Share this article on:

The Effect of Adding Functional Classification to ASA Status for Predicting 30-Day Mortality

Visnjevac, Ognjen MD*†; Davari-Farid, Sina MD*†; Lee, Jun MSc*†; Pourafkari, Leili MD*†; Arora, Pradeep MBBS*†; Dosluoglu, Hasan H. MD*†; Nader, Nader D. MD, PhD*†

doi: 10.1213/ANE.0000000000000740
Patient Safety: Research Report

BACKGROUND: The functional capacity to perform the activities of daily living is identified as an independent predictor of perioperative mortality but is not formally incorporated in the American Society of Anesthesiologists (ASA) classification. Our primary objective was to assess whether functional capacity is an independent predictor of 30-day and long-term mortality in a general population and, if so, to define how it may formally be incorporated into the routine preoperative ASA classification assessment.

METHODS: This retrospective, observational cohort study was conducted using 1998 to 2009 data extracted from the Veterans Affairs Surgical Quality Improvement Program of Western New York, a perioperative prospectively maintained database. Mortality follow-up was performed for all records in 2013. This population-based sample included all patients undergoing any noncardiac surgery (n = 12,324). Each patient’s ASA class (assigned preoperatively) was appended with subclasses A or B, with A representing patients who were functionally independent and B representing partially or fully dependent patients. The primary outcome was all-cause mortality during the follow-up period. Secondary outcomes included 30-day postoperative complications and mortality. Multivariate logistic regression was used to identify independent risk factors for mortality.

RESULTS: The likelihood for mortality was significantly lower for A patients than B patients within each ASA class. The odds ratios for mortality for group A patients significantly favored survival over group B within each ASA class (0.14, 0.29, and 0.50, for ASA class II, III, and IV, respectively, each P < 0.0001). The odds ratio for mortality of IIB over IIIA patients was 1.92 (95% confidence interval [CI], 1.19–3.11; P = 0.01); 1.29 (95% CI, 1.04–1.60; P = 0.03) for IIIB over IVA patients; and 2.03 (95% CI, 0.99–4.12, P=0.11) for IVB over ASA V patients, despite each higher class carrying a greater disease burden, by definition. The area under the curve the receiver operator characteristic curve was 0.811 ± 0.010 for traditional ASA classification in predicting death within 30 days, which improved 4.7% to 0.848 ± 0.008 using the modified ASA classification, P < 0.00001.

CONCLUSIONS: Functional capacity was an independent predictor of mortality within each ASA class, indicating that it should be considered for incorporation into the routine preoperative evaluation. Functional dependence may be an indication for increasing a patient’s ASA class by 1 class-point to better reflect his or her perioperative risk, but prospective validation of these findings is recommended, as this is a preliminary study.

From the Department of *Anesthesiology and Surgery, University at Buffalo, Buffalo, New York.

Accepted for publication January 24, 2015.

Funding: Not funded.

The authors declare no conflicts of interest.

This report was presented, in part, at the International Anesthesia Research Society (IARS), May 17–20, 2014, Montréal, Canada.

Reprints will not be available from the authors.

Address correspondence to Nader D. Nader, MD, PhD, Departments of Anesthesiology and Surgery, University at Buffalo, 3495 Bailey Ave., Buffalo, NY 14215. Address e-mail to nnader@buffalo.edu.

The ASA 6-point physical status (PS) classification is a simple system based on clinical assessment only, without the additional need for laboratory and diagnostic tests or often-cumbersome mathematic formulas.1 It was initially proposed in 1941,2 modified in 1961, and validated as one of the most reliable predictors of mortality.3–13 A multitude of risk-assessment strategies and formulas have been described for surgical patients,5,6,14–18 but the ASA classification system remains the most common routine component of preoperative risk assessment of surgical patients worldwide and provides guidance to surgeons, anesthesiologists, intensivists, and other health care personnel concerning perioperative management and monitoring, including prediction of postoperative intensive care requirements.

In select patient populations (i.e., vascular surgery patients, very elderly patients), however, the ASA classification system has limited risk stratification utility because such patients are almost always confined to 1 of 2 classes (ASA grade III or IV) due to their propensity for having several comorbid conditions.19–21 Distinction of these patients by functional capacity independently predicts 30-day and long-term mortality and strongly suggests that this evaluation should be routinely performed in patients with moderate operative risk.19,20

Functional capacity is not formally incorporated into the definition of the ASA PS classification.1 Thus, we hypothesize that incorporating function capacity into the traditional ASA classification will better predict perioperative risk for both 30-day and long-term mortality after surgery.

