A, Extended LOS in the 75th percentile; B, extended LOS in the 90th percentile; C, inpatient complication rate; and D, inpatient severe complication rate.
A, Inpatient complications vs extended LOS in the 75th percentile (ρ = 0.56, P < .001); B, inpatient severe complications vs extended LOS in the 75th percentile (ρ = 0.49, P < .001); C, inpatient complications vs extended LOS in the 90th percentile (ρ = 0.46, P < .001); and D, inpatient severe complications vs extended LOS in the 90th percentile (ρ = 0.47, P < .001).
Abbreviations: LOS, length of stay; SIRS, systemic inflammatory response syndrome.
Data are presented as percentage of patients unless otherwise indicated. P < .001 for all characteristics.
Surgical site complications include organ-space surgical site infection or wound dehiscence. Pulmonary complications include unplanned reintubation, prolonged mechanical ventilation, pulmonary embolism, and pneumonia. Cardiac complications include cardiac arrest that requires cardiopulmonary resuscitation or myocardial infarction.
Abbreviation: LOS, length of stay.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Krell RW, Girotti ME, Dimick JB. Extended Length of Stay After Surgery: Complications, Inefficient Practice, or Sick Patients? JAMA Surg. 2014;149(8):815–820. doi:10.1001/jamasurg.2014.629
With the health policy focus on shifting risk to hospitals and physicians, hospital leaders are increasing efforts to reduce excessive resource use, such as patients with extended length of stay (LOS) after surgery. However, the degree to which extended LOS represents complications, patient illness, or inefficient practice style is unclear.
To examine the influence of complications on the variance in hospitals’ extended LOS rates after colorectal resections.
Design, Setting, and Participants
In this retrospective cohort study performed from January 1 through December 31, 2009, we analyzed data from the 2009 American College of Surgeons National Surgical Quality Improvement Program. Study participants were 22 664 adults undergoing colorectal resections in 199 hospitals.
Inpatient complications recorded in the American College of Surgeons National Surgical Quality Improvement Program registry. Inpatient complications were identified by the association of the complication's postoperative date with the patient’s surgical discharge date.
Main Outcome and Measure
Hospitals’ risk-adjusted extended LOS rates, defined as the proportion of patients with a hospital stay greater than the 75th percentile for the entire cohort.
A total of 2177 patients (42.8%) with extended LOSs did not have a documented inpatient complication. Although there was wide variation in risk-adjusted extended LOS (14.5%-35.3%) and risk-adjusted inpatient complication (12.1%-28.5%) rates, there was only a weak correlation (Spearman ρ = 0.56, P < .001) between the two. Only 52.0% of the variation in hospitals’ extended LOS rates was attributable to hospitals’ inpatient complication rates.
Conclusions and Relevance
Much of the variation in hospitals’ risk-adjusted extended LOS rates is not attributable to patient illness or complications and therefore most likely represents differences in practice style. Efforts to reduce excess resource use should focus on efficiency of care, such as increased adoption of enhanced recovery pathways.
With the policy emphasis on shifting risk to hospitals and physicians, such as bundled payments and pay for performance, hospital leaders are looking for ways to improve resource use.1-5 Although these policies will encourage hospitals to be more efficient in general, few data are available to help understand costs after surgery. Because hospitals lack detailed cost data, they commonly use length of stay (LOS) as a proxy for resource use.6,7 In this context of value-based payment, hospitals and physicians are increasing efforts to better understand and improve resource use and unnecessarily long postoperative hospital stays.
The best strategy to reduce excessive LOS after surgery is unclear, however. There are 2 common explanations for extended hospital stays after an operation. First, patients experience postoperative complications that extend the LOS through management of the complications (eg, additional operations), so it is possible that hospitals and physicians should focus on preventing and managing complications to improve overall efficiency. Second, differences in LOS are due to practice style differences among hospitals and physicians. There is differential adoption of new surgical technologies, such as minimally invasive approaches, and variable use of other efforts to coordinate care processes, such as enhanced recovery pathways.8,9
A better understanding of the extent to which extended LOS is attributable to patient illness, complications, or practice style differences is essential to targeting efforts for improvement. In this context, we studied the association between extended postoperative LOS and complications and the extent to which complications account for variation in hospitals’ extended LOS rates.
The study protocol was reviewed and deemed not regulated by the University of Michigan Institutional Review Board, so no informed consent was required. We analyzed data from the 2009 American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) clinical registry. Details regarding data abstraction and quality control have been described previously.10 Using relevant Current Procedural Terminology codes, we selected adult patients undergoing inpatient laparoscopic or open colorectal resections from January 1 through December 31, 2009, to form our study cohort.
