Cost is in 2013 US dollars.
eTable 1. Multivariable Analysis of Log Transformed Costs
eTable 2. Multivariable Analysis of Log Transformed Total Payments
eTable 3. Margins Under Fee-for-service
eTable 4. Bundled Payments by DRG and Resulting Margin
eFigure. Comparison of Net Margins (Defined as the Difference Between Total In-Patient Costs and Total Payments) Under a Traditional Volume-Based Payment Model Versus Bundled Payment Model
Customize your JAMA Network experience by selecting one or more topics from the list below.
Gani F, Makary MA, Wick EC, et al. Bundled Payments for Surgical Colectomy Among Medicare Enrollees: Potential Savings vs the Need for Further Reform. JAMA Surg. 2016;151(5):e160202. doi:10.1001/jamasurg.2016.0202
The Bundled Payments for Care Improvement Initiative was proposed by the Centers for Medicare and Medicaid Services to obtain and reward a greater value of care. Still in its infancy, little is known regarding the potential effects of the Bundled Payments for Care Improvement Initiative on hospital payments and net margins.
To investigate the potential effects of the Bundled Payments for Care Improvement Initiative on net margins among Medicare patients undergoing colectomy at a tertiary care hospital.
Design, Setting, and Participants
Cross-sectional retrospective analysis conducted in October 2015. Medicare enrollees undergoing an elective colectomy at a large tertiary care hospital between January 1, 2009, and December 31, 2013, were identified using diagnosis-related group and International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis codes.
Main Outcomes and Measures
Multivariable linear regression analysis was performed to calculate risk-adjusted, diagnosis-related group–specific hospital costs and payments for each patient. Net margins were calculated as the difference between total hospital costs and total payments received.
A total of 821 Medicare enrollees underwent an elective colectomy and met inclusion criteria. The median age of patients was 69 years (interquartile range [IQR], 65-74 years), with 51.3% being female. Postoperative complications were observed among 27.5% of patients (n = 226) and the median length of stay was 8 days (IQR, 5-14 days). The median risk-adjusted cost among all patients was $24 951 (IQR, $16 197-$38 922). Risk-adjusted costs were higher among patients who developed a postoperative complication ($42 537 [IQR, $28 918-$72 316] vs $22 829 [IQR, $14 820-$26 150]; P < .001) and among patients with an observed to expected length of stay greater than 1 ($36 826 [IQR, $24 951-$65 016] vs $16 197 [IQR, $14 182-$23 998]; P < .001). The median payment under the fee-for-service structure was $29 603 (IQR, $17 742-$44 819), resulting in an overall net margin of $3177 (IQR, −$1692 to $10 773), with 33.7% of patients (n = 277) contributing to an overall negative margin. In contrast, under the bundled payment paradigm, the net margin per patient was $3442 (IQR, −$9311 to $8203), with 41.7% of patients (n = 342) contributing to a net negative margin.
Conclusions and Relevance
Postoperative complications, length of stay, and total hospital costs were strongly associated with hospital costs. Payments under the bundled payments system were lower and the proportion of patients contributing to a net negative margin increased. Further study is warranted to define the effect of bundled payments on quality of care and hospital finances.
With an estimated $400 billion spent annually to cover the costs of surgical care, there is increasing focus to improve the value of surgical care via payment reform.1 To this end, the Affordable Care Act of 2012 is accelerating a shift from a traditional fee-for-service payment structure toward a value-based payment model.2 An innovative payment paradigm that has gained significant traction is the Bundled Payments for Care Improvement Initiative (BPCI) from the Centers for Medicare and Medicaid Services (CMS).2,3 Under the BPCI, hospitals receive a single, fixed bundled payment for all services delivered within an episode of care. Implemented in 2010, the BPCI has since expanded to allow participating organizations to enroll into bundled payment agreements (with Medicare) for up to 48 predefined clinical conditions aggregated from the Medicare diagnosis-related group (DRG) system.4 While preliminary studies assessing the efficacy of such payment strategies for cardiac and orthopedic surgical procedures have suggested a potential reduction in wasteful spending, it remains unclear whether this payment structure will be capable of broadly transforming health care delivery and spending.4-6
As hospitals and physicians move to accept greater financial responsibility under this new payment program, a deeper understanding of hospital finances is necessary to identify areas for quality improvement and subsequently determine target prices for payment bundles.7,8 To make the business case for such paradigms, financial data must be highly developed to reflect changes in expenses that occur when patients have comorbidity or disease severity different from the average. Similarly, policies must also adopt more accurate methods that ensure appropriate payments to hospitals caring for patients likely to incur high-cost episodes due to disease severity, medical comorbidity, or socioeconomic circumstances.9,10
Accounting for more than 300 000 hospitalizations and more than $2.4 billion in annual direct hospital costs, colectomy represents a commonly performed procedure that can be targeted under the BPCI to enhance the value of surgical care.7,11 While previous reports have demonstrated large variations in hospital payments and postoperative clinical outcomes following colon surgery, to our knowledge, no previous report has examined the effect of implementing a bundled payment model on hospital finances.4,7,11 Therefore, the current study sought to evaluate the effect of implementing a bundled payment model at a large tertiary care hospital. Specifically, we sought to quantitate the differences in hospital reimbursements when moving from the traditional fee-for-service model to a value-based bundled payment model and the resulting effect on hospital profitability and net margins.
