Free
Perioperative Medicine  |   March 2014
Inpatient Falls after Total Knee Arthroplasty: The Role of Anesthesia Type and Peripheral Nerve Blocks
Stavros G. Memtsoudis, M.D., Ph.D., F.C.C.P.; Thomas Danninger, M.D.; Rehana Rasul, M.P.H., M.A.; Jashvant Poeran, M.D., Ph.D.; Philipp Gerner, B.S.; et al Ottokar Stundner, M.D.; Edward R. Mariano, M.D., M.A.S.; Madhu Mazumdar, Ph.D., M.A., M.S.
Author Affiliations & Notes
  • Stavros G. Memtsoudis, M.D., Ph.D., F.C.C.P.
    From the Department of Anesthesiology, Hospital for Special Surgery, New York, New York (S.G.M., T.D., and O.S.); Department of Public Health, Division of Biostatistics and Epidemiology, Weill Cornell Medical College, New York, New York (R.R., J.P., and M.M.); Stony Brook University, Stony Brook, New York (P.G.); and Department of Anesthesiology, Perioperative and Pain Medicine, Stanford University School of Medicine, Stanford, California (E.R.M.).
  • Thomas Danninger, M.D.
    From the Department of Anesthesiology, Hospital for Special Surgery, New York, New York (S.G.M., T.D., and O.S.); Department of Public Health, Division of Biostatistics and Epidemiology, Weill Cornell Medical College, New York, New York (R.R., J.P., and M.M.); Stony Brook University, Stony Brook, New York (P.G.); and Department of Anesthesiology, Perioperative and Pain Medicine, Stanford University School of Medicine, Stanford, California (E.R.M.).
  • Rehana Rasul, M.P.H., M.A.
    From the Department of Anesthesiology, Hospital for Special Surgery, New York, New York (S.G.M., T.D., and O.S.); Department of Public Health, Division of Biostatistics and Epidemiology, Weill Cornell Medical College, New York, New York (R.R., J.P., and M.M.); Stony Brook University, Stony Brook, New York (P.G.); and Department of Anesthesiology, Perioperative and Pain Medicine, Stanford University School of Medicine, Stanford, California (E.R.M.).
  • Jashvant Poeran, M.D., Ph.D.
    From the Department of Anesthesiology, Hospital for Special Surgery, New York, New York (S.G.M., T.D., and O.S.); Department of Public Health, Division of Biostatistics and Epidemiology, Weill Cornell Medical College, New York, New York (R.R., J.P., and M.M.); Stony Brook University, Stony Brook, New York (P.G.); and Department of Anesthesiology, Perioperative and Pain Medicine, Stanford University School of Medicine, Stanford, California (E.R.M.).
  • Philipp Gerner, B.S.
    From the Department of Anesthesiology, Hospital for Special Surgery, New York, New York (S.G.M., T.D., and O.S.); Department of Public Health, Division of Biostatistics and Epidemiology, Weill Cornell Medical College, New York, New York (R.R., J.P., and M.M.); Stony Brook University, Stony Brook, New York (P.G.); and Department of Anesthesiology, Perioperative and Pain Medicine, Stanford University School of Medicine, Stanford, California (E.R.M.).
  • Ottokar Stundner, M.D.
    From the Department of Anesthesiology, Hospital for Special Surgery, New York, New York (S.G.M., T.D., and O.S.); Department of Public Health, Division of Biostatistics and Epidemiology, Weill Cornell Medical College, New York, New York (R.R., J.P., and M.M.); Stony Brook University, Stony Brook, New York (P.G.); and Department of Anesthesiology, Perioperative and Pain Medicine, Stanford University School of Medicine, Stanford, California (E.R.M.).
  • Edward R. Mariano, M.D., M.A.S.
    From the Department of Anesthesiology, Hospital for Special Surgery, New York, New York (S.G.M., T.D., and O.S.); Department of Public Health, Division of Biostatistics and Epidemiology, Weill Cornell Medical College, New York, New York (R.R., J.P., and M.M.); Stony Brook University, Stony Brook, New York (P.G.); and Department of Anesthesiology, Perioperative and Pain Medicine, Stanford University School of Medicine, Stanford, California (E.R.M.).
  • Madhu Mazumdar, Ph.D., M.A., M.S.
    From the Department of Anesthesiology, Hospital for Special Surgery, New York, New York (S.G.M., T.D., and O.S.); Department of Public Health, Division of Biostatistics and Epidemiology, Weill Cornell Medical College, New York, New York (R.R., J.P., and M.M.); Stony Brook University, Stony Brook, New York (P.G.); and Department of Anesthesiology, Perioperative and Pain Medicine, Stanford University School of Medicine, Stanford, California (E.R.M.).
  • Best of Meeting Abstracts submission originally presented at the American Society of Anesthesiologists Annual Meeting, October 2013. This article is featured in “This Month in Anesthesiology,” page 1A. Corresponding article on page 530.
    Best of Meeting Abstracts submission originally presented at the American Society of Anesthesiologists Annual Meeting, October 2013. This article is featured in “This Month in Anesthesiology,” page 1A. Corresponding article on page 530.×
  • Submitted for publication June 13, 2013. Accepted for publication November 27, 2013.
    Submitted for publication June 13, 2013. Accepted for publication November 27, 2013.×
  • Address correspondence to Dr. Memtsoudis: Department of Anesthesiology, Hospital for Special Surgery, 535 East 70th Street, New York, New York 10021. memtsoudiss@hss.edu. This article may be accessed for personal use at no charge through the Journal Web site, www.anesthesiology.org.
Article Information
Perioperative Medicine / Clinical Science / Regional Anesthesia
Perioperative Medicine   |   March 2014
Inpatient Falls after Total Knee Arthroplasty: The Role of Anesthesia Type and Peripheral Nerve Blocks
Anesthesiology 03 2014, Vol.120, 551-563. doi:https://doi.org/10.1097/ALN.0000000000000120
Anesthesiology 03 2014, Vol.120, 551-563. doi:https://doi.org/10.1097/ALN.0000000000000120
Abstract

Background:: Much controversy remains on the role of anesthesia technique and peripheral nerve blocks (PNBs) in inpatient falls (IFs) after orthopedic procedures. The aim of the study is to characterize cases of IFs, identify risk factors, and study the role of PNB and anesthesia technique in IF risk in total knee arthroplasty patients.

Methods:: The authors selected total knee arthroplasty patients from the national Premier Perspective database (Premier Inc., Charlotte, NC; 2006–2010; n = 191,570, >400 acute care hospitals). The primary outcome was IF. Patient- and healthcare system–related characteristics, anesthesia technique, and presence of PNB were determined for IF and non-IF patients. Independent risk factors for IFs were determined by using conventional and multilevel logistic regression.

Results:: Overall, IF incidence was 1.6% (n = 3,042). Distribution of anesthesia technique was 10.9% neuraxial, 12.9% combined neuraxial/general, and 76.2% general anesthesia. PNB was used in 12.1%. Patients suffering IFs were older (average age, 68.9 vs. 66.3 yr), had higher comorbidity burden (average Deyo index, 0.77 vs. 0.66), and had more major complications, including 30-day mortality (0.8 vs. 0.1%; all P < 0.001). Use of neuraxial anesthesia (IF incidence, 1.3%; n = 280) had lower adjusted odds of IF compared with adjusted odds of IF with the use of general anesthesia alone (IF incidence, 1.6%; n = 2,393): odds ratio, 0.70 (95% CI, 0.56–0.87). PNB was not significantly associated with IF (odds ratio, 0.85 [CI, 0.71–1.03]).

Conclusions:: This study identifies several risk factors for IF in total knee arthroplasty patients. Contrary to common concerns, no association was found between PNB and IF. Further studies should determine the role of anesthesia practices in the context of fall-prevention programs.

Review of more than 190,000 records from 400 hospitals in an administrative database showed an incidence of inpatient falls of 1.6% in this patient group, associated with morbidity and mortality. Peripheral nerve block did not alter the risk of inpatient fall, whereas use of neuraxial anesthesia reduced the risk by 30% compared with general anesthesia.

What We Already Know about This Topic

  • Inpatient falls after lower extremity total joint surgery are associated with significant morbidity and mortality

  • The use of peripheral nerve blockade has been speculated to contribute to the risk of inpatient falls in this patient group

What This Article Tells Us That Is New

  • Review of more than 190,000 records from 400 hospitals in an administrative database showed an incidence of inpatient falls of 1.6% in this patient group, associated with morbidity and mortality

  • Peripheral nerve block did not alter the risk of inpatient fall, whereas use of neuraxial anesthesia reduced the risk by 30% compared with general anesthesia

INPATIENT falls (IFs) bear the risk of severe complications and expose patients to potentially preventable injury. Particularly, patients undergoing a total knee arthroplasty (TKA) may be at risk because this procedure significantly limits their mobility in the perioperative period.1,2  Despite the recognition that IFs constitute a major problem and have been designated as a potentially preventable event by the Centers for Medicare and Medicaid,3,4  surprisingly little national research is available to help elucidate associated characteristics and risk factors for this adverse event in the context of TKA. Although we and others have previously attempted to study the extent of the problem, either using nationally representative databases or institutional data,1,2,5  the role of many potentially contributing factors, such as anesthesia-related variables, has not been captured in these evaluations and thus their impact remains unknown.
It has been suggested that anesthesia-related factors, especially the use of peripheral nerve blocks (PNBs), may contribute to the risk of IFs, by negatively affecting motor function.6–8  Identifying risk factors for IF is important not only to prevent patients from harm, complications, and severe related injuries but also to avert associated economic damages to patients and the healthcare system. In this context, the Centers for Medicare and Medicaid services have categorized IFs as hospital-acquired conditions and may not reimburse for related costs.4 
Therefore, we used a large, national database, previously used by our study group for the study of anesthesia-related outcomes,8–12  that allows for the capture of anesthesia-related procedures to: (1) better characterize patient- and healthcare system–related factors associated with IFs; (2) identify risk factors for this outcome in general; and (3) determine whether the type of anesthesia and use of PNB affect the risk for this event in TKA patients. We hypothesized that, among other factors, older and sicker patients would be at increased risk for IFs, and that the choice of anesthesia and the use of PNB will affect this risk.
Materials and Methods
Data Source and Study Population
Data provided by Premier Perspective Inc. (Charlotte, NC) for this study were collected between January 2006 and September 2010 from approximately 400 acute care hospitals in the United States. The database features information on a patient’s hospitalization which includes patient demographics, hospital characteristics, and complete billing information. In addition, International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes and Current Procedural Terminology codes are provided to determine specific information about diagnoses present during the hospitalization and procedures carried out. Data validity is assured through a standardized process before it is included in the database. Specifically, the validation process is made up of a seven-step integrity analysis, after which approximately 100 sampling and statistical validity and integrity assurance crosschecks are performed.01 Data used in this study were deidentified and thus compliant with the Health Insurance Portability and Accountability Act,02 and therefore this project was exempt from review by our Institutional Review Board (Hospital for Special Surgery, New York, NY). Specific hypotheses and primary outcomes were not evaluated by our Institutional Review Board for this study. The database was queried between February 22, 2013 and June 12, 2013, to collect the necessary data elements.
Study Sample
Unique cases from the database were included in the study if they underwent TKA (ICD-9 CM procedure code 81.54) and had information on type of anesthesia used during the procedure: general anesthesia, neuraxial anesthesia, or a combination of general and neuraxial anesthesia, which was identified using billing information. In addition, cases were restricted to routine admissions and elective procedures.
Study Variables
The primary outcome, IF, was defined by ICD-9-CM diagnosis code E849.7 indicative of “accidents occurring in a residential institution.” As there is no definitive standard in the reporting of IFs though ICD-9 codes, we used the IF coding as used in our hospital (Hospital for Special Surgery) and as described previsouly.2  By restricting our sample to routine admissions and nonemergent procedures only, we further attempted to logically exclude patients who fell in another institutionalized setting other than the hospital where the procedure was performed.
Patient demographic variables analyzed were age, sex, and race (white, black, Hispanic, and other). Healthcare-related variables were hospital location (urban or rural), hospital size (<300, 300–499, or ≥500 beds), hospital teaching status, and individual hospital identifier (deidentified to researchers). Procedure-related variables were type of anesthesia (general, neuraxial, or neuraxial/general), PNB use, type of knee arthroplasty (unilateral or bilateral), year of procedure (to account for trend), hospital costs, and length of hospitalization.
Type of anesthesia or use of PNB was defined using a list of all billing descriptions containing the search term “ANES.” This list (illustrated in appendix 1) was reviewed independently by one anesthesiologist (S.G.M.), one anesthesiology resident (T.D.), and a physician-epidemiologist (J.P.) to classify the type of anesthesia and, in addition, use of PNB. Further, the list was concatenated with Current Procedural Terminology codes indicating type of anesthesia (appendix 1 for list). Together, they provide a comprehensive definition of anesthesia usage. Only relevant and logic codes were included for analysis. Although we feel this is the best approach to identify anesthesia type in this claims-based dataset, there still are missing data for 28% of cases, which is only moderately higher than the missing rate (19%)13  on anesthesia type in the comprehensive National Anesthesia Clinical Outcomes Registry03 specifically designed to capture anesthesia-related information.
Comorbidity variables consisted of two comorbidity measures, that is, grouping according to Deyo–Charlson14  and Elixhauser,15  for which a selection of individual comorbidities was taken into account (obesity, [complicated] diabetes mellitus, congestive heart failure, chronic lung disease, renal failure, metastatic cancer, solid tumor without metastasis, coagulation deficiency, fluid and electrolyte abnormalities, alcohol abuse, drug abuse, psychosis, [complicated] hypertension, bloodloss anemia, and deficiency anemia). In addition, a diagnosis of sleep apnea was considered as this comorbidity was considered important but is not included in either index. Complication and outcome variables included major cardiac complications (acute myocardial infarction or other cardiac-related complications), major pulmonary complications (pulmonary embolism, pneumonia, or other pulmonary complications), deep venous thrombosis, cerebrovascular events, infections, acute renal failure, gastrointestinal complications, 30-day mortality, need for blood transfusion, mechanical ventilation, and critical care service usage. Comorbidities were identified using ICD-9 codes as previously reported.14,15 04 Complication variables were defined using ICD-9 and Current Procedural Terminology codes as listed in appendix 2. Other than missing data for type of anesthesia, no other missing data were present in our current dataset.
Statistics
Inpatient fall risk was quantified in terms of subgroup prevalence and odds ratios (ORs) in a multiple logistic regression model and a multilevel (random intercept) logistic regression model. We specifically did not choose for propensity score analysis because we do not believe it would inform our study more than our current approach because of several reasons put forward by others including risk of considerable sample size reduction and overall similar results in case of large sample sizes.16–18 
Univariable Analysis
Inpatient fall events were characterized by patient demographics, healthcare- and procedure-related variables, and comorbidity measures. Groups were compared using chi-square and t tests for categorical and continuous variables, respectively. Means and SDs were estimated for age, Deyo index, and length of hospital stay. Because hospital cost had a positively skewed distribution, its median and interquartile range were determined, and the Mann–Whitney rank sum test was used to evaluate significance. As Deyo index is a more compact measure (compared with the Elixhauser comorbidity grouping), we chose only to show this comorbidity measure in the univariable analysis.
Multiple Logistic Regression
To determine risk factors associated with IF, a multiple logistic regression model was fitted after the following sequential stages:
First, candidate covariates were chosen based on clinical judgment and significance of P value less than 0.15 in the univariable analyses. These covariates included all patient demographic variables, hospital identifier (hospital-fixed effects), all procedure-related variables, and a selection of comorbidity and complication variables (dementia, sleep apnea, blood transfusion, and individual Elixhauser comorbidities). We chose comorbidity grouping according to Elixhauser15  (instead of grouping according to Deyo–Charlson14 ) because they yielded slightly higher validity in sensitivity analyses. As previously described, we also considered interactions of blood transfusion with (deficiency) anemia.19 
Second, we achieved further variable selection through a nonparametric bootstrapping process on the model from the first stage.20  Specifically, 100 bootstrap samples of size n were randomly drawn with replacement from the original sample of size n. A stepwise procedure was applied to each sample using a forward selection method (with selection entry level of P = 0.20). Because a variable included in the model for most bootstrap samples is expected to have higher probability of being prognostically important, candidate covariates were selected if they were included in more than 70% of the 100 bootstrap sample models.20  Pairwise covariate combinations were evaluated for covariates which failed this 70% cutoff. If the frequency of pairwise combinations was greater than 90%, then the covariate with the higher frequency was selected.
Finally, the variables selected in the second stage were used to fit the final multiple logistic regression model. Because PNB was not selected in this process, we added this to the final model to assess its effect on IF.
Model Diagnostics
The optimism-corrected c-statistic (discrimination) and the Hosmer and Lemeshow test (calibration) were determined to assess the validity of the final model.21  A model with a c-statistic greater than 0.7 is indicative of good discriminatory power, that is, how well the model discriminates between observed data at different levels of the outcome. Calibration indicates the ability of a model to match predicted and observed data; a nonsignificant Hosmer and Lemeshow test value indicates a well-calibrated model.
Multilevel Logistic Regression
To additionally check the effect of anesthesia type and PNB on IF, and the identified risk factors for IF, we fitted a multilevel logistic regression model with random intercepts using the SAS GLIMMIX procedure (appendix 3).22  Multilevel logistic regression is a modification of conventional (single-level) logistic regression; it takes into account the multilevel structure of the Premier Perspective data, for example, procedures per hospital. The technique adjusts for clustering, for example, individuals within hospitals. A random intercept to account for the variation in each hospital was included in the model. Hospitals with less than 50 patients were excluded from this analysis, as previously recommended.23  To determine risk factors for IF, a forward selection procedure (α inclusion, P = 0.05) was used for variable selection from all covariates listed above, except individual hospital identifiers. All statistical analyses were performed using SAS software version 9.3 (SAS Institute, Cary, NC).
Disclosures.
As previous studies have shown significantly increased risks for IFs in cohorts of less than 2,000 patients, we assumed (before database query) our database to be sufficiently powered.6 
On the basis of reviewing the literature (meta-analyses using a mix of observational and interventional trials) evaluating the impact of interventions to prevent IFs, we determined that an alteration of at least 10 to 20% in the OR for IF (in comparison with previously reported IF rates)24,25  would be clinically significant.26–28 
Results
We identified 191,570 records for elective TKA which also had information on anesthesia type listed. Of these, 10.9% were performed under neuraxial, 12.9% under combined general/neuraxial, and 76.2% under general anesthesia, respectively. Of all patients, 12.1% received a PNB. In 1.6% (n = 3,042) of cases, an IF took place.
Table 1 illustrates patient demographics, healthcare- and procedure-related variables, and comorbidity measure variables for patients who suffered an IF versus those who did not. Patients whose course was complicated by an IF were on average older (mean age, 68.9 [SD = 10.3] vs. 66.3 [SD = 10.5] yr; P < 0.001). The incidence of IFs was higher among patients undergoing their surgery under general versus neuraxial or neuraxial/general anesthesia. (1.6% vs. 1.3% vs. 1.5%; P = 0.0018). The proportion of patients suffering an IF who received a PNB or not was not significantly different (1.58 vs. 1.62%; P = 0.6933). In addition, patients suffering an IF had a higher comorbidity burden (mean Deyo index, 0.77 [SD = 1.03] vs. 0.66 [SD = 0.97]; P < 0.001), which was also evident by the higher prevalence of individual comorbidities (table 2).
Table 1.
Patient Demographics, Healthcare-related, Procedure-related, and Comorbidity Measure Variables for Patients Subgrouped by Fall/No Fall
Patient Demographics, Healthcare-related, Procedure-related, and Comorbidity Measure Variables for Patients Subgrouped by Fall/No Fall×
Patient Demographics, Healthcare-related, Procedure-related, and Comorbidity Measure Variables for Patients Subgrouped by Fall/No Fall
Full Size  |  Slide (.ppt)
Table 1.
Patient Demographics, Healthcare-related, Procedure-related, and Comorbidity Measure Variables for Patients Subgrouped by Fall/No Fall
Patient Demographics, Healthcare-related, Procedure-related, and Comorbidity Measure Variables for Patients Subgrouped by Fall/No Fall×
Full Size  |  Slide (.ppt)
×
Table 2.
Prevalence of Selected Comorbidities by Fall/No Fall
Prevalence of Selected Comorbidities by Fall/No Fall×
Prevalence of Selected Comorbidities by Fall/No Fall
Full Size  |  Slide (.ppt)
Table 2.
Prevalence of Selected Comorbidities by Fall/No Fall
Prevalence of Selected Comorbidities by Fall/No Fall×
Full Size  |  Slide (.ppt)
×
Table 3 illustrates the rate of complications for patients who suffered an IF versus those who did not. IF patients had higher rates of major complications, including those affecting the cardiac and pulmonary system, 30-day mortality, and higher rates of usage of critical care services. Moreover, IF patients had a significantly increased length of stay (4.7 [SD = 3.2] vs. 3.5 [SD = 1.8] days; P < 0.001) and higher hospital costs ($17,070 [interquartile range, $13,588–$22,295] vs. $14,508 [interquartile range, $12,034, $17,998]; P < 0.0001); data not shown.
Table 3.
Prevalence of Selected Complications by Fall/No Fall
Prevalence of Selected Complications by Fall/No Fall×
Prevalence of Selected Complications by Fall/No Fall
Full Size  |  Slide (.ppt)
Table 3.
Prevalence of Selected Complications by Fall/No Fall
Prevalence of Selected Complications by Fall/No Fall×
Full Size  |  Slide (.ppt)
×
Table 4 shows patients characterized by use of a PNB. We did not find a difference for age, sex, or Deyo comorbidity burden when comparing patients receiving versus not receiving a PNB. However, patients of minority race received a PNB procedure less commonly.
Table 4.
Patient Demographics, Healthcare-related, Procedure-related, and Comorbidity Measure Variables for Patients with a PNB and Those without
Patient Demographics, Healthcare-related, Procedure-related, and Comorbidity Measure Variables for Patients with a PNB and Those without×
Patient Demographics, Healthcare-related, Procedure-related, and Comorbidity Measure Variables for Patients with a PNB and Those without
Full Size  |  Slide (.ppt)
Table 4.
Patient Demographics, Healthcare-related, Procedure-related, and Comorbidity Measure Variables for Patients with a PNB and Those without
Patient Demographics, Healthcare-related, Procedure-related, and Comorbidity Measure Variables for Patients with a PNB and Those without×
Full Size  |  Slide (.ppt)
×
Multiple Logistic Regression
The final multiple logistic regression model (table 5) did not include the use of PNB as a risk factor for IF as this variable did not reach the required predetermined level for inclusion. Year of procedure and hospital identifiers are not shown in table 5, but both were significant additions to the model (P = 0.015 and P < 0.001, respectively). Advanced age, male sex, and presence of individual comorbidities (fluid and electrolyte abnormalities, psychosis, sleep apnea, obesity, coagulopathy, and bloodloss anemia) were associated with increased odds for IF.
Table 5.
Final Multiple Logistic Regression Model (Adjusted for Year of Procedure and Hospital-fixed Effects through the Hospital Identifier Variable) Displaying OR and 95% CI
Final Multiple Logistic Regression Model (Adjusted for Year of Procedure and Hospital-fixed Effects through the Hospital Identifier Variable) Displaying OR and 95% CI×
Final Multiple Logistic Regression Model (Adjusted for Year of Procedure and Hospital-fixed Effects through the Hospital Identifier Variable) Displaying OR and 95% CI
Full Size  |  Slide (.ppt)
Table 5.
Final Multiple Logistic Regression Model (Adjusted for Year of Procedure and Hospital-fixed Effects through the Hospital Identifier Variable) Displaying OR and 95% CI
Final Multiple Logistic Regression Model (Adjusted for Year of Procedure and Hospital-fixed Effects through the Hospital Identifier Variable) Displaying OR and 95% CI×
Full Size  |  Slide (.ppt)
×
Patients with anemia (and no transfusion) and those receiving a transfusion (without anemia) both had an increased risk for IF (OR, 1.43 [CI, 1.28–1.59] and OR, 1.98 [CI, 1.77–2.21], respectively) compared with patients with no anemia and no transfusion. This risk remained increased for patients with anemia who were transfused.
The use of neuraxial anesthesia was associated with lower odds with regard to IFs compared with the odds with the use of general anesthesia alone (OR, 0.70 [CI, 0.56–0.87]; P < 0.0012). When added to the final model, the use of a PNB (OR, 0.85 [CI, 0.71–1.03; P = 0.0962]) did not alter the odds for IFs.
The optimism-corrected c-statistic associated with the model for the outcome of IFs was 0.85 and the P value for the Hosmer and Lemeshow test for the model was 0.91, indicating good model discrimination and calibration.
Multilevel Logistic Regression
Results of the multilevel (random intercept) logistic regression model are illustrated in appendix 3. This analysis included 190,758 (99.6% of study sample) observations from 273 hospitals. The multilevel model yielded an intercept variance of 2.90 (standard error, 0.35; P < 0.0001). The P value is the result of the test evaluating whether there is a variance of zero; in this case, there was none, suggesting that there are indeed unmeasured explanatory hospital-level variables. However, analogous to the conventional logistic regression, PNB was not a risk factor for IF, and neuraxial anesthesia showed lower risk for IF compared with general anesthesia (OR, 0.71 [CI, 0.57–0.87]; P = 0.0015). Except for bloodloss anemia, all risks found in the single-level conventional regression model were also found in the multilevel model, with similar magnitude and direction of ORs. Also, the c-statistic was similar: 0.85.
Discussion
This analysis of more than 190,000 patients undergoing TKA reveals an overall IF rate of 1.6%. As expected, IFs were associated with worse outcomes as evidenced by higher cardiac and pulmonary complications, 30-day mortality, and higher rates of usage of critical care services. Among other differences in demographic variables, the population suffering from IFs was older with a higher comorbidity burden. When analyzing IFs according to the type of anesthesia, neuraxial anesthesia had lower odds of IF compared with the odds in general anesthesia alone. The usage of PNB had no significant impact on the risk for IFs.
In this study, we found a higher IF prevalence than previously reported in studies using nationally representative and institutional data.1,2  In a previous analysis, we found an incidence of 0.85% for IFs and Mandl et al.1  reported an incidence of 0.9%.2  However, in the former study, we reported an increase in incidence over the 10 yr of observation from 0.4% in 1998 to 1.3% in 2007.2  Thus, these more recent data suggest a further increase in incidence. We can only speculate that this finding may be related to increased rates of reporting and/or increases in IFs due to an increasingly sicker patient population undergoing arthroplasty.29 
We identified several patient factors that were associated with higher odds for IFs. Patient suffering from an IF were on average older, thus suggesting that age-related factors such as reduced motor strength, impaired reflexes, and balance may play a role. Furthermore, male sex was associated with increased odds for IFs. This finding has previously been reported,2  but reasons have to remain speculative. However, men may be less likely to ask for help when ambulating or are willing to take more risk and overestimate their capabilities.30  Interestingly, this sex difference in fall risk has not been observed in nonsurgical populations.31,32 
We identified a number of comorbidities and conditions that were associated with an increased risk for IFs, including sleep apnea or psychosis. Changes in sensorium such as increased sensitivity to postoperative narcotics and daytime alertness associated with the former and various degrees of altered perceptions of surroundings with the latter are likely contributors. Recently published data even suggested an association between sleep apnea and postoperative delirium, thus providing insight into one potential mechanism.33,34  Anemia was found to be a contributing factor to IFs in this and several other studies.35,36  As previously described, also transfusion alone demonstrated an increased risk for adverse outcome, in this case IFs.19  From our retrospective dataset, however, no clinical inferences can be made as we do not have data on hemoglobin levels and actual transfusion triggers during the procedures.
When comparing the types of anesthesia used for TKA, neuraxial anesthesia was associated with a reduced risk for IFs. Recent studies have identified neuraxial anesthesia to be associated with reduced risks for many perioperative complications in the orthopedic population.8,37  However, no data on the risk for IFs were available from these analyses. The reduced risk for IFs may be associated with differential influence on the risk for postoperative cognitive function and delirium.38 
Importantly, the use of PNB had no significant influence on IFs in the studied population. This is in contrast to previously published data.39,40  For example, one study suggested that continuous lumbar plexus blockade was associated with four times greater relative risk of fall compared with groups receiving noncontinuous or no blockade.6  This information suggests that the type of PNB may play an important role in the propensity to cause motor weakness and thus increase the risk for falls.41  Although further research is needed to identify the optimal technique to balance adequate pain control with the risk for motor dysfunction, our data should provide encouragement to not shy away from the use of PNB. It should also be kept in mind that the choice of anesthesia type and the usage of the type of blocks can and should be viewed only as parts of a comprehensive fall-prevention program. Such programs have been instituted across many hospitals with great success.42–44 
The multilevel model in appendix 3 suggested unmeasured explanatory hospital-level variables, which was also indicated by fitting the final single-level model. By including hospital identifiers in the single-level model, we noticed it to influence the effect of anesthesia type on IF. When assessing the final model without including hospital identifiers, the associations between neuraxial versus general anesthesia were consistent with the results presented. However, the ORs of neuraxial/general versus general anesthesia showed a significantly protective association (OR, 0.86 [CI, 0.77–0.96]; P = 0.0073). By including hospital identifiers, the c-statistic greatly improved and results were more consistent with the multilevel model. The association between PNB and IF was consistently nonsignificant between all models. From this exercise, the question arises on what hospital-level factors in particular are influencing the effect of anesthesia type on IF risk, and how this differs between hospitals. As a post hoc exploration, we modeled a logistic regression to evaluate the crude (unadjusted) association between IF and type of anesthesia within each hospital using an interaction. Only hospitals with 10 or more IFs were included which yielded 87,359 patients from 78 hospitals. The interaction between anesthesia type and hospital identifier was significant (P < 0.001). There were lower odds for IF for patients with neuraxial versus general anesthesia in 85% of hospitals. However, the odds of neuraxial/general versus general anesthesia for IF widely varied by hospital; significantly reduced odds in 41% of hospitals. These exploratory results must remain inconclusive due to a bias toward selecting larger hospitals or hospitals with higher IF rates and also because only crude (unadjusted) estimates could be performed due to low IF frequencies. For every covariate to be taken into account, an increment of 10 IF cases would be needed, therefore, requiring an even stricter selection of hospitals. Thus, the hospital-level factors influencing the effect of anesthesia type on IF risk remain to be elucidated.
Our analysis is burdened by a number of limitations. The database used contains limited clinical information and thus some important factors cannot be taken into account. Furthermore, no causal relations can be established from our data source, and associations between identified risk factors and IFs have to remain speculative. In this context, from the data used, we cannot clearly determine which mechanism contributes to the lower risk of IFs in conjunction with type of anesthesia. With regard to information concerning PNB, specific details on the exact type of block, if it was a continuous or single-shot application and doses and type of local anesthetics used, are not readily discernible. Furthermore, we do not have information on the presence of IF prevention programs in the participating hospitals which are likely to influence results.
Furthermore, the lack of more detailed clinical data might have influenced our results, and there may be unobserved confounding. Moreover, IF patients seem to have a higher comorbidity burden as has been demonstrated for patients undergoing knee arthroplasty with general anesthesia.10  To some extent, the addition of hospital identifiers to the multiple logistic regression, and the multilevel regression analysis may have accounted for this in terms of unmeasured confounders at the hospital level. Neuraxial anesthesia remained a factor with a lower risk of IF, and PNB still was not associated with IF. Moreover, the same risk factors identified in the conventional logistic regression model were also seen in the multilevel model.
Some limitations are inherent to the analysis of secondary databases and are related to the use of ICD-9 codes for various outcomes and patient demographics, such as complications and comorbidities. Most importantly, this limitation refers to the ICD-9 definition of IFs. We chose to use the definition as used in our own facility and as reported previsouly.2  However, this definition may vary across hospitals.45  Moreover, one previous study has shown the positive predictive value of IFs determined by ICD-9 coding to be only 54%.24  This study, however, is burdened by a highly selective and local group of patients. In addition, not all codes used in the mentioned study are unanimously representative of falls (e.g., E887 “Fracture, cause unspecified”). As a sensitivity analysis, we studied the difference in the number of IFs using the definition from the aforementioned study (resulted in 243 extra IFs, 8%) and repeated the multivariate regression analysis. This approach yielded similar results regarding effects of anesthesia, PNB, and risk factors. An additional factor counteracting this limitation is that we do not expect variability in coding to be related to the type of anesthesia or PNB use. Moreover, the IF prevalence we observed in the current study (1.6%) is similar to recent studies using either ICD-9 codes (1.8%)24  or data from a central event reporting system (1.5%).25 
The general limitations regarding analysis of secondary databases and use of ICD-9 codes have been described extensively elsewhere. In particular, as with any observational analysis of a complex clinical outcome, unexplained variation remains, demonstrated by, for example, the c-statistic of 0.85. To minimize any untoward influence beyond the usual level of concern, we have used methodologies that have previously been either published or validated in this study.
In conclusion, in this study, we were able to identify independent risk factors for IFs in patients undergoing TKA including advanced age and increasing comorbidity burden. The presence of sleep apnea, psychosis, obesity, coagulopathy, electrolyte abnormalities, and anemia also increased the risk of IFs. The type of anesthesia may represent a modifiable risk factor and the use of neuraxial over general anesthesia may be considered in the context of a fall-prevention program. Contrary to some publications and beliefs, the use of PNB did not significantly alter the risk of IFs in the context of actual clinical practice as shown in this analysis. These data can be used not only to risk stratify patients but also to support the use of interventions to avoid this complication.
Acknowledgments
Dr. Memtsoudis is funded by the Anna Maria and Stephen Kellen Career Development Award, New York, New York. Contributions of Dr. Mazumdar, Dr. Poeran, and Mrs. Rasul on this project were supported in part by funds from the Clinical Translational Science Center, New York, New York.
Competing Interests
The authors declare no competing interests.
*Premier Inc., Premier Perspective Database. More information available at: https://www.premierinc.com/wps/portal/premierinc/public/transforminghealthcare/improvingperformance/servicesprograms/researchservices. Accessed October 10, 2013.
Premier Inc., Premier Perspective Database. More information available at: https://www.premierinc.com/wps/portal/premierinc/public/transforminghealthcare/improvingperformance/servicesprograms/researchservices. Accessed October 10, 2013.×
†U.S. Department of Health and Human Services: Summary of the HIPAA Privacy Rule. Available at: http://www.hhs.gov/ocr/privacy/hipaa/understanding/summary/privacysummary.pdf. Accessed October 10, 2013.
U.S. Department of Health and Human Services: Summary of the HIPAA Privacy Rule. Available at: http://www.hhs.gov/ocr/privacy/hipaa/understanding/summary/privacysummary.pdf. Accessed October 10, 2013.×
‡Anesthesia Quality Institute: National Anesthesia Clinical Outcomes Registry. More information available at: http://www.aqihq.org. Accessed October 10, 2013.
Anesthesia Quality Institute: National Anesthesia Clinical Outcomes Registry. More information available at: http://www.aqihq.org. Accessed October 10, 2013.×
§Healthcare Cost and Utilization Project (HCUP): HCUP comorbidity software Version 3.7; 2012. Available at: http://www.hcup-us.ahrq.gov/toolssoftware/comorbidity/comorbidity.jsp). Accessed October 10, 2013.
Healthcare Cost and Utilization Project (HCUP): HCUP comorbidity software Version 3.7; 2012. Available at: http://www.hcup-us.ahrq.gov/toolssoftware/comorbidity/comorbidity.jsp). Accessed October 10, 2013.×
References
Mandl, LA, Lyman, S, Quinlan, P, Bailey, T, Katz, J, Magid, SK Falls among patients who had elective orthopaedic surgery: A decade of experience from a musculoskeletal specialty hospital.. J Orthop Sports Phys Ther. (2013). 43 91–6 [Article] [PubMed]
Memtsoudis, SG, Dy, CJ, Ma, Y, Chiu, YL, Della Valle, AG, Mazumdar, M In-hospital patient falls after total joint arthroplasty: Incidence, demographics, and risk factors in the United States.. J Arthroplasty. (2012). 27 823–8.e1 [Article] [PubMed]
Mattie, AS, Webster, BL Centers for Medicare and Medicaid Services’ “never events”: An analysis and recommendations to hospitals.. Health Care Manag (Frederick). (2008). 27 338–49 [PubMed]
Adedeji, OM An evaluation of the centers for Medicare & Medicaid Services’ Hospital Acquired Conditions and Present on Admission Indicator Reporting program [Ph.D. thesis].. (2012). School of Public Health, The University of Texas
Fischer, ID, Krauss, MJ, Dunagan, WC, Birge, S, Hitcho, E, Johnson, S, Costantinou, E, Fraser, VJ Patterns and predictors of inpatient falls and fall-related injuries in a large academic hospital.. Infect Control Hosp Epidemiol. (2005). 26 822–7 [Article] [PubMed]
Johnson, RL, Kopp, SL, Hebl, JR, Erwin, PJ, Mantilla, CB Falls and major orthopaedic surgery with peripheral nerve blockade: A systematic review and meta-analysis.. Br J Anaesth. (2013). 110 518–28 [Article] [PubMed]
Kandasami, M, Kinninmonth, AW, Sarungi, M, Baines, J, Scott, NB Femoral nerve block for total knee replacement—A word of caution.. Knee. (2009). 16 98–100 [Article] [PubMed]
Stundner, O, Chiu, YL, Sun, X, Mazumdar, M, Fleischut, P, Poultsides, L, Gerner, P, Fritsch, G, Memtsoudis, SG Comparative perioperative outcomes associated with neuraxial versus general anesthesia for simultaneous bilateral total knee arthroplasty.. Reg Anesth Pain Med. (2012). 37 638–44 [Article] [PubMed]
Memtsoudis, SG, Stundner, O, Sun, X, Chiu, YL, Ma, Y, Fleischut, P, Kerr, GE, Girardi, FP, Walz, JMCritical care in patients undergoing lumbar spine fusion: A population-based study.. J Intensive Care Med. (2013). [Epub ahead of print]
Memtsoudis, SG, Sun, X, Chiu, YL, Stundner, O, Liu, SS, Banerjee, S, Mazumdar, M, Sharrock, NE Perioperative comparative effectiveness of anesthetic technique in orthopedic patients.. Anesthesiology. (2013). 118 1046–58 [Article] [PubMed]
Memtsoudis, SG, Stundner, O, Rasul, R, Sun, X, Chiu, YL, Fleischut, P, Danninger, T, Mazumdar, M Sleep apnea and total joint arthroplasty under various types of anesthesia: A population-based study of perioperative outcomes.. Reg Anesth Pain Med. (2013). 38 274–81 [Article] [PubMed]
Memtsoudis, SG, Sun, X, Chiu, YL, Nurok, M, Stundner, O, Pastores, SM, Mazumdar, M Utilization of critical care services among patients undergoing total hip and knee arthroplasty: Epidemiology and risk factors.. Anesthesiology. (2012). 117 107–16 [Article] [PubMed]
Fleischut, P, Gaber-Baylis, LK, Rasul, R, Faggiani, S, Mazumdar, M, Dutton, R, Memtsoudis, S Variability in anesthetic care for total knee arthroplasty in the United States.. 38th Annual Regional Anesthesia Meeting and Workshops May 4, 2013 Boston, MA A064
Deyo, RA, Cherkin, DC, Ciol, MA Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases.. J Clin Epidemiol. (1992). 45 613–9 [Article] [PubMed]
Elixhauser, A, Steiner, C, Harris, DR, Coffey, RM Comorbidity measures for use with administrative data.. Med Care. (1998). 36 8–27 [Article] [PubMed]
Cepeda, MS, Boston, R, Farrar, JT, Strom, BL Comparison of logistic regression versus propensity score when the number of events is low and there are multiple confounders.. Am J Epidemiol. (2003). 158 280–7 [Article] [PubMed]
Shah, BR, Laupacis, A, Hux, JE, Austin, PC Propensity score methods gave similar results to traditional regression modeling in observational studies: A systematic review.. J Clin Epidemiol. (2005). 58 550–9 [Article] [PubMed]
Winkelmayer, WC, Kurth, T Propensity scores: Help or hype?. Nephrol Dial Transplant. (2004). 19 1671–3 [Article] [PubMed]
Shander, A, Javidroozi, M, Ozawa, S, Hare, GM What is really dangerous: Anaemia or transfusion?. Br J Anaesth. (2011). 107suppl 1 i41–59 [Article] [PubMed]
Sauerbrei, W, Schumacher, M A bootstrap resampling procedure for model building: Application to the Cox regression model.. Stat Med. (1992). 11 2093–109 [Article] [PubMed]
Gonen, M Analyzing Receiver Operating Characteristic Curves with SAS. (2007). Cary SAS Publications
SAS Institute Inc., The GLIMMIX Procedure, SAS/STAT® 9.3 User’s Guide. (2011). Cary SAS Institute Inc. 2805–53
Moineddin, R, Matheson, FI, Glazier, RH A simulation study of sample size for multilevel logistic regression models.. BMC Med Res Methodol. (2007). 7 34 [Article] [PubMed]
Hougland, P, Nebeker, J, Pickard, S, Van Tuinen, M, Masheter, C, Elder, S, Williams, S, Xu, WHenriksen, K, Battles, JB, Keyes, MA, Grady, MLUsing ICD-9-CM codes in hospital claims data to detect adverse events in patient safety surveillance. Advances in Patient Safety: New Directions and Alternative Approaches (Vol. 1: Assessment). (2008). Rockville Agency for Healthcare Research and Quality 235–52. Edited by
Kolla, BP, Lovely, JK, Mansukhani, MP, Morgenthaler, TI Zolpidem is independently associated with increased risk of inpatient falls.. J Hosp Med. (2013). 8 1–6 [Article] [PubMed]
Coussement, J, De Paepe, L, Schwendimann, R, Denhaerynck, K, Dejaeger, E, Milisen, K Interventions for preventing falls in acute- and chronic-care hospitals: A systematic review and meta-analysis.. J Am Geriatr Soc. (2008). 56 29–36 [Article] [PubMed]
Oliver, D, Connelly, JB, Victor, CR, Shaw, FE, Whitehead, A, Genc, Y, Vanoli, A, Martin, FC, Gosney, MA Strategies to prevent falls and fractures in hospitals and care homes and effect of cognitive impairment: Systematic review and meta-analyses.. BMJ. (2007). 334 82 [Article] [PubMed]
DiBardino, D, Cohen, ER, Didwania, A Meta-analysis: Multidisciplinary fall prevention strategies in the acute care inpatient population.. J Hosp Med. (2012). 7 497–3 [Article] [PubMed]
Kirksey, M, Chiu, YL, Ma, Y, Della Valle, AG, Poultsides, L, Gerner, P, Memtsoudis, SG Trends in in-hospital major morbidity and mortality after total joint arthroplasty: United States 1998–2008.. Anesth Analg. (2012). 115 321–7 [Article] [PubMed]
Addis, ME, Mahalik, JR Men, masculinity, and the contexts of help seeking.. Am Psychol. (2003). 58 5–14 [Article] [PubMed]
Centers for Disease Control and Prevention (CDC), Self-reported falls and fall-related injuries among persons aged > or =65 years: United States 2006.. MMWR Morb Mortal Wkly Rep. (2008). 57 225–9 [PubMed]
Morrison, A, Fan, T, Sen, SS, Weisenfluh, L Epidemiology of falls and osteoporotic fractures: A systematic review.. Clinicoecon Outcomes Res. (2013). 5 9–18 [PubMed]
Flink, BJ, Rivelli, SK, Cox, EA, White, WD, Falcone, G, Vail, TP, Young, CC, Bolognesi, MP, Krystal, AD, Trzepacz, PT, Moon, RE, Kwatra, MM Obstructive sleep apnea and incidence of postoperative delirium after elective knee replacement in the nondemented elderly.. Anesthesiology. (2012). 116 788–96 [Article] [PubMed]
Lombardi, C, Rocchi, R, Montagna, P, Silani, V, Parati, G Obstructive sleep apnea syndrome: A cause of acute delirium.. J Clin Sleep Med. (2009). 5 569–70 [PubMed]
Dharmarajan, TS, Avula, S, Norkus, EP Anemia increases risk for falls in hospitalized older adults: An evaluation of falls in 362 hospitalized, ambulatory, long-term care, and community patients.. J Am Med Dir Assoc. (2007). 83 suppl 2 e9–15 [Article] [PubMed]
Guse, CE, Porinsky, R Risk factors associated with hospitalization for unintentional falls: Wisconsin hospital discharge data for patients aged 65 and over.. WMJ. (2003). 102 37–42 [PubMed]
Neuman, MD, Silber, JH, Elkassabany, NM, Ludwig, JM, Fleisher, LA Comparative effectiveness of regional versus general anesthesia for hip fracture surgery in adults.. Anesthesiology. (2012). 117 72–92 [Article] [PubMed]
Papaioannou, A, Fraidakis, O, Michaloudis, D, Balalis, C, Askitopoulou, H The impact of the type of anaesthesia on cognitive status and delirium during the first postoperative days in elderly patients.. Eur J Anaesthesiol. (2005). 22 492–9 [Article] [PubMed]
Ilfeld, BM, Duke, KB, Donohue, MC The association between lower extremity continuous peripheral nerve blocks and patient falls after knee and hip arthroplasty.. Anesth Analg. (2010). 111 1552–4 [Article] [PubMed]
Sharma, S, Iorio, R, Specht, LM, Davies-Lepie, S, Healy, WL Complications of femoral nerve block for total knee arthroplasty.. Clin Orthop Relat Res. (2010). 468 135–40 [Article] [PubMed]
Charous, MT, Madison, SJ, Suresh, PJ, Sandhu, NS, Loland, VJ, Mariano, ER, Donohue, MC, Dutton, PH, Ferguson, EJ, Ilfeld, BM Continuous femoral nerve blocks: Varying local anesthetic delivery method (bolus versus basal) to minimize quadriceps motor block while maintaining sensory block.. Anesthesiology. (2011). 115 774–81 [Article] [PubMed]
Cameron, ID, Murray, GR, Gillespie, LD, Robertson, MC, Hill, KD, Cumming, RG, Kerse, N Interventions for preventing falls in older people in nursing care facilities and hospitals.. Cochrane Database Syst Rev. (2010).  CD005465
Miake-Lye, IM, Hempel, S, Ganz, DA, Shekelle, PG Inpatient fall prevention programs as a patient safety strategy: A systematic review.. Ann Intern Med. (2013). 1585 Pt 2 390–6 [Article] [PubMed]
Oliver, D, Healey, F, Haines, TP Preventing falls and fall-related injuries in hospitals.. Clin Geriatr Med. (2010). 26 645–92 [Article] [PubMed]
Currie, LHughes, RGFall and injury prevention. Patient Safety and Quality: An Evidence-Based Handbook for Nurses. (2008). Rockville Agency for Healthcare Research and Quality 195–50. Edited by
List of All Billing Descriptions Containing the Search Term “ANES”; Classified into “General Anesthesia,” “Neuraxial Anesthesia,” “General and Neuraxial Anesthesia Combined,” or Neither
Image not available
Full Size  |  Slide (.ppt)
ICD-9-CM Diagnosis Codes for Major Complications and Outcomes
Image not available
Full Size  |  Slide (.ppt)
Results from the Multilevel Regression Model (Adjusted for Year of Procedure) with OR and 95% CI
Image not available
Full Size  |  Slide (.ppt)
Table 1.
Patient Demographics, Healthcare-related, Procedure-related, and Comorbidity Measure Variables for Patients Subgrouped by Fall/No Fall
Patient Demographics, Healthcare-related, Procedure-related, and Comorbidity Measure Variables for Patients Subgrouped by Fall/No Fall×
Patient Demographics, Healthcare-related, Procedure-related, and Comorbidity Measure Variables for Patients Subgrouped by Fall/No Fall
Full Size  |  Slide (.ppt)
Table 1.
Patient Demographics, Healthcare-related, Procedure-related, and Comorbidity Measure Variables for Patients Subgrouped by Fall/No Fall
Patient Demographics, Healthcare-related, Procedure-related, and Comorbidity Measure Variables for Patients Subgrouped by Fall/No Fall×
Full Size  |  Slide (.ppt)
×
Table 2.
Prevalence of Selected Comorbidities by Fall/No Fall
Prevalence of Selected Comorbidities by Fall/No Fall×
Prevalence of Selected Comorbidities by Fall/No Fall
Full Size  |  Slide (.ppt)
Table 2.
Prevalence of Selected Comorbidities by Fall/No Fall
Prevalence of Selected Comorbidities by Fall/No Fall×
Full Size  |  Slide (.ppt)
×
Table 3.
Prevalence of Selected Complications by Fall/No Fall
Prevalence of Selected Complications by Fall/No Fall×
Prevalence of Selected Complications by Fall/No Fall
Full Size  |  Slide (.ppt)
Table 3.
Prevalence of Selected Complications by Fall/No Fall
Prevalence of Selected Complications by Fall/No Fall×
Full Size  |  Slide (.ppt)
×
Table 4.
Patient Demographics, Healthcare-related, Procedure-related, and Comorbidity Measure Variables for Patients with a PNB and Those without
Patient Demographics, Healthcare-related, Procedure-related, and Comorbidity Measure Variables for Patients with a PNB and Those without×
Patient Demographics, Healthcare-related, Procedure-related, and Comorbidity Measure Variables for Patients with a PNB and Those without
Full Size  |  Slide (.ppt)
Table 4.
Patient Demographics, Healthcare-related, Procedure-related, and Comorbidity Measure Variables for Patients with a PNB and Those without
Patient Demographics, Healthcare-related, Procedure-related, and Comorbidity Measure Variables for Patients with a PNB and Those without×
Full Size  |  Slide (.ppt)
×
Table 5.
Final Multiple Logistic Regression Model (Adjusted for Year of Procedure and Hospital-fixed Effects through the Hospital Identifier Variable) Displaying OR and 95% CI
Final Multiple Logistic Regression Model (Adjusted for Year of Procedure and Hospital-fixed Effects through the Hospital Identifier Variable) Displaying OR and 95% CI×
Final Multiple Logistic Regression Model (Adjusted for Year of Procedure and Hospital-fixed Effects through the Hospital Identifier Variable) Displaying OR and 95% CI
Full Size  |  Slide (.ppt)
Table 5.
Final Multiple Logistic Regression Model (Adjusted for Year of Procedure and Hospital-fixed Effects through the Hospital Identifier Variable) Displaying OR and 95% CI
Final Multiple Logistic Regression Model (Adjusted for Year of Procedure and Hospital-fixed Effects through the Hospital Identifier Variable) Displaying OR and 95% CI×
Full Size  |  Slide (.ppt)
×
Image not available
Full Size  |  Slide (.ppt)
Image not available
Full Size  |  Slide (.ppt)
Image not available
Full Size  |  Slide (.ppt)