Massarweh NN, Legner VJ, Symons RG, McCormick WC, Flum DR. Impact of Advancing Age on Abdominal Surgical Outcomes. Arch Surg. 2009;144(12):1108-1114. doi:10.1001/archsurg.2009.204
To describe the population-level risk of adverse outcomes among older adults undergoing common abdominal surgical procedures.
Retrospective, population-based cohort study.
Washington State hospital discharge database.
A total of 101 318 adults 65 years or older who underwent common abdominal procedures such as cholecystectomy, colectomy, and hysterectomy from 1987 through 2004.
Main Outcome Measures
Ninety-day rates of postsurgical morbidity and mortality.
The 90-day cumulative incidence of complications was 17.3%, with a 90-day mortality rate of 5.4%. Advancing age was associated with increasing frequency of complications (65-69 years, 14.6%; 70-74 years, 16.1%; 75-79 years, 18.8%; 80-84 years, 19.9%; 85-89 years, 22.6%; and ≥90 years, 22.7%; trend test, P < .001) and mortality (65-69 years, 2.5%; 70-74 years, 3.8%; 75-79 years, 6.0%; 80-84 years, 8.1%; 85-89 years, 12.6%; and ≥90 years, 16.7%; trend test, P < .001). After adjusting for demographic, patient, and surgical characteristics as well as hospital volume, the odds of early postoperative death increased considerably with each advance in age category. These associations were found among patients with both cancer and noncancer diagnoses and for both elective and nonelective admissions (trend test, P < .001).
Among older adults, the risk of complications and early death after commonly performed abdominal procedures is greater than previously reported. These rates should be considered in ongoing quality improvement initiatives and may be helpful when counseling patients regarding abdominal operations.
Adults 65 years or older constitute the fastest-growing segment of the population.1 Current estimates predict that, in the year 2020, 1 in 6 Americans will be 65 years old or older and 15% of older adults will be greater than 85 years.1 Approximately 2 million older Americans undergo abdominal surgical operations each year.2 Older adults may have greater difficulty recovering from surgery because of decreased physiologic reserve,3- 5 but many case series have reported acceptable rates of adverse outcomes after a variety of abdominal surgical procedures.6- 13
For clinicians, patients, and families considering abdominal surgical procedures, informed decision making is challenging because of limited data regarding the risks of adverse perioperative events associated with advancing age. In counseling patients regarding risk, some rely on the extent of comorbid illness and “physiologic age” rather than age measured in years.14- 18 Community-level risk of adverse outcomes in this growing and potentially higher-risk population has yet to be addressed. The purpose of this study was to describe the population-level risk of adverse outcomes among older patients undergoing common abdominal surgical procedures and to assess the impact of advancing chronologic age on adverse outcome.
Our group's related work19 describes the Washington State Community Hospital Discharge Data (CHARS) data set and the study design.
Patients 65 years or older who had a valid Washington State zip code of residence were included. We searched all CHARS reports from 1987 through 2004 for International Classification of Diseases, Ninth Revision (ICD-9) procedure codes representing an abdominal operation as the primary procedure during hospitalization. Abdominal operations included appendectomy (ICD-9 code 47.0), cholecystectomy (ICD-9 code 51.2), colectomy (ICD-9 codes 45.7 and 45.8), esophagectomy (ICD-9 code 42.4), gastrectomy (ICD-9 code 43.0), hysterectomy (ICD-9 codes 68.3, 68.4, and 68.6-68.9), nephrectomy (ICD-9 code 55.5), oophorectomy (ICD-9 codes 65.0 and 65.3-65.6), pancreatectomy (ICD-9 codes 52.5-52.7), and prostatectomy (ICD-9 codes 60.3-60.6). Subcategories for each code, if present, were included with the appropriate abdominal surgical procedure category (hysterectomy and oophorectomy were combined).
Patients were divided into 6 age categories: 65 to 69 years, 70 to 74 years, 75 to 79 years, 80 to 84 years, 85 to 89 years, and 90 years or older. Complications were defined as either a specified postoperative complication during the index hospitalization (eg, postoperative pneumonia) or a more general complication (eg, bacterial pneumonia) if it occurred within 90 days of discharge (an appendix with ICD-9 codes for included complications can be accessed at http://depts.washington.edu/sorce/elderly_outcomes_appendix.html). Subjects were identified as having died within 90 days of admission on the basis of length of hospital stay, discharge status, and number of days between discharge and death. Neighborhood income was defined as the average household income for the patient's zip code of residence and was stratified into quartile ranges (category 1, $0-$36 239; category 2, $36 240-$43 320; category 3, $43 321-$51 736; category 4, ≥$51 737). Admission priority was classified as either elective or not elective (urgent or emergent). To adjust for potential confounding based on comorbid conditions, the comorbidity index, initially described by Charlson et al20 (score range, 0-3, with 3 indicating greatest comorbidity) and subsequently adapted by Deyo and colleagues21 for use with administrative data utilizing ICD-9 diagnostic codes, was calculated for each patient and was based on the index admission. Hospital volume was calculated from the average number of abdominal operations performed per year at each hospital and was stratified into quartile ranges (category 1, 1-6 operations; category 2, 7-30 operations; category 3, 31-90 operations; and category 4, ≥91 operations).
Categorical variables were compared by means of χ2 statistics, and continuous variables were compared with analysis of variance. A nonparametric test for trend across ordered groups was applied to evaluate significant changes in rates across age categories. Multivariate logistic regression models were constructed to evaluate the odds of 90-day postoperative complications and the odds of 90-day death on the basis of age category. Reference categories within the models were defined as follows: age, 65-69 years; sex, female; income category, 1; Charlson index, 0; admission source, home; surgical procedure, appendectomy; urgency of admission, elective; surgery year, 1987 to 1991; postoperative complications, none; and hospital volume category, 1. Models were constructed in 4 steps. The first model included age stratified into 5-year intervals. The second model evaluated the variable of interest (age category) and patient demographic characteristics including sex and income. The third model evaluated the effect of model 2 variables with the addition of other clinical characteristics such as comorbidity, type of operation, source of admission, urgency of admission, year of surgery, and postoperative complications. The fourth model included model 3 variables as well as hospital volume. All models were evaluated by means of generalized estimating equations and included hospital identification as the clustering variable. These models were developed by a nonparsimonious approach and included the foregoing variables of interest. Model fit was assessed by means of generalized Pearson residuals.22 Statistical analysis was performed with Stata statistical analysis software, version 9 (StataCorp, College Station, Texas).
A total of 101 318 patients 65 years or older (mean [SD] age, 74.4 [6.9] years) underwent abdominal operations between 1987 and 2004. Demographic, clinical, and hospital characteristics are listed in Table 1. Approximately 88.2% of the procedures were cholecystectomy, colectomy, hysterectomy/oophorectomy, or prostatectomy. Greater than 20% of patients were 80 years or older (23.2%). The most common comorbid conditions included chronic obstructive pulmonary disease (12.2%), diabetes mellitus (8.9%), congestive heart failure (6.5%), and a history of myocardial infarction (3.1%). We identified 41.5% of patients as having a malignant diagnosis. Of patients with malignant neoplasms, 31.9% had metastatic disease.
The 90-day mortality rate was 5.4%. Mortality rates for each procedure ranged from 0.7% for prostatectomy to 16.0% for gastrectomy. We identified a direct relationship between increasing 90-day mortality and advancing age (65-69 years, 2.5%; 70-74 years, 3.8%; 75-79 years, 6.0%; 80-84 years, 8.1%; 85-89 years, 12.6%; and ≥90 years, 16.7%; trend test, P < .001). The progressive increase in mortality with advancing age was statistically significant for every abdominal operation (trend test, P < .01 for esophagectomy and < .001 for all other operations) except pancreatectomy (Table 2). The trend of increasing mortality with each 5-year age category was identified for elective (65-69 years, 1.4%; 70-74 years, 2.2%; 75-79 years, 3.7%; 80-84 years, 4.8%; 85-89 years, 7.5%; and ≥90 years, 10.0%; trend test, P < .001) and nonelective (65-69 years, 5.0%; 70-74 years, 6.7%; 75-79 years, 9.2%; 80-84 years, 11.8%; 85-89 years, 16.7%; and ≥90 years, 20.4%; trend test, P < .001) admissions. This trend was also identified for operations related to both malignant (65-69 years, 2.6%; 70-74 years, 3.7%; 75-79 years, 6.2%; 80-84 years, 8.4%; 85-89 years, 12.7%; and ≥90 years, 16.8%; trend test, P < .001) and benign (65-69 years, 2.5%; 70-74 years, 3.8%; 75-79 years, 5.8%; 80-84 years, 8.0%; 85-89 years, 12.5%; and ≥90 years, 16.7%; trend test, P < .001) disease.
Postoperative complications occurred within 90 days in 17.3% of patients (range, 10.0% for prostatectomy to 48.8% for esophagectomy). The most common postoperative complications were postoperative pneumonia (8.3%), acute renal failure (5.2%), and surgical wound infection (4.2%). The frequency of 90-day postoperative complications increased with each 5-year incremental increase in age (65-69 years, 14.6%; 70-74 years, 16.1%; 75-79 years, 18.8%; 80-84 years, 19.9%; 85-89 years, 22.6%; and ≥90 years, 22.7%; trend test, P < .001). The progressive increase in complications with advancing age was statistically significant for 5 of the abdominal operations (appendectomy, cholecystectomy, colectomy, hysterectomy/oophorectomy, and prostatectomy; trend test, P < .001 [Table 3]). The trend of increasing complications with each 5-year age category was identified for elective (65-69 years, 13.4%; 70-74 years, 14.5%; 75-79 years, 16.5%; 80-84 years, 17.5%; 85-89 years, 19.5%; and ≥90 years, 20.1%; trend test, P < .001) and nonelective (65-69 years, 17.2%; 70-74 years, 19.0%; 75-79 years, 22.1%; 80-84 years, 22.7%; 85-89 years, 24.8%; and ≥90 years, 24.2%; trend test, P < .001) admissions. This trend was also identified for operations related to both malignant (65-69 years, 14.9%; 70-74 years, 16.0%; 75-79 years, 19.2%; 80-84 years, 19.3%; 85-89 years, 21.9%; and ≥90 years, 22.4%; trend test, P < .001) and benign (65-69 years, 14.1%; 70-74 years, 16.2%; 75-79 years, 18.5%; 80-84 years, 20.4%; 85-89 years, 22.8%; and ≥90 years, 22.9%; trend test, P < .001) disease.
Multivariate analysis demonstrated a relationship between advancing age and 90-day mortality (Table 4). Even after adjusting for patient demographics, comorbidity, source and urgency of admission, type of operation, postoperative complications, hospital volume, and discharge year, there was at least a 2-fold increase in the odds of early death for each 5-year incremental increase in age beyond 80 years. The odds of 90-day complications were 30% higher (95% confidence interval [CI], 1.2-1.4) and the odds of 90-day mortality were 2.7 times higher (95% CI, 2.5-3.0) for patients 80 years or older compared with those aged 65 to 69 years. We also performed an analysis that used age as a continuous variable and adjusted for demographic, patient, and surgical characteristics, as well as hospital volume. We found a 1% increase in the odds of postoperative complications (odds ratio, 1.01; 95% CI, 1.01-1.02) and a 6% increase in the odds of mortality (odds ratio, 1.06; 95% CI, 1.05-1.06) when comparing a patient undergoing abdominal surgery with a counterpart 1 year younger (starting with age 65 years). The occurrence of any postoperative complication was associated with increased odds of 90-day mortality (OR, 3.8; 95% CI, 3.6-4.2).
The presence of postoperative complications was associated with an increase in 90-day mortality. Older patients with increasing numbers of complications fared much worse than their younger counterparts (Table 5). For example, a 90-year-old patient with 3 postoperative complications had a 44.2% mortality rate compared with a rate of 15.9% for a 65- to 69-year-old. (This trend was true for all age categories and across varying numbers of complications.) After adjusting for all possible variables, advancing age was associated with an increase in the odds of a postoperative complication at 90 days. However, this increase was more gradual than the association between age and mortality (a 10% increase for ages 70-74 years [95% CI, 1.0-1.1]; a 20% increase for ages 75-79 years [95% CI, 1.1-1.2] and 80-84 years [95% CI, 1.3-1.3]; and a 30% increase for ages 85-89 years [95% CI, 1.2-1.5] and ≥90 years [95% CI, 1.2-1.4]).
This population-based study of older adults undergoing abdominal surgery in Washington State found a high rate of adverse outcomes among patients of advanced age. Our principal purpose in performing this study was to describe early outcomes in elderly surgical patients. After adjusting for a number of confounding variables, advancing age was associated with an increased frequency of complications and mortality at 90 days. Notably, with respect to the year surgery was performed, more contemporary periods were associated with decreased odds of mortality. However, even after adjusting for this variable, the odds of 90-day mortality were directly associated with advancing patient age.
Interestingly, patients who underwent surgery for malignant neoplasms fared no worse than those who had operations for nonmalignant reasons. Superior outcomes have been widely reported for a number of cancer and noncancer operations performed at high-volume hospitals and by high-volume practitioners.23- 27 Compared with the lower-volume categories, the odds of 90-day mortality for patients who underwent surgery at hospitals in the highest-volume category was slightly lower; however, even the highest-volume category was associated with a 2-fold increase in the odds of early mortality.
Health care providers counsel patients and families considering operative procedures regarding the risks and benefits of those interventions. Because many of these risks are not well quantified, providers often rely on a patient's comorbid disease burden and that person's ill-defined “physiologic age,” rather than actual chronologic age, as a predictor of postoperative adverse outcomes. Case series tend to support this approach but may represent both selection and publication bias. For example, in many series of older patients undergoing surgery, increasing age was not associated with adverse postoperative outcomes.6,7,13,28 In each, comorbid disease was considered a more important predictor of adverse outcome than was chronologic age. In contrast, we found not only a significant effect of advancing age but also markedly different complication and mortality rates across different age categories. Even after adjusting for other patient and clinical characteristics, as well as hospital volume, 90-year-old patients had 4-fold higher odds of early postoperative death than did patients aged 65 to 69 years.
Different trends were reported in a series of 4315 patients undergoing elective noncardiac surgery at an urban academic medical center.29 That study suggested that the low rate of complications (12.5%) and death (2.6%) in the oldest cohort (patients older than 80 years) was acceptable and not prohibitive. Our results suggest that many abdominal procedures, even elective operations, have high adverse outcome rates for the oldest patients. This may affect the decision-making process for elective as well as nonelective procedures. One reason our study demonstrated higher rates compared with the single-center study may be that we included only patients undergoing abdominal surgery, which could be associated with worse outcomes than other types of surgical procedures. A large study in the veteran population disclosed similarly high complication and mortality rates in older patients undergoing a variety of noncardiac operations when age was examined as a dichotomous variable, with 80 years used as a cutoff.30 We were able to explore in detail the association between incremental increases in age and the risk of adverse postoperative outcomes after individual abdominal operations. Although increased levels of comorbid illness were associated with worse outcomes, advancing age was more strongly and independently associated.
A recent study primarily of male veterans that used the National Surgical Quality Improvement Program data set suggested that the occurrence of postoperative complications was a more important determinant of survival than was advancing age.31 We found age-unadjusted mortality rates similar to those of the National Surgical Quality Improvement Program study, with higher 90-day mortality in those experiencing any postoperative complication (14.9% vs 3.4%). However, in our study, advancing age was associated with dramatically increasing 90-day mortality rates, and this effect was accentuated by the presence of a postoperative complication. Our study differs from the National Surgical Quality Improvement Program study in that we focused on adults 65 years or older, more than half of whom were women, and we explored a more homogeneous mix of procedures, looking only at abdominal operations. For these reasons, our results may be more generalizable to the typical older adult living in the community and undergoing these procedures.
As the population ages, more emphasis is placed on trying to reduce morbidity and delay illness, with the goal of converting a gradually declining disability curve to a more rectangular one in which older patients remain at high functional levels until death.32 Our study provides an analysis of one of the largest cohorts of elderly patients and indicates that the stress of surgery has a major effect on older individuals and is associated with high rates of morbidity and mortality. The ensuing burden of premature morbidity creates significant stress not only on the affected individual but also on their families and the entire health care system. Strategies to reduce the impact of advancing age on surgical morbidity and mortality have yet to be fully explored. Health care providers should be aware of the adverse impact that surgery has on older adults and develop interventions to help reduce or prevent the occurrence of adverse outcomes.
This study had several limitations. The CHARS administrative discharge database provides data on nearly all hospitalized patients from Washington State. Because CHARS does not constitute a national sample, the racial, socioeconomic, and rural/urban makeup of Washington State residents may be different from those of the United States as a whole and therefore may not be completely generalizable to the entire US elderly population. Furthermore, because the CHARS data set is administrative, it shares many of the well-known limitations associated with the use of this type of data. The nature of administrative data makes validation difficult, and we were unable to measure, and therefore control for, a number of important variables such as cardiovascular performance (based on echocardiographic findings), preoperative and postoperative medication administration (in particular, β-blockers and perioperative antibiotics), preoperative and postoperative laboratory values (allowing determination of renal and other organ function), family involvement, and other pertinent surgical and anesthetic factors that may have led to overrepresentation of mortality risk. Alternatively, our results might also have underrepresented the mortality risk in this population. Because we used the vital record for the state of Washington to capture mortality data, data from patients who underwent surgery and then left the state and died were not captured.
The presence of a postoperative complication was determined by means of ICD-9 procedure and diagnosis codes, which poses another limitation. Although the specificity of using these codes compared with medical record review has been found to be quite high (97.4%), the sensitivity for capturing a diagnosis was only 59.9%.33 This is an important point because we likely missed a number of events that bear on outcome in this population (eg, myocardial ischemia as opposed to infarction). Others have used the Agency for Healthcare Research and Quality–Patient Safety Indicators system, which may be more inclusive.34,35 In addition, we used a very conservative definition for the occurrence of complications. A complication was included during the index admission only if the diagnosis code specified that it was a postoperative adverse outcome. More general adverse outcomes were included only if readmission to a hospital occurred within a specified period. Because of the way we defined postoperative complications, it is possible that we underestimated the frequency of adverse outcomes, and the true rate may be higher than what we have reported. Others have used the Complications Screening Program in an attempt to capture a higher number of complications.36 We chose to use a very conservative definition for complications to present a “best-case scenario,” knowing that outcomes are likely even worse than we report.
We used the Charlson comorbidity index to control for the confounding influences of medical illness on mortality and classified comorbidity on the basis of diagnoses present during the index admission (not from previous admissions). Because this index is based on the completeness and accuracy of diagnostic coding, by using this method we may have underestimated the degree of comorbidity in our sample. The database also contained limited information regarding socioeconomic status, which may be an important factor associated with postoperative outcomes. We used mean income in the geographic area of residence to adjust for socioeconomic differences, which may be suboptimal, necessitating future studies to more fully explore the relationship. In addition, we were able to adjust for urgency of admission (eg, elective admission), but this is not the same as an elective operation and must be considered when interpreting our results. The CHARS data set contains information on patients who were selectively offered and accepted surgery, thereby excluding patients who were too ill or too frail to undergo surgery. These excluded patients might be expected to have worse surgical outcomes because of their physical frailty. Thus, care should be taken when applying our data to more frail elderly patients because their risk may be higher than reported herein.
In conclusion, we found that the risk of complications and mortality after abdominal procedures increased significantly for each 5-year age interval beyond 65 years for common abdominal operations. Early mortality rates in elderly patients were higher than previously reported in a cohort of approximately equal size from the Veterans Affairs population.31 The increase in mortality with advancing age remained significant after adjusting for demographic, patient, and surgical characteristics, as well as hospital volume. Older adults may be less able to adapt to the stress of surgery or to the added stress of any postoperative complication, greatly increasing their risk of early mortality. These effects appear to be additive, highlighting the need for interventions to both prevent decline among older patients and avoid postsurgical complications.
Correspondence: Nader N. Massarweh, MD, Department of Surgery, University of Washington, 1959 NE Pacific St, PO Box 356410, Seattle, WA 98195-6410 (email@example.com).
Accepted for Publication: November 21, 2008.
Author Contributions:Study concept and design: Legner, McCormick, and Flum. Acquisition of data: Symons and Flum. Analysis and interpretation of data: Massarweh, Legner, Symons, and Flum. Drafting of the manuscript: Massarweh and Legner. Critical revision of the manuscript for important intellectual content: Massarweh, Symons, McCormick, and Flum. Statistical analysis: Massarweh, Legner, Symons, and Flum. Obtained funding: Legner. Administrative, technical, and material support: Massarweh, Symons, and Flum. Study supervision: McCormick and Flum.
Financial Disclosure: None reported.
Funding/Support: This research was supported by the National Institutes of Health Roadmap Multidisciplinary Clinical Research Career Development Award Grant (8K12RR023265-02) from the National Institutes of Health and the Hartford/American Federation for Aging Research.
Role of the Sponsors: The funding agencies played no role in the design or performance of this study or in the creation or submission of the manuscript.