Back to Top | Article Outline

METHODS

Study Design

The study protocol was reviewed and approved by the Institutional Review Board at the Buffalo Veterans Affairs Medical Center with waiver of informed consent. A total of 12,324 patients presented to the Veterans Affairs Western New York Healthcare System between 1998 and 2010 to undergo at least 1 noncardiac surgery. Long-term mortality and death within 30 days were the primary outcome variables, and postoperative complications were the secondary outcome variable. Data for each participant were prospectively entered into the Veterans Affairs Surgical Quality Improvement Program (VASQIP) database at the time of surgery. This database does not include ASA VI patients undergoing organ harvest. Each patient’s functional status, defined by one’s ability to independently perform all activities of daily living (ADL), was prospectively assessed by preoperative anesthesia personnel during the patient interview and then entered into the VASQIP database. In the VASQIP database, patients deemed to be functionally independent to perform their ADL were coded as 1, whereas patients who were partially dependent on others for at least 1 ADL were coded as 2, and those fully dependent on others for their ADL were coded as 3. For the purposes of this study, functionally independent patients were re-coded by adding a suffix of A to the ASA class, and those who were either partially or fully dependent were identified by adding the suffix B.

A board-certified anesthesiologist examined all patients before any operative intervention and determined each patient’s ASA class.1 A data collector blinded to A and B subgroup allocation documented mortality data at the time of follow-up in 2013. This Veterans Affairs computerized medical records system is thorough and allowed for reliable documentation of mortality data for all patients, including dates of death.22

By definition, there could not be any patients classified as ASA IB (a perfectly healthy patient not able to perform his or her ADL) or ASA VA (a moribund patient able to independently perform all ADL). Thus, ASA classes I and V were assessed and compared with other classes without A or B subgrouping. All other ASA classes and subclasses were compared with their immediately bordering counterparts to evaluate the hypotheses (i.e., ASA IIB patients were compared with IIA patients and IIIA patients).

Each patient’s ASA class, functional status, comorbidities, clinical and demographic characteristics, preoperative laboratory data, anesthesia type (local, regional, epidural/spinal, general), procedures performed, length of stay, and perioperative morbidity and mortality were incorporated into the data for analyses, then compared among groups. Board-certified surgeons performed all procedures. Information on the date of death was available for all patients in our series with the use of the computerized charting system of the Buffalo Veterans Affairs Hospital. Definitions of collected and analyzed data are available through a publication by the American College of Surgeons.23

Back to Top | Article Outline

Statistical Analysis

Data analysis was performed using NCSS version 2007 (NCSS, LLC, Kaysville, UT). Demographic characteristics and morbidities were reported using descriptive statistics. The Fisher exact test was used to compare categorical variables between the 2 groups (A and B) for binary data; χ2 analysis was used when there were ≥3 categories; and nonparametric Mann-Whitney U tests were used for continuous variables. Odds ratios (ORs) were reported with 95% confidence interval (CI) and P values <0.05 were considered statistically significant. Receiver operator characteristic (ROC) curves were plotted for traditional and proposed modified ASA classification in predicting long-term survival and death within 30 days. Area under the curve (AUC) for the ROC were calculated and used to express the predictive value of each model. Comparisons were made between the ROC curves using the paired (nonindependent) method described by DeLong et al.24 Univariate log-rank modeling was used to identify potential factors associated with survival.25 Significant variables selected from the univariate models were included in a stepwise multivariate logistic regression to evaluate their independent prognostic effects with the binary variable of all-cause mortality as the primary outcome. Because of significant covariance between ASA classification and the presence of comorbid medical conditions, ASA classes were deleted from multivariate logistic regression analysis. Logistic regression model was used to explore the relationship between risk factors and 30-day mortality rate. We randomly selected half of the events and nonevents to form a training set and used remaining data to form a validation set. Stepwise model selection method was used. We considered variables and first-order interactions. We applied our model to the validation set. The Hosmer-Lemeshow test was used to assess goodness of fit for the final model. ORs and the corresponding 95% CIs are estimated by logistic regression model and reported with their corresponding P values. Bonferroni correction was used to adjust for 3 comparisons to maintain a family-wise error rate at overall 0.05. Kaplan-Meier analysis was used to compare overall survival for each group.26 All P values are 2-sided with α level = 0.05.