We examined extended postoperative LOS, which we defined as a postoperative hospital stay greater than the 75th percentile for the entire cohort. We also examined LOS greater than the 90th percentile in sensitivity analyses. Hospitals’ extended LOS rates were defined as the proportion of patients with extended LOSs. We also assessed complications (eg, wound dehiscence; superficial, deep, or organ-space surgical site infection; myocardial infarction; cardiac arrest; prolonged ventilator requirement; unplanned reintubation; pneumonia; progressive renal insufficiency; acute renal failure; coma; stroke; deep venous thrombosis or pulmonary embolism; bleeding requiring transfusion of >4 U of blood; graft or prosthetic failure; urinary tract infection; and sepsis or septic shock) and severe complications (those listed above but excluding deep venous thrombosis, urinary tract infection, progressive renal insufficiency, and superficial or deep surgical site infection). Because inpatient complications would most likely prolong hospital stay, we focused our assessment on complications that occurred before the patient’s discharge date.
Patient variables recorded in the clinical registry include age; race; sex; indication for operation (from International Classification of Diseases, Ninth Revision codes); height; weight; functional status; American Society of Anesthesiologists (ASA) class; cardiac, pulmonary, renal, neurologic, endocrine, hematologic and vascular comorbidities; long-term corticosteroid therapy; disseminated cancer; prior operation; 10% or greater weight loss before operation; preoperative sepsis; open wound or transfusion requirement; and preoperative laboratory values. We reclassified continuous variables as categorical variables with 5 levels for model entry.
First, we assessed the proportion of patients with extended LOSs who did not experience an inpatient complication or severe complication. Second, we conducted 2 hospital analyses: the first assessed the correlation between hospitals’ risk-adjusted extended LOS and complication rates, and the second assessed the extent to which different complications explained the variation in hospitals’ risk-adjusted extended LOS rates.
We started by calculating hospitals’ risk-adjusted extended LOS and complication rates. All risk-adjustment models included patient age, sex, race, ASA class, comorbidities and laboratory variables, and procedural (eg, laparoscopic case and emergency procedure) variables to generate predicted outcome probabilities. Model discrimination was fair (C statistic = 0.774-0.818), and calibration was adequate (Hosmer-Lemeshow χ2 = 6.17-15.2).11 Dividing each hospital’s observed outcome rate by the sum of its predicted probabilities generates observed to expected outcome ratios, which when multiplied by the cohort’s outcome rate yield hospitals’ risk-adjusted rates. To further account for random outcome variation, we adjusted hospitals’ risk-adjusted rates using shrinkage estimators derived from hierarchical regression models.12-14 We then used the Spearman rank correlation test to compare hospitals’ risk-adjusted extended LOS and complication rates.
To assess the extent to which complications explained the variation in hospitals’ risk-adjusted extended LOS rates, we constructed a hierarchical logistic regression model for extended LOS with the hospital specified as the higher level. We serially assessed the proportional change in hospital-level random intercept variance after adding complications (patient-level and hospital-level complication rates) to the hierarchical model.14 Finally, we substituted specific severe complication types (eg, surgical site [organ-space surgical site infection or wound dehiscence], pulmonary [unplanned reintubation, prolonged mechanical ventilation, pulmonary embolism, or pneumonia], cardiac [cardiac arrest or myocardial infarction], and sepsis or septic shock). All models adjusted for patient age, sex, race, ASA class, comorbidities, laboratory values, and procedural variables as above.
We performed all analyses using STATA statistical software, version 12 (Stata Corp). All statistical tests were 2-sided with P < .05 considered significant.
We identified 22 664 patients undergoing colorectal resections in 199 hospitals participating in the ACS-NSQIP in 2009. The median, 75th percentile, and 90th percentile LOSs were 6, 9, and 16 days, respectively. Patients with extended LOS were older, had more comorbidities, underwent more emergency procedures, and more often had resections for obstructive reasons (Table 1). Although patients with extended LOS were more likely to have complications, a large proportion (2177 [42.8%]) did not have a documented complication or severe complication (2844 [55.9%]) (Table 1).
There was wide variation in hospitals’ risk-adjusted outcome rates but a weak correlation among outcomes (Figure 1 and Figure 2). For example, risk-adjusted extended LOS rates (range, 14.5%-35.3%) and complication rates (range, 12.1%-28.5%) had weak correlation (Spearman ρ = 0.56, P < .001) (Figure 2A). The correlation between extended LOS and severe complications was weaker (Spearman ρ = 0.49, P < .001) (Figure 2B). When extended LOS was defined as the 90th percentile, the correlation between extended LOS and complications was weaker still (Figure 2C and D).
Table 2 provides the proportion of hospitals’ risk-adjusted extended LOS rate variation attributable to complications. Complications explained more of the hospitals’ extended LOS rate variation (36.9%) than severe complications (31.2%). Similarly, the hospitals’ complication rates explained more (52.0%) of the extended LOS rate variation than the hospitals’ severe complication rates (47.0%). Surgical site and cardiac complications explained extended LOS rate variation equally (35.5% and 35.4%, respectively) and to a greater extent than pulmonary or septic complications (33.6% and 30.4%, respectively) (Table 2). When LOS was defined as the 90th percentile, cardiac complications accounted for more hospitals’ extended LOS rate variation (52.1%) than other complication types (surgical site, 47.7%; septic, 32.9%; and pulmonary, 32.3%).
With policy initiatives, such as bundled payments and pay for performance, hospital leaders have increased efforts to reduce excessive resource use.3-5 Postoperative LOS is a common proxy for episode resource use. A better understanding of the association between extended LOS and complications will help hospitals and physicians focus their efforts to reduce resource use. In this study, we found that a considerable proportion of patients with extended LOS do not have documented complications after a common and morbid procedure. There was weak correlation between hospitals’ risk-adjusted extended LOS and complication rates. Moreover, we found that 63.1% of the variation in extended LOS is attributable to hospital complication rates.
Studies7,15,16 that used administrative and clinical registry data found that a considerable proportion of patients with apparently uncomplicated hospital courses have extended LOSs. Conversely, another study found that patients with normal LOSs still have clinically relevant complications.17 Our study affirms these findings and further quantifies how little variation in hospitals’ extended LOS rates is explained by complications, even after accounting for patient illness. These results suggest that much of the variation in resource use surrounding surgical episodes may be caused by practice style differences rather than differences in technical quality or patient illness.
There is increased attention on understanding and implementing measures that address the efficiency of care provision. In other patient populations, care coordination and extended care facility availability influence LOS to a large degree.8,18 For surgical patients, emerging evidence suggests that process interventions, such as enhanced recovery pathways, are effective at reducing LOS without increasing overall complication rates, but the efficacy of such interventions on a large scale remains unclear.9,19-22 With different uptake and implementation of enhanced recovery for patients with colectomies, it would be reasonable to assume that practice style differences underlie at least a portion of the unexplained variation in hospitals’ extended LOS rates.
Our study has some important limitations. First, our data set lacked colectomy-specific complications that may better explain extended LOS, such as prolonged postoperative ileus, although the expected ileus rate for the cohort is far less than the amount of unexplained extended LOS.23 Second, although our risk-adjustment models accounted for patient illness, procedure type, and acuity, we lacked data on factors such as patient rurality, access to transportation, discharge planning, and care coordination, which undoubtedly influence LOS as well. Third, we analyzed a common gastrointestinal procedure, and our results may not apply to different procedures. Fourth, although LOS is a common proxy for hospital resource use, price index–adjusted total payments remain a more fair measure of resource use.6 Finally, our data represent a subset of hospitals with a presumed interest in quality improvement, and as such our results may not be generalizable to all hospitals.
Much of the variation among hospitals in their resource use remains unexplained after accounting for patient illness and complications. With increasing emphasis on improving the overall efficiency of episode-based care, a better understanding of practice style variation and how it contributes to differences in resource use should help guide improvement efforts apart from improving complication rates. In addition to focusing efforts on complication prevention, hospitals should also focus efforts on implementing and refining processes that eliminate inefficient practice.
Accepted for Publication: December 2, 2013.
Corresponding Author: Robert W. Krell, MD, Center for Healthcare Outcomes and Policy, 2800 Plymouth Rd, Bldg 16, Office 016-100N-13, Ann Arbor, MI 48109 (email@example.com).
Published Online: June 25, 2014. doi:10.1001/jamasurg.2014.629.
Author Contributions: Dr Dimick had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: All authors.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: All authors.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: All authors.
Obtained funding: Dimick.
Administrative, technical, or material support: Dimick.
Study supervision: Dimick.
Conflict of Interest Disclosures: Dr Krell reported having received a payment from Blue Cross/Blue Shield of Michigan for data entry unrelated to the submitted work. Dr Dimick reported having a financial interest in ArborMetrix Inc. No other disclosures were reported.
Funding/Support: This study is supported by grant 5T32CA009672-22 from the National Institutes of Health (Dr Krell), Career Development Award K08 HS017765 from the Agency for Healthcare Research and Quality ( Dr Dimick), and research grant R21DK084397 from the National Institute of Diabetes and Digestive and Kidney Diseases (Dr Dimick).
Role of the Sponsors: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: The ACS-NSQIP and the hospitals participating in the ACS-NSQIP are the source of the original data and cannot verify or be held responsible for the statistical validity of the data analysis or the conclusions derived by the authors.
Create a personal account or sign in to: