services on discharge. Although purely speculative, patients cared for by dedicated orthopedic surgery staff may develop added confidence from the experience of the allied care staff and feel less of a need for postdismissal services.

Role of Hospitalists in Specialized Care Pathways

Hospitalists are known to improve efficiency without reducing patient satisfaction. Their role has been demonstrated in different patient populations.1, 2, 3638 In a study of hip fracture patients, a hospitalist care model demonstrated a reduction in length of stay and time to surgery, without compromising long‐term outcomes.4, 39 Utilizing a hospitalist/midlevel care provider team approach to reduce LOS in units with a static number of beds can possibly increase bed turnover and prevent triaging of patients onto NON units. This is but one example of how a medical‐surgical partnership can improve outcomes. However, in an era where cost‐effective and regulatory practices require optimal resource allocation, hospitalists are in a key position to foster quality improvement projects, promote patient safety measures, and enhance systems care delivery. Becoming involved in designing specialized clinical units, with an emphasis on a multidisciplinary care approach, and developing their relationships with hospital administrators and nursing staff should be among their priorities. The Society of Hospital Medicine has also been committed to the care of the elderly through its core competencies40 and the orthopedic population that will benefit from such process changes and care pathways. Hospital innovations such as the implementation of SOS‐type units not only for other medical‐surgical partnerships but also for site‐based units caring for geriatric patients can be top priorities for hospitalists.

Strengths and Applicability

Our results are important in that they can likely be applied to both large tertiary‐care centers and smaller community‐based centers that perform specialized orthopedic surgeries. Nurses on specialized orthopedic units are very familiar with this postoperative population and have developed expertise in the care of these patients. These experienced nurses can likely be found on orthopedic units in tertiary‐care centers or surgical units in smaller facilities. Furthermore, our results support the benefits of interdisciplinary advanced teamwork. When an interdisciplinary group of health care providers works together on a daily basis, certain habits and patterns inevitably develop that often are unplanned and may be difficult to measure. This enhanced patient flow may not occur if these patients are cared for by providers unfamiliar with each other's work patterns. The importance of optimized teamwork is not hospital‐size dependent. Only primary elective knee arthroplasties were included to minimize confounding bias by bilateral or revision surgeries or indications such as septic arthritis, which are known to lead to increased length of stay, costs and complications.41

Limitations

Our study has the limitations of its retrospective nonrandomized study design, and only a prospective, randomized investigation could definitively address our aims. By excluding sicker patients, such as those referred with complicated health issues or high‐risk patients who required admission in advance of the proposed surgery for monitoring of perioperative anticoagulation issues, our estimates of possible differences between our comparison groups may have been conservative. We are unaware of how these sicker patients would fare on either nursing unit. Furthermore, what occurs in the hospital setting may not only have an impact on the hospital stay but may also influence long‐term outcomes. This is impossible to assess with analysis of administrative databases.

We relied on the complete and accurate recording of data from various databases, depending on the validity of data entry and collection. With a large cohort of patients, any errors in documentation or abstraction would be expected to be similar in both groups. Furthermore, confounding variables such as patient comorbidities are extracted from administrative data sets whose personnel might not be as familiar with the medical aspects of patient care. We used linear and logistic regression analyses to account for known differences in baseline characteristics despite the sample sizes being proportionally larger in the SOS group. Although we attribute the shortened length of stay in the SOS group to the interdisciplinary team approach, we were unable to determine to what extent this was a result of nursing staff or discharge planning. By using administrative databases, we were unable to abstract the consensus time and date of discharge, when all hospital staff deemed the patient ready for discharge, and hence relied on the actual time of discharge, which can be heavily reliant on availability at skilled nursing facilities. In addition, it was unknown whether patients discharged from SOS units were, by matter of protocol, discharged earlier in the day. Nevertheless, this small difference in length of stay can improve patient flow by opening up postoperative patient beds. Furthermore, such data sets are unable to provide information on patient satisfaction or quality‐of‐life measures, both of which are important determinants in specialized care pathways.42 The patient population served by our institution is generally ethnically homogeneous, thereby limiting potential generalizations to tertiary‐care centers or geographical areas with a population similar to ours. Our study also was not intended as a formal cost‐effectiveness analysis; hence, the impact of possible startup costs to begin a similar nursing unit was not explored. Although differences in practice management can be considered a limitation of not only operative but also perioperative care, we neither expected nor encountered any significant or drastic alterations during the study period, and year of surgery was adjusted for in our analysis. However, prospective randomized controlled studies testing specific clinical pathways and practice‐related innovations are needed to better examine these outcomes.