Back to Top | Article Outline

RESULTS

There were 12,324 surgical patients in this population-based sample. Of these, 381 patients (3.1%) were classified as ASA I; 3705 (30.1%) as ASA II; 6513 (52.8%) as ASA III; 1632 (13.2%) as ASA IV; and 93 (0.8%) as ASA V. Majority of the patients (95.2%) were male, and the proportion of males in group B was slightly but significantly larger than the proportion of males in group A (Table 1).

Table 1

Table 1

Demographic, comorbidity, and preoperative laboratory data for groups A and B were reported in Tables 1 and 2. Comparisons showed that many factors differed between groups A and B, with group B patients generally presenting with more comorbidities (Table 2). Group B patients also had significantly more postoperative complications (Table 3). Thirty-day mortality, myocardial infarction, cardiac arrest, postoperative pneumonia, urinary tract infections, wound dehiscence, renal insufficiency, return to the operating room for additional intervention, and hospital length of stay were all significantly worse in the group B (Table 3).

Table 2

Table 2

Table 3

Table 3

Postoperative mortality was significantly higher in group B patients within each ASA class, when compared with their counterparts in group A, both in the early postoperative period and long term (Fig. 1). This relationship is re-emphasized in Table 4, with ORs for mortality from each ASA class A group significantly in favor of survival compared with the corresponding B group (0.14, 0.29, and 0.50, each with P < 0.0001; Table 4).

Table 4

Table 4

Figure 1

Figure 1

Group B patients in any one ASA class had significantly greater mortality than group A patients in the subsequent, sicker, ASA class. The OR for mortality of IIB patients compared with IIIA patients was 1.92 (95% CI, 1.19–3.11, P = 0.01); OR for mortality for IIIB patients over IVA patients was 1.29 (95% CI, 1.04–1.60, P = 0.03); and OR for mortality for IVB patients over ASA V patients was 2.03 (95% CI, 0.99–4.12, P = 0.11). The AUC for the ROC curve was 0.811 ± 0.010 for traditional ASA classification in predicting death within 30 days, which improved 4.6% to 0.848 ± 0.008 using the modified ASA classification, P < 0.00001 (Fig. 2).

Figure 2

Figure 2

Multivariate logistic regression analysis with the binary variable of all-cause mortality as the primary outcome was performed. With stepwise model selection, when the significance level for entering an explanatory variable into the model is 0.01, 4 covariates were included into the model. The OR and the corresponding 95% CIs are listed in Table 5. Of note, while many demographic, comorbidity, and preoperative laboratory factors differed between groups A and B, as depicted in Tables 1 and 2, few were found to be independently strong, statistically significant predictors of mortality after stepwise model selection for multivariate regression was performed (Table 5). AUC for this model using the training set was 0.871. The AUC for this model using the validation set was 0.833. This suggests that the model is stable.

Table 5

Table 5

Group B patients were 4.57 times more likely to experience mortality within 30 days (OR, 4.57; 95% CI, 2.24–9.35; P < 0.0001). Emergency surgery, albumin, and blood urea nitrogen were also identified in the stepwise model as significant predictors of mortality (Table 5). Risk of mortality decreased 43% with each gram-per-deciliter rise in albumin (OR, 0.568; 95% CI, 0.399–0.809; P = 0.0017) while emergency surgery presented a 2.48 times increased risk of mortality (OR, 2.481; 95% CI, 1.340–4.587; P = 0.0038). Appendix 1 provides a broader list of variables considered for multivariate analysis. Of note, history of cardiovascular risk factors (myocardial infarction, heart failure, cerebrovascular accidents, angina, coronary revascularization, renal insufficiency, and critical limb ischemia) was not found to be significant.

Back to Top | Article Outline

DISCUSSION

Based on our findings, adding functional capacity to traditional ASA classification improves its predictive value for short-term and long-term mortality and the occurrence of postoperative complications. Functionally independent patients were found to have fewer postoperative complications, including significantly less short-term mortality, when compared with their functionally dependent counterparts (1.8% vs 12.9%). Similarly, mortality was significantly greater for the functionally dependent patients throughout the entire follow-up period and the modest, but statistically significant, improvement in ROC suggests better prediction of mortality when adding functional status to the current ASA PS definitions.

To our knowledge, this is the first study to specifically evaluate whether there is evidence for introducing an evaluation preoperative functional capacity into the ASA classification for surgical patients. Data presented in Tables 1 to 3 demonstrate objective differences in disease states, laboratory values, and morbidities between functionally independent and functionally dependent individuals, thereby validating the choice of investigating functional status as a surrogate for a patient’s overall disease state. Our findings provide evidence that preoperative functional status is an independent predictor of postoperative mortality and should be considered for incorporation into the ASA classification of patients’ PS.

Although differences in mortality between dependent and independent patients were clear within each ASA class, it is important to emphasize that dependent patients in any one ASA class had significantly greater mortality than independent patients in the subsequent higher and sicker ASA class. This indicates that the likelihood of mortality for functionally dependent patients in any one class was not only greater than that for independent patients in the same ASA class but also greater than independent patients in the subsequent higher (sicker) class, despite the presumption that the higher class carries a greater disease burden by definition. Not only does this blur the lines defining ASA classes, it provides evidence to consider whether it would be appropriate to amend the official definition of ASA PS classification system to increase all functionally dependent patients’ ASA classes by 1 class-point to better predict perioperative risks.

The divergence in mortality between functionally dependent and independent patients was found in select patient populations before this study. Dependent octogenarian patients had 23% greater mortality than their independent peers at 30 days postoperatively, and a significant divergence in mortality continued for the entirety of that study’s follow-up period. It is important to note that data for the current study overlaps data for the 1049 ASA grade III octogenarian patients in the previous study,27 as both were assessed using the Western New York VASQIP database. Because the octogenarian data represent <10% of the sample analyzed in this study, concern for skewed outcomes in this study was minimal, as effects were expected to be meager. Functionally dependent vascular surgery patients were found to have a 130% increased risk of mortality at 30 days postoperatively,19 and 28% greater mortality than their independent counterparts at 12 months.20 The capacity to perform ADL significantly affects cognition, mental health, and quality of life.28,29 Given that functional capacity has been shown to be a reliable predictor of outcomes in this study and others,19,20 it may be reasonable to conclude that functional capacity may be an unsophisticated but summative representation of each patient’s multifactorial disease state, thereby potentially explaining why this variable independently and reliably predicts postoperative mortality with such a high degree of significance.

It is important to note that differences in the 30-day occurrence of adverse events between groups were not limited to mortality. The dependent patients had a greater proportion of myocardial infarctions, cardiac arrests, cerebrovascular accidents, pneumonia, urinary tract infections, renal insufficiency, wound infections, and failure to wean from mechanical ventilation.

There were baseline differences between independent and dependent patients, but only those variables subsequently found to be independent predictors of mortality by multivariate analysis were ultimately considered significant. Along with functional capacity, age was found to be a significant independent predictor of mortality by multivariate analysis (Appendix 1), a finding consistent with other studies.19–21 Despite the significance of age and the baseline age difference between independent and dependent groups of patients, functional dependency was found to carry a greater risk of mortality. Prothrombin time, serum albumin level, blood urea nitrogen, total bilirubin, a history of chronic obstructive pulmonary diseases, recent prior pneumonia, current use of tobacco, and disseminated cancer were also identified as independent contributing risk factors for postoperative mortality (Appendix 1). Interestingly, history of coronary revascularization, angina, heart failure, critical limb ischemia, renal failure, cerebrovascular accident, and a “Do Not Resuscitate” status were not independently associated with increased risks of postoperative mortality.

Without doubt, postoperative outcomes are dependent on a constellation of comorbidities, preoperative and intraoperative factors, and postoperative complications,5,6,14–18 but multifactorial formulas to predict these outcomes are often cumbersome and difficult to incorporate for routine preoperative risk assessment, thereby emphasizing the advantages of the simple and easy-to-use ASA classification system.13 Functional status assessment can be easily incorporated into the current ASA classification, as it can be assessed by patient interview alone and does not require any calculations, formulas, laboratory data, or diagnostic tests. This simplicity begets its utility when considering its adaptation to the current ASA PS definitions, but it must be recognized that functional status is not an all-encompassing predictor, as suggested by the statistically significant, but arguably modest, rise in ROC (0.811–0.848, P < 0.00001) and the multiple other significant covariates identified as independent predictors of mortality.

Despite prospective collection of data, a major weakness of this study stems from its retrospective nature, carrying with it the usual limitations of a retrospective study. A large sample size was a useful tool for minimizing some of these limitations. Because this is a single-center study that includes mostly men at a Veterans Affairs Medical Center, it raises questions regarding potential gender bias and population homogeneity and presents a major limitation for generalizing data from the veteran’s population to the general public. As this was an observational study, the possibility of residual and persistent confounding factors exists, despite the large sample size, plethora of variables and patient factors investigated, and use of multivariate statistical analyses. The precise proportions and contributions of biopsychosocial, genetic, and pathophysiologic factors that contribute to each patient’s functional status and, subsequently, to their perioperative risk are yet to be fully elucidated and represent a potential topic of future study. This preliminary study presents a clinically relevant risk stratification tool by way of functional status assessment, but further prospective research in more diverse patient populations will be required before considering formal modification of the ASA PS system. More robust, prospective longitudinal studies involving metabolic markers, nutritional and functional analyses, and multidisciplinary assessments may further identify which patients are at greatest risk for postoperative complications and those with the potential for modifiable preoperative risk factor interventions.

Back to Top | Article Outline

APPENDIX 1

Multivariate Analysis for Predicting Mortality in a Broad Surgical Population

Multivariate Analysis for Predicting Mortality in a Broad Surgical Population

Back to Top | Article Outline

CONCLUSIONS

Functional dependence proved to be a reliable independent predictor of postoperative mortality within each ASA class. It must be emphasized that these data have provided evidence that functionally dependent patients within each ASA class had greater risk for mortality than functionally independent patients in the subsequent higher (sicker) class, despite the higher class carrying a greater disease burden by definition. This could provide a basis for redefining the ASA PS classification system, but future prospective validation is necessary before any formal changes are considered.

Back to Top | Article Outline

DISCLOSURES

Name: Ognjen Visnjevac, MD.

Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.

Attestation: Ognjen Visnjevac has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Name: Sina Davari-Farid, MD.

Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.

Attestation: Sina Davari-Farid has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Name: Jun Lee, MSc.

Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.

Attestation: Jun Lee has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Name: Leili Pourafkari, MD.

Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.

Attestation: Leili Pourafkari has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Name: Pradeep Arora, MBBS.

Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.

Attestation: Pradeep Arora has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Name: Hasan H. Dosluoglu, MD.

Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.

Attestation: Hasan H. Dosluoglu has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Name: Nader D. Nader, MD, PhD.

Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.

Attestation: Nader D. Nader has seen the original study data, reviewed the analysis of the data, approved the final manuscript, and is the author responsible for archiving the study files.

This manuscript was handled by: Sorin J. Brull, MD, FCARCSI (Hon).

Back to Top | Article Outline

REFERENCES

1. American. Society of Anesthesiologists A. ASA Physical Status Classification System. 1995–2013 Available at https://www.asahq.org/resources/clinical-information/asa-physical-status-classification-system. Accessed October 15, 2014
2. Saklad M. Grading patients for surgical procedures. Anesthesiology. 1941;2:281–4
3. Davenport DL, Bowe EA, Henderson WG, Khuri SF, Mentzer RM Jr.. National Surgical Quality Improvement Program (NSQIP) risk factors can be used to validate American Society of Anesthesiologists Physical Status Classification (ASA PS) levels. Ann Surg. 2006;243:636–41
4. Dripps RD, Lamont A, Eckenhoff JE. The role of anesthesia in surgical mortality. JAMA. 1961;178:261–6
5. Glance LG, Lustik SJ, Hannan EL, Osler TM, Mukamel DB, Qian F, Dick AW. The Surgical Mortality Probability Model: derivation and validation of a simple risk prediction rule for noncardiac surgery. Ann Surg. 2012;255:696–702
6. Hamel MB, Henderson WG, Khuri SF, Daley J. Surgical outcomes for patients aged 80 and older: morbidity and mortality from major noncardiac surgery. J Am Geriatr Soc. 2005;53:424–9
7. Khuri SF, Daley J, Henderson W, Hur K, Gibbs JO, Barbour G, Demakis J, Irvin G III, Stremple JF, Grover F, McDonald G, Passaro E Jr, Fabri PJ, Spencer J, Hammermeister K, Aust JB. Risk adjustment of the postoperative mortality rate for the comparative assessment of the quality of surgical care: results of the National Veterans Affairs Surgical Risk Study. J Am Coll Surg. 1997;185:315–27
8. Marx GF, Mateo CV, Orkin LR. Computer analysis of postanesthetic deaths. Anesthesiology. 1973;39:54–8
9. Menke H, Klein A, John KD, Junginger T. Predictive value of ASA classification for the assessment of the perioperative risk. Int Surg. 1993;78:266–70
10. Prause G, Ratzenhofer-Comenda B, Pierer G, Smolle-Jüttner F, Glanzer H, Smolle J. Can ASA grade or Goldman’s cardiac risk index predict peri-operative mortality? A study of 16,227 patients. Anaesthesia. 1997;52:203–6
11. Story DA, Fink M, Leslie K, Myles PS, Yap SJ, Beavis V, Kerridge RK, McNicol PL. Perioperative mortality risk score using pre- and postoperative risk factors in older patients. Anaesth Intensive Care. 2009;37:392–8
12. Vacanti CJ, VanHouten RJ, Hill RC. A statistical analysis of the relationship of physical status to postoperative mortality in 68,388 cases. Anesth Analg. 1970;49:564–6
13. Wolters U, Wolf T, Stützer H, Schröder T. ASA classification and perioperative variables as predictors of postoperative outcome. Br J Anaesth. 1996;77:217–22
14. Copeland GP, Jones D, Walters M. POSSUM: a scoring system for surgical audit. Br J Surg. 1991;78:355–60
15. Donati A, Ruzzi M, Adrario E, Pelaia P, Coluzzi F, Gabbanelli V, Pietropaoli P. A new and feasible model for predicting operative risk. Br J Anaesth. 2004;93:393–9
16. Lloyd H, Ahmed I, Taylor S, Blake JR. Index for predicting mortality in elderly surgical patients. Br J Surg. 2005;92:487–92
17. Sutton R, Bann S, Brooks M, Sarin S. The Surgical Risk Scale as an improved tool for risk-adjusted analysis in comparative surgical audit. Br J Surg. 2002;89:763–8
18. Tiret L, Hatton F, Desmonts JM, Vourc’h G. Prediction of outcome of anaesthesia in patients over 40 years: a multifactorial risk index. Stat Med. 1988;7:947–54
19. Crawford RS, Cambria RP, Abularrage CJ, Conrad MF, Lancaster RT, Watkins MT, LaMuraglia GM. Preoperative functional status predicts perioperative outcomes after infrainguinal bypass surgery. J Vasc Surg. 2010;51:351–8
20. Dosluoglu HH, Wang J, Defranks-Anain L, Rainstein M, Nader ND. A simple subclassification of American Society of Anesthesiology III patients undergoing peripheral revascularization based on functional capacity. J Vasc Surg. 2008;47:766–72
21. Visnjevac O, Nader ND. Is ASA classification a valid prognosticator in octogenarians? Probably not. J Anesthesia Clinic Res. 2013;4:350
22. Sohn MW, Arnold N, Maynard C, Hynes DM. Accuracy and completeness of mortality data in the Department of Veterans Affairs. Popul Health Metr. 2006;4:2
23. American College of Surgeons A. User Guide for the Participant Use Data File. 2007 Chicago, IL American College of Surgeons Inc.:1–40
24. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics. 1988;44:837–45
25. Cox DR. Regression models and life tables J R Stat Soc. 1972;34:187–220
26. Kaplan EL, Meier P. Nonparametric estimation from incomplete observation. J Am Stat Assoc. 1958;53:457–81
27. Visnjevac O, Lee J, Pourafkari L, Dosluoglu HH, Nader ND. Functional capacity as a significant independent predictor of postoperative mortality for octogenarian ASA III patients. J Gerontol A Biol Sci Med Sci. 2014;69:1229–35
28. Rajan KB, Hebert LE, Scherr PA, Mendes de Leon CF, Evans DA. Disability in basic and instrumental activities of daily living is associated with faster rate of decline in cognitive function of older adults. J Gerontol A Biol Sci Med Sci. 2013;68:624–30
29. Ritchie CS, Locher JL, Roth DL, McVie T, Sawyer P, Allman R. Unintentional weight loss predicts decline in activities of daily living function and life-space mobility over 4 years among community-dwelling older adults. J Gerontol A Biol Sci Med Sci. 2008;63:67–75
© 2015 International Anesthesia Research Society