- Email: [email protected]
Vascular surgeon-hospitalist comanagement improves in-hospital mortality at the expense of increased in-hospital cost
From the Society for Clinical Vascular Surgery
Vascular surgeon-hospitalist comanagement improves in-hospital mortality at the expense of increased in-hospital cost Rami O. Tadros, MD,a Melissa L. Tardiff, BA,a Peter L. Faries, MD,a Michael Stoner, MD,b Chien Yi M. Png, BA,a David Kaplan, MPA,a Ageliki G. Vouyouka, MD,a and Michael L. Marin, MD,a New York and Rochester, NY
ABSTRACT Objective: We have shown that vascular surgeon- hospitalist co management resulted in improved in-hospital mortality rates. We now aim to assess the impact of the hospitalist co management service (HCS) on healthcare cost. Methods: A total of 1558 patients were divided into three cohorts and compared: 516 in 2012, 525 in 2013, and 517 in 2014. The HCS began in January 2013. Data were standardized for six vascular surgeons that were present 2012-2014. New attendings were excluded. Ten hospitalists participated. Case mix index (CMI), contribution margin, total hospital charges (THCs), length of stay (LOS), actual direct costs (ADCs), and actual variable indirect costs (AVICs) were compared. Analysis of variance with post-hoc tests, t-tests, and linear regressions were performed. Results: THC rose by a mean difference of $14,578.31 between 2012 and 2014 (P < .001) with a significant difference found between all groups during the study period (P ¼ .0004). ADC increased more than AVIC; however, both significantly increased over time (P ¼ .0002 and P ¼ .014, respectively). A mean $3326.63 increase in ADC was observed from 2012 to 2014 (P < .0001). AVIC only increased by an average $392.86 during the study period (P ¼ .01). This increased cost was observed in the context of a higher CMI and longer LOS. CMI increased from 2.25 in 2012 to 2.53 in 2014 (P ¼ .006). LOS increased by a mean 1.02 days between 2012 and 2014 (P ¼ .016), and significantly during the study period overall (P ¼ .018). After adjusting for CMI, LOS increases by only 0.61 days between 2012 and 2014 (P ¼ .07). In a final regression model, THC is independently predicted by comanagement, CMI, and LOS. After adjusting for CMI and LOS, the increase in THC because of comanagement (2012 vs 2014) accounts for only $4073.08 of the total increase (P < .001). During this time, 30-day readmission rates decreased by w7% (P ¼ .005), while related 30-day readmission rates decreased by w2% (P ¼ .32). Physician contribution margin remained unchanged over the 3-year period (P ¼ .76). The most prevalent diagnosis-related group was consistent across all years. Variation in the principal diagnosis code was observed with the prevalence of circulatory disorders because of type II diabetes replacing atherosclerosis with gangrene as the most prevalent diagnosis in 2013 and 2014 compared with 2012. Conclusions: In-hospital cost is significantly higher since the start of the HCS. This surge may relate to increased CMI, LOS, and improved coding. This increase in cost may be justified as we have observed sustained reduction in in-hospital mortality and slightly improved readmission rates. (J Vasc Surg 2016;-:1-7.)
The implementation of hospitalist comanagement services in comanaging surgical inpatients is a relatively recent phenomenon. Nonetheless, statistics show that such cooperation between medical physicians and
From the Division of Vascular Surgery, Department of Surgery, Icahn School of Medicine at Mount Sinai, New Yorka; and the Division of Vascular Surgery, Department of Surgery, University of Rochester Medical Center, Rochester.b Author conflict of interest: none. Presented as an oral presentation at the Forty-fourth Annual Symposium of the Society for Clinical Vascular Surgery, Las Vegas, Nev, March 12-16, 2016. Correspondence: Rami O. Tadros, MD, Division of Vascular Surgery, Department of Surgery, The Icahn School of Medicine at Mount Sinai, 1425 Madison Ave, 4th floor, New York, NY 10029 (e-mail: [email protected]). The editors and reviewers of this article have no relevant financial relationships to disclose per the JVS policy that requires reviewers to decline review of any manuscript for which they may have a conflict of interest. 0741-5214 Copyright Ó 2016 by the Society for Vascular Surgery. Published by Elsevier Inc. http://dx.doi.org/10.1016/j.jvs.2016.09.042
surgeons in the joint care and oversight of surgical patients may be justified. According to the Centers for Disease Control and Prevention, 51.4 million surgical procedures were performed in 2010.1 Further, the American College of Surgeons National Surgical Quality Improvement Program estimates that a single postoperative complication increases the total cost of care by 54%.2 It additionally shows that a hospital’s average profit margin is drastically reduced, from 23% to only 3.4%, for patients with complications.2 As the number of surgeries performed each year increases, so does the prevalence of postoperative complications and subsequently, costs to the hospital. The introduction of hospitalists to a surgical comanagement service allows for coordination of medical care and more oversight of patients during their postoperative course of treatment. Comanagement may prove essential to the prevention of postoperative complications, the reduction of in-hospital mortality rates, and ultimately, the containment of hospital costs. 1
2
Tadros et al
Journal of Vascular Surgery ---
A vascular surgery hospitalist comanagement service was first implemented at the Mount Sinai Medical Center and the Icahn School of Medicine at Mount Sinai in early 2013.3 Since then, the program has proven effective at managing complex medical comorbidities and improving in-hospital mortality rates among vascular surgery patients.4 In addition, comanagement was found to have a positive impact on vascular surgery pain outcomes as well as patient satisfaction.5 Because vascular surgery patients often face multiple medical comorbidities, their expected morbidity and mortality after surgery is much higher than the average patient.6 Therefore, this high-risk cohort of patients may uniquely benefit from a hospitalist comanagement service. Although the effect of a comanagement service on in-hospital mortality rates and postoperative pain outcomes has been assessed, the program’s impact on total hospital costs has yet to be studied. The main purpose of this study was to determine the impact of a collaborative, hospitalist-led approach in the vascular surgery service on the cost of care.