Question What are the potential effects of the Bundled Payments for Care Improvement Initiative on net margins among Medicare patients undergoing colectomy at a tertiary care hospital?
Findings In this financial analysis of 821 Medicare enrollees undergoing an elective colectomy, 33.7% of patients contributed to an overall negative margin under a fee-for-service payment model. In contrast, under the bundled payment paradigm, 41.7% of patients contributed to a net negative margin.
Meaning Payments under the bundled payments system were lower and the proportion of patients contributing to a net negative margin increased.
This retrospective financial analysis was performed in October 2015 using administrative claims data collected for Medicare enrollees discharged from the Johns Hopkins Hospital between January 1, 2009, and December 31, 2013. Patient demographics, such as age, sex, race/ethnicity, admission status (elective vs emergent), length of stay (LOS), and up to 25 diagnostic and procedural codes for the index admission coded under the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) lexicon were collected for each patient. Patients undergoing a colectomy were identified using relevant ICD-9-CM procedure codes. Patients younger than 65 years or patients undergoing emergent procedures were excluded from further analysis.
To further ensure the homogeneity of the patient population, only patients with a final DRG (All Patient Refined [APR]–DRG) corresponding to those identified by CMS under the BPCI were selected for further analysis: 221 (major small- and large-bowel procedures) and 223 (other small- and large-bowel procedures).12 Patient comorbidity was classified according to the Charlson Comorbidity Index (CCI), categorizing patients into 1 of 3 comorbidity groups (ie, CCI score = 0, 1, or ≥2).13 Postoperative complications were defined using ICD-9-CM diagnostic codes representing surgical site infection/wound dehiscence, sepsis/septic shock, venous thromboembolism, stroke, myocardial infarction, pneumonia, and gastrointestinal bleeding.14 These codes were selected because they have previously been validated in analysis using administrative claims data.
The Johns Hopkins University institutional review board approved this study. Data used within the current study were deidentified and therefore, patient consent was waived.
For each patient record, detailed financial information, including total costs, fixed costs, variable costs, and net payments, were extracted from the institutional EPSi cost accounting system (Allscripts Healthcare Solutions Inc). For the purpose of this study, physician professional fees or salary data were not collected. Fixed costs were defined as costs that do not vary with patient volume and represent costs related to structural/building costs, maintenance, and utilities.15 In contrast, variable costs were defined as costs that vary with patient volume.15 Total hospital costs were calculated as the sum of fixed and variable costs for each patient. Payments represented the total amount of money collected for each patient after accounting for contractual discounts and revenues collected from patients under coinsurance or copayment. Hospital profitability for each patient was measured by the total margin or net profit and was calculated as the difference between payments received and total hospital costs. All financial data were inflation adjusted to and are reported as 2013 US dollars.16
Categorical variables are reported as percentages and whole numbers and compared using the Pearson χ2 test. Continuous variables are reported as medians with interquartile ranges (IQRs) or means with SDs and compared using the nonparametric Wilcoxon rank-sum test or t test as appropriate. As financial variables were right skewed, all financial data were log-transformed before entering into a multivariable linear regression model that adjusted for patient demographics (age, sex, and race/ethnicity), postoperative complications, LOS, procedure type, and APR-DRG. Potential interactions between patient comorbidity, LOS, and postoperative complications were tested by including interactions terms within the multivariable analysis. Because no statistically significant interactions were noted, an interaction term was not specified within the final model. Results of the multivariable regression analyses were then used to calculate risk-adjusted total hospital costs, fixed costs, and variable costs for each patient discharge. Using a previously described method, results from the multivariable analysis were also used to generate an average payment for each DRG representing the bundled payment.17 Specifically, for each DRG, the median payment received was used to represent the average bundled payment and used for subsequent calculations of the margins and profitability among patients within each specific DRG.17-20
Because patients may be clustered to specific physicians, an additional clustered hierarchical analysis was performed with a random-effects intercept specified at the level of the surgeon (eTable 1, eTable 2, eTable 3, and eTable 4 in the Supplement). Of note, because no differences in the risk-adjusted outcomes or subsequent conclusions were noted, the results presented herein are those from the fixed-effects model. Statistical significance was defined a priori as a P < .05. All analyses were performed using Stata statistical software version 14.0 for Windows (StataCorp).
The study identified 821 Medicare enrollees who underwent an elective colectomy at Johns Hopkins Hospital between January 1, 2009, and December 31, 2013, and met inclusion criteria. The median age of the cohort was 69 years (IQR, 65-74 years), with 22.1% of patients (n = 181) older than 75 years and 51.3% of patients (n = 421) being female (Table 1). Overall, the most common patient race/ethnicity was white (n = 604; 73.6%) followed by African American (n = 176; 21.4%) and Asian (n = 13; 1.6%). Comorbidities were common, with more than half of the patient population presenting with a CCI score of 2 or greater (n = 432; 52.6%). More than 87% of patients (n = 715) were billed under the APR-DRG 221, representing major small- and large-bowel procedures, while 12.9% (n = 106) were billed under the APR-DRG 223, representing other small- and large-bowel procedures. While age and race/ethnicity did not vary between these 2 patient groups, patients billed under the DRG 221 were proportionally more likely to be female (53.6% vs 35.9%) and present with greater comorbidity (CCI score ≥ 2: 55.8% vs 31.1%; P < .001).
The median LOS for all patients was 8 days (IQR, 5-14 days); 27.5% (n = 226) developed 1 or more postoperative complications following an elective colectomy (Table 2). Differences in postoperative clinical outcomes were also noted between the 2 patient groups. For example, postoperative morbidity was more common among patients billed under the DRG 221, with 28.5% of patients (n = 204) developing 1 or more postoperative complication compared with 20.8% of patients (n = 22) billed under the DRG 223. The median LOS was 4 days longer among patients billed under the DRG code 221 (median LOS: 9 days, IQR, 5-15 days vs 5 days, IQR, 3-10 days; P < .001) and a greater proportion of patients belonging to this subgroup were discharged home with additional care (43.5%, n = 311 vs 21.7%, n = 23; P < .001).
The median risk-adjusted fixed and variable costs among all patients undergoing an elective colectomy were $10 886 (IQR, $6988-$17 060) and $14 008 (IQR, $9178-$21 684), respectively. This resulted in a median total hospital cost of $24 951 (IQR, $16 197-$38 922). Of note, total hospital costs were higher among patients billed under the DRG 221 compared with those billed under the DRG 223 (median total hospital costs: $25 468 [IQR, $18 709-$40 586] vs $13 192 [IQR, $11 420-$21 298]; P < .001; Figure 1). Similarly, total hospital costs were also higher among patients who developed a postoperative complication compared with patients who did not. Specifically, total hospital costs were more than $20 000 greater among patients who developed a postoperative complication (median total hospital costs: $42 537 [IQR, $28 918-$72 316] vs $22 829 [IQR, $14 820-$26 150]; P < .001). Additionally, an increasing LOS was also associated with an increase in total hospital costs as patients with an observed to expected LOS of more than 1 incurred an almost 2-fold higher cost vs patients with an observed to expected LOS of 1 or less (median total hospital costs: $36 826 [IQR, $24 951-$65 016] vs $16 197 [IQR, $14 182-$23 998]; P < .001).