METHODS Patient population and data collection. This study was performed at the Mount Sinai Medical Center and Icahn School of Medicine at Mount Sinai, an urban tertiary care hospital and medical school located in New York City.4 A retrospective review of deidentified hospital administrative data was conducted on 1558 patients treated at Mount Sinai between 2012 and 2014. The Office for Excellence in Patient Care Reporting System through the Icahn School of Medicine at Mount Sinai, as well as institutional Department of Surgery finance records, provided the data used for this study. Institutional Review Board approval of the study was waived because patient specific information was not studied. Informed consent was also waived because the data collected were retrospective and deidentified.4 The vascular surgeon-hospitalist comanagement service was implemented in January 2013. Patients were divided into three cohorts: 516 treated in 2012 prior to comanagement, 525 treated in 2013 during the first year of comanagement, and 517 treated in 2014 during the second year of comanagement. Because of discrepancies between the total number of billing encounters obtained from The Department of Surgery financial data and the number of hospital discharges from the Office for Excellence in Patient Care Reporting System, select cases missing financial data were excluded from this study. For this reason, the numbers are slightly off from our previous study assessing the same comanagement program.4 Data were standardized for six vascular surgeon attendings who were present 2012-2014 and saw patients in all three cohorts. Two new attendings that joined the Mount Sinai faculty in July 2013 were
2016
excluded. Ten hospitalists also participated to make up the remainder of the comanagement service. The primary endpoints captured from the data collected were case mix index (CMI), contribution margin (CM), total hospital charges (THCs), actual direct costs (ADCs), actual variable indirect costs (AVICs), length of stay (LOS), and readmission rates (RARs). CMI is defined as the average relative weight of diagnosis-related groups (DRGs), for a given hospital or division; in our case, the calculation of CMI values was division- and physician-specific. Because DRGs classify inpatient stays into groups based on a patient’s age, sex, condition, procedure, and overall case complexity or comorbidity, our CMI reflects the clinical complexity of cases and provides an idea of how resources were allocated to treat patients in a given group. CM is the percent contribution, or dollar difference between total net charges to patients and relative variable costs; it can be used to measure the percentage that a given case is contributing to overall hospital profitability.7 THCs include all charges associated with a given patient’s admission that are billed to a third party payer. ADCs, or controllable costs, are those that can be easily identified within the department and are usually assigned to products and services.7 These included labor by physicians and other procedure-specific staff as well as fixed costs for supplies and equipment. AVICs typically include overhead, such as administration, information technology, and human resources, in addition to any variable costs that were required to care for and treat patients on a case-by-case basis. LOS describes the duration (in days) of a patient’s hospitalization, calculated by subtracting the day of admission from the day of discharge. Readmission rates will include general readmissions, including any admission to Mount Sinai Hospital within 30 days of discharge from the patient’s original hospital stay, and related readmissions, including any readmissions to Mount Sinai Hospital within 30 days of discharge for an event directly related to a patient’s primary admission. All Department of Surgery financial data were automatically adjusted for inflation prior to use in our analysis. Statistical analysis. All data were analyzed using SPSS software (IBM Corp, Armonk, NY). Normalized comparisons were made. Analysis of variance with post-hoc tests and t-tests were performed for each variable collected. Multiple linear regression models were used to assess the impact of confounding by CMI and LOS on the relationship between comanagement and THC. By creating a final regression model with CMI, LOS and comanagement as independent predictors of THC, we are able to estimate the association of each variable with THC when all other variables are held constant. An analysis of covariance was also used to assess CMI as a possible covariate in the relationship between comanagement and LOS. For regression analyses only, the data
Journal of Vascular Surgery Volume
-,
Number
Tadros et al
3
-
Fig 1. Comparison of unadjusted total hospital charges (THCs) over the 3-year study period, before and after the comanagement service was implemented. aSignificant in analysis of variance analysis to P < .01 when compared with the pre-comanagement period (2012).
was split into two groups: patients treated in 2012 (precomanagement) and patients treated in 2014 (postcomanagement). Results were considered significant to the P value of <.05 level. Outliers above a 40-day LOS were excluded from all analyses.
RESULTS After implementation of the hospitalist comanagement program, THCs rose by a mean $14,578.31 between 2012 and 2014 (2012: $52,374.76 vs 2014: $66,953.07; P < .001) (Fig 1). While the difference in THC over the study period as a whole was significant (P ¼ .0004), differences between 2012 and 2013 were not (2012: $52,374.76 vs 2013: $58,135.35; P ¼ .262). ADCs increased more over the study period than AVICs; however, both significantly increased over time (P ¼ .0002 and P ¼ .014, respectively). From 2012 to 2014, a mean $3326.63 increase in ADC was observed (P < .0001). During the same time period, AVIC only increased by a mean $392.86 (P ¼ .01). Normalized comparisons of these metrics remained statistically significant. These increased costs were observed in the context of a higher CMI over the 3-year period (P ¼ .006). A CMI of 2.25 observed in 2012 rose significantly to 2.47 in 2013 (P ¼ .039) and to 2.53 in 2014 (P ¼ .006). LOS also increased significantly during the study period (P ¼ .018), with a mean 1.02 days increase from 2012 to 2014 (2012: 5.28 days vs 2014: 6.30 days; P ¼ .016). To evaluate for possible confounding, an analysis of covariance was performed to predict LOS based on both implementation of the comanagement service (2012 vs 2014) and CMI. After adjusting for increased CMI, the impact of the comanagement service on LOS was no longer statistically significant (2012: 5.48 days vs 2014: 6.09 days;
P ¼ .07; Fig 2). Further, increased CMI was found to be independently associated with increased THC (P < .001). A linear regression analysis predicting LOS based on comanagement and CMI confirmed that for every 1.0-unit increase in CMI, LOS is expected to increase by 1.67 days (P < .001). THCs also increased in the context of a higher CMI, as well as longer LOS. As such, a hierarchical regression analysis was performed for all possible variables that could predict THCs (Table I). This table displays four separate regression models: in the first, only the impact of the comanagement program on THC was assessed; in the second, comanagement and CMI were examined; the third, comanagement and LOS; and the fourth, comanagement, CMI, and LOS, combined. In model 1, before controlling for any additional variables, THC is predicted to increase by $14,578.31 from 2012 to 2014 because of the comanagement program. In model 2, CMI is introduced into the equation as a possible covariate. After controlling for the increase in CMI over the study period, only $7830.96 of the total increase in THC can be attributed to implementation of the comanagement service (P ¼ .008). This prediction model accounts for 37% of the variance in THC (R2 ¼ .374). Model 3 examines the impact of the comanagement program and LOS on THC. After controlling for the increase in LOS across the study period, the increase in THC because of the comanagement program is significantly reduced, to $5511.91 (P < .001). Meanwhile, a 1-day increase in LOS independently results in a $7984.09 increase in THC (P < .001). This prediction model accounts for 82.7% of the total variance in THC (R2 ¼ .827), thereby demonstrating the significant impact of an increased LOS on THC. Our fourth and final model assessed the combined impact of comanagement, CMI and LOS on THC. In model 4, only $4073.08 of the increase in THC from 2012 to 2014 can be attributed to the comanagement program (P < .001). Although still a significant independent predictor of THC, this is largely reduced from the $14,578.31 increase observed between pre- and post-comanagement periods before controlling for any confounding factors. For every 1-day increase in LOS, model 4 predicts a $7063.68 increase in THC (P < .001) and for every 1.0-unit increase in CMI, a $9312.05 increase is expected (P < .001). This prediction model accounts for approximately 88% of the total variance in THC (R2 ¼ .882). According to semipartial correlations, introduction of the comanagement program predicts only 0.15% of the variance in THC, whereas CMI predicts 5.52% and LOS 53%. Regression analyses were also performed for 2012 and 2013. Although the difference in THC was not statistically significant during this time (2012: $52,374.76 vs 2013: $58,135.35; P ¼ .262), we anticipated similar findings to our regression analysis comparing 2012 and 2014. Although the increases in CMI and LOS were smaller
4
Journal of Vascular Surgery
Tadros et al
---
2016
Fig 2. Difference in length of stay (LOS) between 2012 and 2014 before and after controlling for case mix index (CMI). aSignificant in analysis of variance analysis to P < .01. Table I. Summary of hierarchical regression analysis for variables predicting total hospital charges (THCs) Model 1 Variables Comanagement (2012 vs 2014)
Model 2
B
P value
14,600
<.001
CMI
P value
B 7830
.008
24,000
<.001
LOS R2, %
1.5
Model 3
37.4
Model 4
B
P value
B
P value
5510
<.001
4070
<.001
9310
<.001
7980
<.001
7060
<.001
82.7
88.2
CMI, Case mix index; LOS, length of stay. Under the final model in the analysis (model 4), each 1.0-unit increase in CMI results in a $9310 increase in cost and each 1-day increase in LOS results in a $7060 increase in cost. The B value for comanagement demonstrates the contribution of comanagement to THCs in each of the four models; from model 1 to model 4, the incremental cost of comanagement decreases from $14,600 to $4070 after adjusting for other independent predictors of THC (CMI and LOS). Overall, the final model proves that comanagement was not the sole driver of increased costs and demonstrates the significant, independent effects of the comanagement program, CMI, and LOS on THCs. In model 1, the coefficient (B) reflects the change in THCs across groups (2012 to 2014). In models 2-4, the coefficient (B) for comanagement reflects the change in THCs from 2012 to 2014, after controlling for any other variables in the model. The coefficient (B) for CMI and LOS reflect the expected changes in THC for every 1.0-unit increase in CMI and 1 day increase in LOS, respectively.
between 2012 and 2013 (LOS: 5.28 to 5.60 days; P ¼ .64, CMI: 2.25 to 2.47; P ¼ .01), after controlling for both in a regression analysis, the difference in THC because of to comanagement decreases significantly (2012: $52,980.91 vs 2013: $53,898.25; P ¼ .37). According to the regression model, for every 1-day increase in LOS, THC is expected to increase by $6971.38, and for every 1.0-unit increase in CMI, THC will increase by $7900.05. Meanwhile, the difference in THC because of comanagement amounts to only $917.34 between 2012 and 2013 after controlling for LOS and CMI, down from a $5760.59 cost difference before performing this adjustment. Readmission to Mount Sinai Hospital after initial discharge was also calculated. Across the study period
and after implementation of the comanagement service, 30-day RARs decreased by w7% (2012: 23.1% vs 2014: 16.1%; P ¼ .005). Related 30-day RARs, which include any readmissions within 30 days of discharge that are directly related to the initial admission, decreased by almost 2%, although not statistically significant (2012: 11.5% vs 2014: 9.5%; P ¼ .32). Physician CM remained unchanged over the study period (P ¼ .76). The most prevalent DRG was consistent across all years during the study period (Table II). Some variation in the principal diagnosis code was observed with the prevalence of circulatory disorders because of type II diabetes replacing atherosclerosis with gangrene as the most prevalent diagnosis in 2013 and 2014, as compared with 2012 (Table III).
Journal of Vascular Surgery Volume
-,
Number
Tadros et al
5
-
Table II. Top three diagnosis-related group (DRG) codes by year 2012 Principal DRG code
Principle DRG description
2013 Principal DRG code
Principle DRG description
2014 Principle DRG code
Principle DRG description
1.
253
Other vascular procedures with CC
253
Other vascular procedures with CC
253
Other vascular procedures with CC
2.
254
Other vascular procedures without CC/MCC
238
Major cardiovascular procedures without MCC
238
Major cardiovascular procedures without MCC
3.
238
Major cardiovascular procedures without MCC
252
Other vascular procedures with MCC
252
Other vascular procedures with MCC
CC, Complication or comorbidity; MCC, major complication or comorbidity.
Table III. Top three diagnosis (DX) codes by year 2012 Principal DX code
Principal DX description
2013 Principal DX code
Principal DX description
2014 Principal DX code
250.7
2. 440.21 Atherosclerosis of native arteries of the extremities with intermittent claudication
440.24 Atherosclerosis of native arteries of the extremities with gangrene
250.72 Diabetes with peripheral circulatory disorders, type II or unspecified type, controlled
440.21 Atherosclerosis of native arteries Diabetes with peripheral of the extremities with circulatory disorders, type II or unspecified type, not stated as intermittent claudication uncontrolled
440.24 Atherosclerosis of native arteries of the extremities with gangrene
3. 250.7
Diabetes with peripheral 250.7 circulatory disorders, type II or unspecified type, not stated as uncontrolled
Principal DX description
1. 440.24 Atherosclerosis of native arteries of the extremities with gangrene
DISCUSSION Current literature on the effects of hospitalist comanagement programs in vascular surgery patients is limited.3,4 Further, there is even less information available on the costs associated with the implementation of a comanagement program in this particular patient demographic. The majority of studies that have examined the effects of hospitalist comanagement services focus on orthopedic and neurology patients.8-11 Unfortunately, it is not always possible to draw comparisons to these studies, as vascular surgery inpatients tend to differ from other types of surgical patients. Patients undergoing vascular surgery are often plagued by old age, atherosclerotic disease, and a wide range of comorbidities, including diabetes, renal failure, and heart disease, among others.6 Such comorbidities complicate the management of postoperative vascular surgery patients, as well as increase the likelihood of perioperative morbidity and mortality.4 Therefore, it is possible that the results observed in this high-risk group of patients will be inconsistent with those of patients having fewer concomitant illnesses as previously studied in the literature. Our study observed higher THCs after the hospitalist comanagement service was initiated in January 2013;
Diabetes with peripheral circulatory disorders, type II or unspecified type, not stated as uncontrolled
these effects extended well into 2014. However, as results of the regression analyses demonstrate, implementation of the comanagement service only accounts for a portion of the increase in THC observed over the study period. The increase in CMI over the study period, and subsequent increase in length of hospital stay, also significantly contributed to the rise in THC. According to the final regression model, for every 1.0-unit increase in CMI and 1-day increase in LOS, a $9625.22 and $7210.31 increase in THC are expected, respectively. Although CMI was originally designed to calculate hospital payments, it is often used as a tracker of disease severity.12 A high CMI may be indicative of a large volume of complications and comorbidities being treated. Because the average CMI was significantly higher in 2013 and 2014, increases in THCs during this time are expected because of increased resources necessary to treat this more medically complex group of patients. Similarly, patients with a higher CMI are likely to have a longer hospital stay. In turn, as our regression models demonstrate, increased LOS leads to significantly greater hospital costs. We have reason to believe that the increase in CMI, and subsequent increase in LOS, across the study period was attributable, in part, to the creation of the Mount Sinai Health System in 2013. This merger
6
Journal of Vascular Surgery
Tadros et al
---
combined the Mount Sinai Hospital with many community-oriented hospitals part of Continuum Health Partners. With the creation of a conglomerate health system came ease in transferring some of the most complex cases and acutely sick patients to the Mount Sinai Hospital from other Mount Sinai sites, thereby elevating CMI, LOS, and total costs to the hospital. Improved documentation and coding over the study period may have also contributed to the rise in THC by allowing for the capture of a larger number of charges and diagnoses. Years of undercharging combined with the movement toward more accurate billing may have intensified the increase in THC, as billing practices improved at the same time that higher DRGs, for more complex cases, were being reimbursed. It is also always possible that the hospitalists participating in comanagement were more effective coders and, as such, billed more accurately for charges incurred during patients’ hospital stays; this would contribute to the increase in CMI observed during the comanagement period when hospitalists were in use. Unfortunately, it is difficult to ascertain the numeric effect that improved billing and codingdby all members of the comanagement team or by hospitalists, alonedhad on THC, but we can only assume that it was a contributory factor. Other studies assessing the effects of hospitalist comanagement programs on the cost of care are highly variable. Auerbach et al8 examined the impact of comanagement between neurosurgeons and hospitalists and observed decreased costs, but found no significant differences in LOS or patient mortality. A study on geriatric hip-fracture patients observed a significant decrease in LOS and THCs per patient, yet observed no improvement in in-hospital mortality rates.9 A population-based cohort study examining a national sample of Medicare patients went further to assess the impact of hospitalist care on costs after discharge. Although a decrease in LOS and hospital charges was initially observed, the study found that Medicare costs increased significantly in the 30 days after discharge.13 An increase in subsequent emergency department visits and 30-day RARs suggest that initial cost savings because of comanagement were merely offset by higher medical utilization and costs after discharge from the hospital. The impact of hospitalist comanagement at Northwestern University Feinberg School of Medicine examined recently by Hinami et al14 found results more similar to our own. This 854-bed urban care center used hospitalists in conjunction with surgeons from 17 different surgical specialties. When using unadjusted comparisons, both LOS and cost were observed to be higher in the comanaged compared with noncomanaged cohort. It was only after performing a risk adjustment that the mean LOS decreased, although the cost of care did not significantly differ between
2016
groups. To explain the higher unadjusted costs that were observed, Hinami et al14 suggests extra testing and care administered by hospitalists. These costs appeared to be warranted, however, as comanaged patients experienced significantly fewer in-hospital deaths. Despite mixed findings in the literature on the effect of hospitalist comanagement services on hospital costs, we firmly believe that the higher costs observed in our study may be justified. Vascular surgery patients represent a unique cohort that may benefit drastically from the effects of a hospitalist service. Despite an increase in CMI in the comanaged group of patients, Tadros et al4 observed a decrease in in-hospital mortality rates from 1.75% to 0.37% after implementation of the vascular surgeon-hospitalist comanagement service. Decreased average mortality rates in this high risk, medically complex group of comanaged patients demonstrate the effectiveness of a comanagement service in drastically improving in-hospital mortality. Study limitations. First, this study was conducted at an urban, single-institution and academic medical center, so it is possible that the results are not generalizable to the national public. Further, the nature of this study was observational, as patients who were treated in 2012 before comanagement were merely compared with those treated in 2013 and 2014, when the comanagement service was in use. Because of this, we are only able to perceive associations between groups, as randomization did not occur. Individuals were not assigned to either comanagement or noncomanagement interventions, so we were not able to control for a variety of factors that could occur, such as increased CMI in the patients treated under comanagement. With statistical adjustments, however, we were able to control for some of the sampling bias that occurred as a result of the nonrandomized nature of this study. In addition, our cost data lacked information regarding where the changes in cost resided (ie, more laboratory tests, X rays, medications). We also had limited access to data on reimbursement to the hospital, rendering us unable to assess whether or not the extra costs incurred during the comanagement period were offset by higher reimbursements and total revenue. Lastly, no patientspecific data was used in the conduct of this study, so the scope of the analysis was subsequently limited. Because of this, we were unable to examine any measures that required access to patient charts, specifically, including whether the addition of hospitalist physicians had any profound impact on the transition of care to patients’ primary care providers or the frequency of readmissions to outside hospitals.
CONCLUSIONS Overall, our findings have demonstrated that although THCs and ADCs increased over the study period, these
Journal of Vascular Surgery Volume
-,
Number
Tadros et al
7
-
findings were undoubtedly confounded by the increase in both CMI and LOS. After adjusting for these factors, the difference in THC between non-comanaged and comanaged groups was not nearly as high. Our previous work demonstrated an improved in-hospital mortality rate among comanaged vascular surgery inpatients.4 The fact that these improved in-hospital mortality rates were observed in the context of a higher CMI adds further credence to the benefits of comanagement.
4.
5.
6. 7.
AUTHOR CONTRIBUTIONS Conception and design: RT, MT, MS, DK Analysis and interpretation: RT, MT, PF, MS, AV Data collection: RT, MT, CP Writing the article: RT, MT Critical revision of the article: RT, MT, PF, CP, DK, AV, MM Final approval of the article: RT, MT, PF, DK, MM Statistical analysis: RT, MT, MS, CP Obtained funding: Not applicable Overall responsibility: RT RT and MT contributed equally to this article.
8.
9.
10.
11.
REFERENCES 1. Centers for Disease Control and Prevention. National Center for Health Statistics [Internet]. Atlanta (GA): Centers for Disease Control and Prevention. National Hospital Discharge Survey, 2010 [updated April 29, 2015]. Available at: www.cdc. gov/nchs/fastats/inpatient-surgery.htm Accessed March 4, 2016. 2. American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) [Internet]. Preventing complications, reducing costs, improving surgical care. Chicago: American College of Surgeons. Available at: https:// www.facs.org/w/media/files/quality%20programs/nsqip/nsqip infobook1012.ashx. Accessed April 1, 2016. 3. FOJP Service Corporation. The expanding role of hospital medicine and the co-management of patients. The
12.
13.
14.
Quarterly Journal for Health Care Practice and Risk Management 2013;21:1-20. Tadros RO, Faries PL, Malik R, Vouyouka AG, Ting W, Dunn A, et al. The effect of a hospitalist comanagement service on vascular surgery inpatients. J Vasc Surg 2015;61:1550-5. Png CY, Faries PL, Qian LY, Lee IT, Chander R, Finlay DJ, et al. Vascular surgery pain outcomes improved by implementing hospitalist comanagement service. J Pain Manage Med 2016;2:1-3. Crimi E, Hill CC. Postoperative ICU management of vascular surgery patients. Anesthesiol Clin 2014;32:735-57. Herkimer AG. Cost characteristics and behavior. In: Davies M, editor. Understanding hospital financial management. 2nd ed. Rockville, MD: Aspen Publishers, Inc; 1978. Auerbach AD, Wachter RM, Cheng HQ, Maselli J, McDermott M, Vittinghoff E, et al. Comangement of surgical patients between neurosurgeons and hospitalists. Arch Intern Med 2010;170:2004-10. Della Rocca GJ, Moylan KC, Christ BD, Volgas DA, Stannard JP, Mehr DR. Comangement of geriatric patients with hip fractures: a retrospective, controlled, cohort study. Geriatr Orthop Surg Rehabil 2013;4:10-5. Pinzur MS, Gurza E, Kristopaitis T, Monson R, Wall MJ, Porter A, et al. Hospitalist-orthopedic comanagement of high-risk patients undergoing lower extremity reconstruction surgery. Orthopedics 2009;32:495. Swart E, Vasudeva E, Makhni EC, Macaulay W, Bozic KJ. Dedicated perioperative hip fracture comanagement programs are cost-effective in high-volume centers: an economic analysis. Clin Orthop Relat Res 2016;474:222-33. Mendez CM, Harrington DW, Christenson P, Spellberg B. Impact of hospital variables on case mix index as a marker of disease severity. Popul Health Manage 2014;17:28-34. Kuo Y, Goodwin JS. Association of hospitalist care with medical utilization after discharge: evidence of cost shift from a cohort study. Ann Intern Med 2011;155:152-9. Hinami K, Feinglass J, Ferranti DE, Williams MV. Potential role of comanagement in “rescue” of surgical patients. Am J Manag Care 2011;17:333-9.
Submitted Jun 8, 2016; accepted Sep 21, 2016.
Vascular surgeon-hospitalist comanagement improves in-hospital mortality at the expense of increased in-hospital cost Rami O. Tadros, MD,a Melissa L. Tardiff, BA,a Peter L. Faries, MD,a Michael Stoner, MD,b Chien Yi M. Png, BA,a David Kaplan, MPA,a Ageliki G. Vouyouka, MD,a and Michael L. Marin, MD,a New York and Rochester, NY
ABSTRACT Objective: We have shown that vascular surgeon- hospitalist co management resulted in improved in-hospital mortality rates. We now aim to assess the impact of the hospitalist co management service (HCS) on healthcare cost. Methods: A total of 1558 patients were divided into three cohorts and compared: 516 in 2012, 525 in 2013, and 517 in 2014. The HCS began in January 2013. Data were standardized for six vascular surgeons that were present 2012-2014. New attendings were excluded. Ten hospitalists participated. Case mix index (CMI), contribution margin, total hospital charges (THCs), length of stay (LOS), actual direct costs (ADCs), and actual variable indirect costs (AVICs) were compared. Analysis of variance with post-hoc tests, t-tests, and linear regressions were performed. Results: THC rose by a mean difference of $14,578.31 between 2012 and 2014 (P < .001) with a significant difference found between all groups during the study period (P ¼ .0004). ADC increased more than AVIC; however, both significantly increased over time (P ¼ .0002 and P ¼ .014, respectively). A mean $3326.63 increase in ADC was observed from 2012 to 2014 (P < .0001). AVIC only increased by an average $392.86 during the study period (P ¼ .01). This increased cost was observed in the context of a higher CMI and longer LOS. CMI increased from 2.25 in 2012 to 2.53 in 2014 (P ¼ .006). LOS increased by a mean 1.02 days between 2012 and 2014 (P ¼ .016), and significantly during the study period overall (P ¼ .018). After adjusting for CMI, LOS increases by only 0.61 days between 2012 and 2014 (P ¼ .07). In a final regression model, THC is independently predicted by comanagement, CMI, and LOS. After adjusting for CMI and LOS, the increase in THC because of comanagement (2012 vs 2014) accounts for only $4073.08 of the total increase (P < .001). During this time, 30-day readmission rates decreased by w7% (P ¼ .005), while related 30-day readmission rates decreased by w2% (P ¼ .32). Physician contribution margin remained unchanged over the 3-year period (P ¼ .76). The most prevalent diagnosis-related group was consistent across all years. Variation in the principal diagnosis code was observed with the prevalence of circulatory disorders because of type II diabetes replacing atherosclerosis with gangrene as the most prevalent diagnosis in 2013 and 2014 compared with 2012. Conclusions: In-hospital cost is significantly higher since the start of the HCS. This surge may relate to increased CMI, LOS, and improved coding. This increase in cost may be justified as we have observed sustained reduction in in-hospital mortality and slightly improved readmission rates. (J Vasc Surg 2016;-:1-7.)
The implementation of hospitalist comanagement services in comanaging surgical inpatients is a relatively recent phenomenon. Nonetheless, statistics show that such cooperation between medical physicians and
From the Division of Vascular Surgery, Department of Surgery, Icahn School of Medicine at Mount Sinai, New Yorka; and the Division of Vascular Surgery, Department of Surgery, University of Rochester Medical Center, Rochester.b Author conflict of interest: none. Presented as an oral presentation at the Forty-fourth Annual Symposium of the Society for Clinical Vascular Surgery, Las Vegas, Nev, March 12-16, 2016. Correspondence: Rami O. Tadros, MD, Division of Vascular Surgery, Department of Surgery, The Icahn School of Medicine at Mount Sinai, 1425 Madison Ave, 4th floor, New York, NY 10029 (e-mail: [email protected]). The editors and reviewers of this article have no relevant financial relationships to disclose per the JVS policy that requires reviewers to decline review of any manuscript for which they may have a conflict of interest. 0741-5214 Copyright Ó 2016 by the Society for Vascular Surgery. Published by Elsevier Inc. http://dx.doi.org/10.1016/j.jvs.2016.09.042
surgeons in the joint care and oversight of surgical patients may be justified. According to the Centers for Disease Control and Prevention, 51.4 million surgical procedures were performed in 2010.1 Further, the American College of Surgeons National Surgical Quality Improvement Program estimates that a single postoperative complication increases the total cost of care by 54%.2 It additionally shows that a hospital’s average profit margin is drastically reduced, from 23% to only 3.4%, for patients with complications.2 As the number of surgeries performed each year increases, so does the prevalence of postoperative complications and subsequently, costs to the hospital. The introduction of hospitalists to a surgical comanagement service allows for coordination of medical care and more oversight of patients during their postoperative course of treatment. Comanagement may prove essential to the prevention of postoperative complications, the reduction of in-hospital mortality rates, and ultimately, the containment of hospital costs. 1
2
Tadros et al
Journal of Vascular Surgery ---
A vascular surgery hospitalist comanagement service was first implemented at the Mount Sinai Medical Center and the Icahn School of Medicine at Mount Sinai in early 2013.3 Since then, the program has proven effective at managing complex medical comorbidities and improving in-hospital mortality rates among vascular surgery patients.4 In addition, comanagement was found to have a positive impact on vascular surgery pain outcomes as well as patient satisfaction.5 Because vascular surgery patients often face multiple medical comorbidities, their expected morbidity and mortality after surgery is much higher than the average patient.6 Therefore, this high-risk cohort of patients may uniquely benefit from a hospitalist comanagement service. Although the effect of a comanagement service on in-hospital mortality rates and postoperative pain outcomes has been assessed, the program’s impact on total hospital costs has yet to be studied. The main purpose of this study was to determine the impact of a collaborative, hospitalist-led approach in the vascular surgery service on the cost of care.
METHODS Patient population and data collection. This study was performed at the Mount Sinai Medical Center and Icahn School of Medicine at Mount Sinai, an urban tertiary care hospital and medical school located in New York City.4 A retrospective review of deidentified hospital administrative data was conducted on 1558 patients treated at Mount Sinai between 2012 and 2014. The Office for Excellence in Patient Care Reporting System through the Icahn School of Medicine at Mount Sinai, as well as institutional Department of Surgery finance records, provided the data used for this study. Institutional Review Board approval of the study was waived because patient specific information was not studied. Informed consent was also waived because the data collected were retrospective and deidentified.4 The vascular surgeon-hospitalist comanagement service was implemented in January 2013. Patients were divided into three cohorts: 516 treated in 2012 prior to comanagement, 525 treated in 2013 during the first year of comanagement, and 517 treated in 2014 during the second year of comanagement. Because of discrepancies between the total number of billing encounters obtained from The Department of Surgery financial data and the number of hospital discharges from the Office for Excellence in Patient Care Reporting System, select cases missing financial data were excluded from this study. For this reason, the numbers are slightly off from our previous study assessing the same comanagement program.4 Data were standardized for six vascular surgeon attendings who were present 2012-2014 and saw patients in all three cohorts. Two new attendings that joined the Mount Sinai faculty in July 2013 were
2016
excluded. Ten hospitalists also participated to make up the remainder of the comanagement service. The primary endpoints captured from the data collected were case mix index (CMI), contribution margin (CM), total hospital charges (THCs), actual direct costs (ADCs), actual variable indirect costs (AVICs), length of stay (LOS), and readmission rates (RARs). CMI is defined as the average relative weight of diagnosis-related groups (DRGs), for a given hospital or division; in our case, the calculation of CMI values was division- and physician-specific. Because DRGs classify inpatient stays into groups based on a patient’s age, sex, condition, procedure, and overall case complexity or comorbidity, our CMI reflects the clinical complexity of cases and provides an idea of how resources were allocated to treat patients in a given group. CM is the percent contribution, or dollar difference between total net charges to patients and relative variable costs; it can be used to measure the percentage that a given case is contributing to overall hospital profitability.7 THCs include all charges associated with a given patient’s admission that are billed to a third party payer. ADCs, or controllable costs, are those that can be easily identified within the department and are usually assigned to products and services.7 These included labor by physicians and other procedure-specific staff as well as fixed costs for supplies and equipment. AVICs typically include overhead, such as administration, information technology, and human resources, in addition to any variable costs that were required to care for and treat patients on a case-by-case basis. LOS describes the duration (in days) of a patient’s hospitalization, calculated by subtracting the day of admission from the day of discharge. Readmission rates will include general readmissions, including any admission to Mount Sinai Hospital within 30 days of discharge from the patient’s original hospital stay, and related readmissions, including any readmissions to Mount Sinai Hospital within 30 days of discharge for an event directly related to a patient’s primary admission. All Department of Surgery financial data were automatically adjusted for inflation prior to use in our analysis. Statistical analysis. All data were analyzed using SPSS software (IBM Corp, Armonk, NY). Normalized comparisons were made. Analysis of variance with post-hoc tests and t-tests were performed for each variable collected. Multiple linear regression models were used to assess the impact of confounding by CMI and LOS on the relationship between comanagement and THC. By creating a final regression model with CMI, LOS and comanagement as independent predictors of THC, we are able to estimate the association of each variable with THC when all other variables are held constant. An analysis of covariance was also used to assess CMI as a possible covariate in the relationship between comanagement and LOS. For regression analyses only, the data
Journal of Vascular Surgery Volume
-,
Number
Tadros et al
3
-
Fig 1. Comparison of unadjusted total hospital charges (THCs) over the 3-year study period, before and after the comanagement service was implemented. aSignificant in analysis of variance analysis to P < .01 when compared with the pre-comanagement period (2012).
was split into two groups: patients treated in 2012 (precomanagement) and patients treated in 2014 (postcomanagement). Results were considered significant to the P value of <.05 level. Outliers above a 40-day LOS were excluded from all analyses.
RESULTS After implementation of the hospitalist comanagement program, THCs rose by a mean $14,578.31 between 2012 and 2014 (2012: $52,374.76 vs 2014: $66,953.07; P < .001) (Fig 1). While the difference in THC over the study period as a whole was significant (P ¼ .0004), differences between 2012 and 2013 were not (2012: $52,374.76 vs 2013: $58,135.35; P ¼ .262). ADCs increased more over the study period than AVICs; however, both significantly increased over time (P ¼ .0002 and P ¼ .014, respectively). From 2012 to 2014, a mean $3326.63 increase in ADC was observed (P < .0001). During the same time period, AVIC only increased by a mean $392.86 (P ¼ .01). Normalized comparisons of these metrics remained statistically significant. These increased costs were observed in the context of a higher CMI over the 3-year period (P ¼ .006). A CMI of 2.25 observed in 2012 rose significantly to 2.47 in 2013 (P ¼ .039) and to 2.53 in 2014 (P ¼ .006). LOS also increased significantly during the study period (P ¼ .018), with a mean 1.02 days increase from 2012 to 2014 (2012: 5.28 days vs 2014: 6.30 days; P ¼ .016). To evaluate for possible confounding, an analysis of covariance was performed to predict LOS based on both implementation of the comanagement service (2012 vs 2014) and CMI. After adjusting for increased CMI, the impact of the comanagement service on LOS was no longer statistically significant (2012: 5.48 days vs 2014: 6.09 days;
P ¼ .07; Fig 2). Further, increased CMI was found to be independently associated with increased THC (P < .001). A linear regression analysis predicting LOS based on comanagement and CMI confirmed that for every 1.0-unit increase in CMI, LOS is expected to increase by 1.67 days (P < .001). THCs also increased in the context of a higher CMI, as well as longer LOS. As such, a hierarchical regression analysis was performed for all possible variables that could predict THCs (Table I). This table displays four separate regression models: in the first, only the impact of the comanagement program on THC was assessed; in the second, comanagement and CMI were examined; the third, comanagement and LOS; and the fourth, comanagement, CMI, and LOS, combined. In model 1, before controlling for any additional variables, THC is predicted to increase by $14,578.31 from 2012 to 2014 because of the comanagement program. In model 2, CMI is introduced into the equation as a possible covariate. After controlling for the increase in CMI over the study period, only $7830.96 of the total increase in THC can be attributed to implementation of the comanagement service (P ¼ .008). This prediction model accounts for 37% of the variance in THC (R2 ¼ .374). Model 3 examines the impact of the comanagement program and LOS on THC. After controlling for the increase in LOS across the study period, the increase in THC because of the comanagement program is significantly reduced, to $5511.91 (P < .001). Meanwhile, a 1-day increase in LOS independently results in a $7984.09 increase in THC (P < .001). This prediction model accounts for 82.7% of the total variance in THC (R2 ¼ .827), thereby demonstrating the significant impact of an increased LOS on THC. Our fourth and final model assessed the combined impact of comanagement, CMI and LOS on THC. In model 4, only $4073.08 of the increase in THC from 2012 to 2014 can be attributed to the comanagement program (P < .001). Although still a significant independent predictor of THC, this is largely reduced from the $14,578.31 increase observed between pre- and post-comanagement periods before controlling for any confounding factors. For every 1-day increase in LOS, model 4 predicts a $7063.68 increase in THC (P < .001) and for every 1.0-unit increase in CMI, a $9312.05 increase is expected (P < .001). This prediction model accounts for approximately 88% of the total variance in THC (R2 ¼ .882). According to semipartial correlations, introduction of the comanagement program predicts only 0.15% of the variance in THC, whereas CMI predicts 5.52% and LOS 53%. Regression analyses were also performed for 2012 and 2013. Although the difference in THC was not statistically significant during this time (2012: $52,374.76 vs 2013: $58,135.35; P ¼ .262), we anticipated similar findings to our regression analysis comparing 2012 and 2014. Although the increases in CMI and LOS were smaller
4
Journal of Vascular Surgery
Tadros et al
---
2016
Fig 2. Difference in length of stay (LOS) between 2012 and 2014 before and after controlling for case mix index (CMI). aSignificant in analysis of variance analysis to P < .01. Table I. Summary of hierarchical regression analysis for variables predicting total hospital charges (THCs) Model 1 Variables Comanagement (2012 vs 2014)
Model 2
B
P value
14,600
<.001
CMI
P value
B 7830
.008
24,000
<.001
LOS R2, %
1.5
Model 3
37.4
Model 4
B
P value
B
P value
5510
<.001
4070
<.001
9310
<.001
7980
<.001
7060
<.001
82.7
88.2
CMI, Case mix index; LOS, length of stay. Under the final model in the analysis (model 4), each 1.0-unit increase in CMI results in a $9310 increase in cost and each 1-day increase in LOS results in a $7060 increase in cost. The B value for comanagement demonstrates the contribution of comanagement to THCs in each of the four models; from model 1 to model 4, the incremental cost of comanagement decreases from $14,600 to $4070 after adjusting for other independent predictors of THC (CMI and LOS). Overall, the final model proves that comanagement was not the sole driver of increased costs and demonstrates the significant, independent effects of the comanagement program, CMI, and LOS on THCs. In model 1, the coefficient (B) reflects the change in THCs across groups (2012 to 2014). In models 2-4, the coefficient (B) for comanagement reflects the change in THCs from 2012 to 2014, after controlling for any other variables in the model. The coefficient (B) for CMI and LOS reflect the expected changes in THC for every 1.0-unit increase in CMI and 1 day increase in LOS, respectively.
between 2012 and 2013 (LOS: 5.28 to 5.60 days; P ¼ .64, CMI: 2.25 to 2.47; P ¼ .01), after controlling for both in a regression analysis, the difference in THC because of to comanagement decreases significantly (2012: $52,980.91 vs 2013: $53,898.25; P ¼ .37). According to the regression model, for every 1-day increase in LOS, THC is expected to increase by $6971.38, and for every 1.0-unit increase in CMI, THC will increase by $7900.05. Meanwhile, the difference in THC because of comanagement amounts to only $917.34 between 2012 and 2013 after controlling for LOS and CMI, down from a $5760.59 cost difference before performing this adjustment. Readmission to Mount Sinai Hospital after initial discharge was also calculated. Across the study period
and after implementation of the comanagement service, 30-day RARs decreased by w7% (2012: 23.1% vs 2014: 16.1%; P ¼ .005). Related 30-day RARs, which include any readmissions within 30 days of discharge that are directly related to the initial admission, decreased by almost 2%, although not statistically significant (2012: 11.5% vs 2014: 9.5%; P ¼ .32). Physician CM remained unchanged over the study period (P ¼ .76). The most prevalent DRG was consistent across all years during the study period (Table II). Some variation in the principal diagnosis code was observed with the prevalence of circulatory disorders because of type II diabetes replacing atherosclerosis with gangrene as the most prevalent diagnosis in 2013 and 2014, as compared with 2012 (Table III).
Journal of Vascular Surgery Volume
-,
Number
Tadros et al
5
-
Table II. Top three diagnosis-related group (DRG) codes by year 2012 Principal DRG code
Principle DRG description
2013 Principal DRG code
Principle DRG description
2014 Principle DRG code
Principle DRG description
1.
253
Other vascular procedures with CC
253
Other vascular procedures with CC
253
Other vascular procedures with CC
2.
254
Other vascular procedures without CC/MCC
238
Major cardiovascular procedures without MCC
238
Major cardiovascular procedures without MCC
3.
238
Major cardiovascular procedures without MCC
252
Other vascular procedures with MCC
252
Other vascular procedures with MCC
CC, Complication or comorbidity; MCC, major complication or comorbidity.
Table III. Top three diagnosis (DX) codes by year 2012 Principal DX code
Principal DX description
2013 Principal DX code
Principal DX description
2014 Principal DX code
250.7
2. 440.21 Atherosclerosis of native arteries of the extremities with intermittent claudication
440.24 Atherosclerosis of native arteries of the extremities with gangrene
250.72 Diabetes with peripheral circulatory disorders, type II or unspecified type, controlled
440.21 Atherosclerosis of native arteries Diabetes with peripheral of the extremities with circulatory disorders, type II or unspecified type, not stated as intermittent claudication uncontrolled
440.24 Atherosclerosis of native arteries of the extremities with gangrene
3. 250.7
Diabetes with peripheral 250.7 circulatory disorders, type II or unspecified type, not stated as uncontrolled
Principal DX description
1. 440.24 Atherosclerosis of native arteries of the extremities with gangrene
DISCUSSION Current literature on the effects of hospitalist comanagement programs in vascular surgery patients is limited.3,4 Further, there is even less information available on the costs associated with the implementation of a comanagement program in this particular patient demographic. The majority of studies that have examined the effects of hospitalist comanagement services focus on orthopedic and neurology patients.8-11 Unfortunately, it is not always possible to draw comparisons to these studies, as vascular surgery inpatients tend to differ from other types of surgical patients. Patients undergoing vascular surgery are often plagued by old age, atherosclerotic disease, and a wide range of comorbidities, including diabetes, renal failure, and heart disease, among others.6 Such comorbidities complicate the management of postoperative vascular surgery patients, as well as increase the likelihood of perioperative morbidity and mortality.4 Therefore, it is possible that the results observed in this high-risk group of patients will be inconsistent with those of patients having fewer concomitant illnesses as previously studied in the literature. Our study observed higher THCs after the hospitalist comanagement service was initiated in January 2013;
Diabetes with peripheral circulatory disorders, type II or unspecified type, not stated as uncontrolled
these effects extended well into 2014. However, as results of the regression analyses demonstrate, implementation of the comanagement service only accounts for a portion of the increase in THC observed over the study period. The increase in CMI over the study period, and subsequent increase in length of hospital stay, also significantly contributed to the rise in THC. According to the final regression model, for every 1.0-unit increase in CMI and 1-day increase in LOS, a $9625.22 and $7210.31 increase in THC are expected, respectively. Although CMI was originally designed to calculate hospital payments, it is often used as a tracker of disease severity.12 A high CMI may be indicative of a large volume of complications and comorbidities being treated. Because the average CMI was significantly higher in 2013 and 2014, increases in THCs during this time are expected because of increased resources necessary to treat this more medically complex group of patients. Similarly, patients with a higher CMI are likely to have a longer hospital stay. In turn, as our regression models demonstrate, increased LOS leads to significantly greater hospital costs. We have reason to believe that the increase in CMI, and subsequent increase in LOS, across the study period was attributable, in part, to the creation of the Mount Sinai Health System in 2013. This merger
6
Journal of Vascular Surgery
Tadros et al
---
combined the Mount Sinai Hospital with many community-oriented hospitals part of Continuum Health Partners. With the creation of a conglomerate health system came ease in transferring some of the most complex cases and acutely sick patients to the Mount Sinai Hospital from other Mount Sinai sites, thereby elevating CMI, LOS, and total costs to the hospital. Improved documentation and coding over the study period may have also contributed to the rise in THC by allowing for the capture of a larger number of charges and diagnoses. Years of undercharging combined with the movement toward more accurate billing may have intensified the increase in THC, as billing practices improved at the same time that higher DRGs, for more complex cases, were being reimbursed. It is also always possible that the hospitalists participating in comanagement were more effective coders and, as such, billed more accurately for charges incurred during patients’ hospital stays; this would contribute to the increase in CMI observed during the comanagement period when hospitalists were in use. Unfortunately, it is difficult to ascertain the numeric effect that improved billing and codingdby all members of the comanagement team or by hospitalists, alonedhad on THC, but we can only assume that it was a contributory factor. Other studies assessing the effects of hospitalist comanagement programs on the cost of care are highly variable. Auerbach et al8 examined the impact of comanagement between neurosurgeons and hospitalists and observed decreased costs, but found no significant differences in LOS or patient mortality. A study on geriatric hip-fracture patients observed a significant decrease in LOS and THCs per patient, yet observed no improvement in in-hospital mortality rates.9 A population-based cohort study examining a national sample of Medicare patients went further to assess the impact of hospitalist care on costs after discharge. Although a decrease in LOS and hospital charges was initially observed, the study found that Medicare costs increased significantly in the 30 days after discharge.13 An increase in subsequent emergency department visits and 30-day RARs suggest that initial cost savings because of comanagement were merely offset by higher medical utilization and costs after discharge from the hospital. The impact of hospitalist comanagement at Northwestern University Feinberg School of Medicine examined recently by Hinami et al14 found results more similar to our own. This 854-bed urban care center used hospitalists in conjunction with surgeons from 17 different surgical specialties. When using unadjusted comparisons, both LOS and cost were observed to be higher in the comanaged compared with noncomanaged cohort. It was only after performing a risk adjustment that the mean LOS decreased, although the cost of care did not significantly differ between
2016
groups. To explain the higher unadjusted costs that were observed, Hinami et al14 suggests extra testing and care administered by hospitalists. These costs appeared to be warranted, however, as comanaged patients experienced significantly fewer in-hospital deaths. Despite mixed findings in the literature on the effect of hospitalist comanagement services on hospital costs, we firmly believe that the higher costs observed in our study may be justified. Vascular surgery patients represent a unique cohort that may benefit drastically from the effects of a hospitalist service. Despite an increase in CMI in the comanaged group of patients, Tadros et al4 observed a decrease in in-hospital mortality rates from 1.75% to 0.37% after implementation of the vascular surgeon-hospitalist comanagement service. Decreased average mortality rates in this high risk, medically complex group of comanaged patients demonstrate the effectiveness of a comanagement service in drastically improving in-hospital mortality. Study limitations. First, this study was conducted at an urban, single-institution and academic medical center, so it is possible that the results are not generalizable to the national public. Further, the nature of this study was observational, as patients who were treated in 2012 before comanagement were merely compared with those treated in 2013 and 2014, when the comanagement service was in use. Because of this, we are only able to perceive associations between groups, as randomization did not occur. Individuals were not assigned to either comanagement or noncomanagement interventions, so we were not able to control for a variety of factors that could occur, such as increased CMI in the patients treated under comanagement. With statistical adjustments, however, we were able to control for some of the sampling bias that occurred as a result of the nonrandomized nature of this study. In addition, our cost data lacked information regarding where the changes in cost resided (ie, more laboratory tests, X rays, medications). We also had limited access to data on reimbursement to the hospital, rendering us unable to assess whether or not the extra costs incurred during the comanagement period were offset by higher reimbursements and total revenue. Lastly, no patientspecific data was used in the conduct of this study, so the scope of the analysis was subsequently limited. Because of this, we were unable to examine any measures that required access to patient charts, specifically, including whether the addition of hospitalist physicians had any profound impact on the transition of care to patients’ primary care providers or the frequency of readmissions to outside hospitals.
CONCLUSIONS Overall, our findings have demonstrated that although THCs and ADCs increased over the study period, these
Journal of Vascular Surgery Volume
-,
Number
Tadros et al
7
-
findings were undoubtedly confounded by the increase in both CMI and LOS. After adjusting for these factors, the difference in THC between non-comanaged and comanaged groups was not nearly as high. Our previous work demonstrated an improved in-hospital mortality rate among comanaged vascular surgery inpatients.4 The fact that these improved in-hospital mortality rates were observed in the context of a higher CMI adds further credence to the benefits of comanagement.
4.
5.
6. 7.
AUTHOR CONTRIBUTIONS Conception and design: RT, MT, MS, DK Analysis and interpretation: RT, MT, PF, MS, AV Data collection: RT, MT, CP Writing the article: RT, MT Critical revision of the article: RT, MT, PF, CP, DK, AV, MM Final approval of the article: RT, MT, PF, DK, MM Statistical analysis: RT, MT, MS, CP Obtained funding: Not applicable Overall responsibility: RT RT and MT contributed equally to this article.
8.
9.
10.
11.
REFERENCES 1. Centers for Disease Control and Prevention. National Center for Health Statistics [Internet]. Atlanta (GA): Centers for Disease Control and Prevention. National Hospital Discharge Survey, 2010 [updated April 29, 2015]. Available at: www.cdc. gov/nchs/fastats/inpatient-surgery.htm Accessed March 4, 2016. 2. American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) [Internet]. Preventing complications, reducing costs, improving surgical care. Chicago: American College of Surgeons. Available at: https:// www.facs.org/w/media/files/quality%20programs/nsqip/nsqip infobook1012.ashx. Accessed April 1, 2016. 3. FOJP Service Corporation. The expanding role of hospital medicine and the co-management of patients. The
12.
13.
14.
Quarterly Journal for Health Care Practice and Risk Management 2013;21:1-20. Tadros RO, Faries PL, Malik R, Vouyouka AG, Ting W, Dunn A, et al. The effect of a hospitalist comanagement service on vascular surgery inpatients. J Vasc Surg 2015;61:1550-5. Png CY, Faries PL, Qian LY, Lee IT, Chander R, Finlay DJ, et al. Vascular surgery pain outcomes improved by implementing hospitalist comanagement service. J Pain Manage Med 2016;2:1-3. Crimi E, Hill CC. Postoperative ICU management of vascular surgery patients. Anesthesiol Clin 2014;32:735-57. Herkimer AG. Cost characteristics and behavior. In: Davies M, editor. Understanding hospital financial management. 2nd ed. Rockville, MD: Aspen Publishers, Inc; 1978. Auerbach AD, Wachter RM, Cheng HQ, Maselli J, McDermott M, Vittinghoff E, et al. Comangement of surgical patients between neurosurgeons and hospitalists. Arch Intern Med 2010;170:2004-10. Della Rocca GJ, Moylan KC, Christ BD, Volgas DA, Stannard JP, Mehr DR. Comangement of geriatric patients with hip fractures: a retrospective, controlled, cohort study. Geriatr Orthop Surg Rehabil 2013;4:10-5. Pinzur MS, Gurza E, Kristopaitis T, Monson R, Wall MJ, Porter A, et al. Hospitalist-orthopedic comanagement of high-risk patients undergoing lower extremity reconstruction surgery. Orthopedics 2009;32:495. Swart E, Vasudeva E, Makhni EC, Macaulay W, Bozic KJ. Dedicated perioperative hip fracture comanagement programs are cost-effective in high-volume centers: an economic analysis. Clin Orthop Relat Res 2016;474:222-33. Mendez CM, Harrington DW, Christenson P, Spellberg B. Impact of hospital variables on case mix index as a marker of disease severity. Popul Health Manage 2014;17:28-34. Kuo Y, Goodwin JS. Association of hospitalist care with medical utilization after discharge: evidence of cost shift from a cohort study. Ann Intern Med 2011;155:152-9. Hinami K, Feinglass J, Ferranti DE, Williams MV. Potential role of comanagement in “rescue” of surgical patients. Am J Manag Care 2011;17:333-9.
Submitted Jun 8, 2016; accepted Sep 21, 2016.