The risk-adjusted median payments for all patients was $29 603 (IQR, $17 742-$44 819). Total payments were higher for patients who developed a postoperative complication and those who demonstrated a longer LOS. Specifically, the total hospital costs and the total payments received were noted to increase with an increasing number of postoperative complications and there was a particular increase noted among patients who developed 5 or more postoperative complications (Figure 2). Similarly, an increasing LOS was associated with an increase in hospital costs and therefore total hospital payments. For example, the total payments made increased from $15 611 (IQR, $13 920-$17 159) among patients with an LOS of fewer than 5 days to $41 578 (IQR, $37 825-$44 818) among patients who stayed 11 days or longer after an elective colectomy (Figure 3). Under the fee-for-service payment model, the overall net margin was $3177 (IQR, −$1692 to $10 773), with 33.7% of patients (n = 277) contributing to an overall negative margin (eFigure in the Supplement). Interestingly, the overall net margin was noted to increase (greater profitability) with an increasing LOS for the index admission and with an increasing number of postoperative complications.
Results from the multivariable analysis were used to calculate risk-adjusted average payments (ie, bundled payment) for all patients undergoing an elective colectomy. Under this bundled payment model, the DRG-specific bundled payments were $30 150 for DRG 221 and $13 966 for DRG 223. This resulted in a median net margin of $3442 (IQR, −$9311 to $8203). Of note, the proportion of patients contributing to an overall negative margin under the bundled payment model markedly increased, with 41.7% of patients (n = 342) contributing to an overall negative margin (eFigure in the Supplement). Further, under the bundled payment model, a negative margin was also noted among patients who developed a postoperative complication, with the difference in cost and payments increasing with an increasing number of postoperative complications. For example, while the median net margin was $6124 (IQR, −$2095 to $13 953) among patients who did not develop a postoperative complication, a net negative margin of $12 387 (IQR, −$44 817 to −$55) was noted among patients who developed 1 or more postoperative complications. Similarly, patients who demonstrated a longer LOS were also associated with an increasing deficit in payments as patients with an LOS of 15 days or more contributed to an average negative margin of $38 848.
Despite an increasing sum of money spent each year to cover health care costs, the quality of surgical care in the United States ranks low.21,22 In an effort to curtail spending while maintaining the quality of care, recent years have seen increased enthusiasm in the shift from a fee-for-service payment model to a bundled payment structure.2 In particular, the BPCI (under which hospitals are reimbursed a single, fixed amount for all services provided under an episode of care) has been adopted by CMS to improve the value of surgical care.9 Although a limited number of reports targeting cardiac and orthopedic procedures have demonstrated potential savings under the BPCI, the effect of such a payment structure on hospital profitability and whether it will be widely applicable to improving health care delivery remains unknown.4,9
Using detailed financial information collected from patients undergoing an elective colectomy at a large, tertiary care hospital, the current study investigated the effect on hospital finances of moving from a fee-for-service payment model to a bundled payment model. Our financial analysis demonstrated that current payment models may allow for the misalignment of cost-containment strategies and the delivery of a high quality of care. Higher payments and greater profits were noted among patients who developed a postoperative complication and among patients with a greater than expected LOS. In contrast, payments under a bundled payment model were associated with an overall decreased reimbursement and decreased hospital profits. Of particular note, the number of patients who contributed to a net negative margin increased from 33.7% under the traditional fee-for-service model to more than 40% under the bundled payment model.
Variations in surgical costs have been targeted by policymakers as a means to decrease the financial burden of health care.23-25 While several previous studies have characterized this variation in hospital costs and payments, to our knowledge, few have focused on explaining this variation in care. Most studies have attributed the increased cost to the development of postoperative complications or preexisting medical comorbidity.11,15,17,24,25 Further, these reports suggest that costly surgical complications can be targeted as a means to improving the value of surgical care.26 The current study builds on this body of work, demonstrating that patients who developed a postoperative complication incurred the highest overall cost. The study is important in that it focused not on episode-based costs, but rather on other financial variables to explain further the association between hospital profitability and postoperative complications. Of note, the study demonstrated that current payment models may allow for the misalignment of financial incentives and efforts for quality improvement. Specifically, postoperative complications were associated with increased hospital payments and greater profit margins. Furthermore, payments and margins increased with an increasing number of postoperative complications. While hospitals and surgeons clearly do not desire more complications, these data suggest that institutions that take on more complicated patients will be at particular financial risk in a bundle-payment model. In turn, although low costs and high quality are often compatible within the business world, this remains an elusive goal for health care.22
To increase the value of surgical care, the CMS implemented the BPCI to incentivize hospitals to improve the quality of care. While preliminary data suggest a potential alignment of cost containment and quality-improvement efforts, it remains largely unknown whether such strategies are applicable to all payors or markets.4,6,18 In this current study, we noted a substantial improvement in the outcomes per dollar spent with lower margins/profits for patients who developed a postoperative complication and among those who had a longer than expected LOS. Collectively, results from the current study and those from previous reports serve to further make the business case for this value-based payment model.18
Despite the potential benefits for surgical care, policymakers and physicians must exercise caution in adopting the bundled payment structure in its current capacity to avoid its potential negative consequences. While bundled payments led to overall savings in surgical care, a significantly greater proportion of patients contributed to a net negative margin under this payment structure. Specifically, 33.7% of the cohort contributed to a negative margin under fee-for-service reimbursements, while more than 40% of bundled payments resulted in a shortfall between reimbursement and hospital costs. Given that increased costs are likely necessary to deliver a greater intensity of care for patients who develop postoperative complications or those who present with preexisting medical comorbidity, adequate risk adjustment across an episode of care is necessary to ensure that physicians caring for larger proportions of high-risk patients are adequately reimbursed.10
Risk-adjustment models must account for the interplay between patient characteristics (medical comorbidities, primary diagnosis, and socioeconomic factors) and the differential use of services included within each payment bundle.10 Failure to do so may disincentivize hospitals caring for high-risk patients, resulting in risk-averse behaviors and therefore having the unwanted effect of decreasing access to surgical care for some high-risk patients. Additionally, as savings achieved under the bundled payment model can be easily offset by an increase in overall episode volume, implementation of such a payment structure can potentially lead to narrower networks as hospitals and physicians contract with select payors to coordinate care. Therefore, policies must ensure a free and open marketplace where patients are given accurate price and quality measures, thereby equipping them with the autonomy to choose the most cost-effective and highest quality of surgical care.
Results of the current study should be interpreted in light of the following limitations. First, the current study reported on patients at a single, tertiary care referral center and therefore results from the current analysis may not be generalizable to the entire surgical population and could not account for variations in payment models across the United States. However, our results highlighted the financial implications of bundled payments on hospitals with a similar patient mix and serve to reiterate the need for further conversations around the appropriate methods and data needed to determine appropriate payments for hospitals. Second, the analysis was subject to the inherent limitations of administrative data, which include the lack of disease-specific/detailed clinical data, such as tumor grade, disease severity, and severity of postoperative complications.27 To overcome any potential coding discrepancies, we selected for appropriate patients using both ICD-9-CM procedure codes and relevant DRGs. Additionally, postoperative complications were identified using a set of diagnosis codes previously validated for use in administrative data.14 Last, the analysis focused only on inpatient costs and therefore could not comment on the costs associated with physician fees or postdischarge care, which have been demonstrated to be important drivers of variations in health care spending. As such, we could not comment on differences in hospital profitability and margins under other relevant payment models of the BPCI.
Results from the current study demonstrated that current payment paradigms may allow for a significant mismatch in financial incentive, quality of surgical care, and willingness of hospitals to take on high-risk patients in a bundled payment system. Although there may be a business case for a shift to a value-based payment model, it remains to be seen whether such efforts can effectively be implemented and coordinated between different physicians/stakeholders to truly deliver high-valued surgical care. Additionally, while bundled payments appear to be a future reality, further thought is warranted regarding the calculation of fixed payments and appropriate risk adjustment to avoid risk-averse practices among hospitals and physicians.
Corresponding Author: Timothy M. Pawlik, MD, MPH, PhD, Department of Surgery, Johns Hopkins Hospital, 600 N Wolfe St, Blalock 688, Baltimore, MD 21287 (email@example.com).
Accepted for Publication: January 15, 2016.
Published Online: March 16, 2016. doi:10.1001/jamasurg.2016.0202.
Author Contributions: Drs Gani and Pawlik had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Gani, Makary, Efron, Safar, Hundt, Pawlik.
Acquisition, analysis, or interpretation of data: Gani, Makary, Wick, Fang, Hundt, Pawlik.
Drafting of the manuscript: Gani, Pawlik.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Gani.
Administrative, technical, or material support: Gani, Efron, Fang, Hundt, Pawlik.
Study supervision: Gani, Makary, Wick, Safar, Pawlik.
Conflict of Interest Disclosures: None reported.
Create a personal account or sign in to: