Abstract
OBJECTIVE:
To understand variation in intraoperative and postoperative utilization for common general surgery procedures.
SUMMARY BACKGROUND DATA:
Reducing surgical costs is paramount to the viability of hospitals.
METHODS:
Retrospective analysis of electronic health record data for 7,762 operations from two health systems. Adult patients undergoing laparoscopic cholecystectomy, appendectomy, and inguinal/femoral hernia repair between November 1st, 2013 and November 30th, 2017 were reviewed for 3 utilization measures: intraoperative disposable supply costs, procedure time, and postoperative length of stay (LOS). Crossed hierarchical regression models were fit to understand case-mixed adjusted variation in utilization across surgeons and locations and to rank surgeons.
RESULTS:
The number of surgeons performing each type of operation ranged from 20 to 63. The variation explained by surgeons ranged from 8.9% to 38.2% for supply costs, from 15.1% to 54.6% for procedure time, and from 1.3% to 7.0% for postoperative LOS. The variation explained by location ranged from 12.1% to 26.3% for supply costs, from 0.2% to 2.5% for procedure time, and from 0.0% to 31.8% for postoperative LOS. There was a positive correlation (ρ=0.49, p=0.03) between surgeons’ higher supply costs and longer procedure times for hernia repair, but there was no correlation between other utilization measures for hernia repair and no correlation between any of the utilization measures for laparoscopic appendectomy or cholecystectomy.
CONCLUSIONS:
Surgeons are significant drivers of variation in surgical supply costs and procedure time, but much less so for postoperative LOS. Intraoperative and postoperative utilization profiles can be generated for individual surgeons and may be an important tool for reducing surgical costs.
This retrospective analysis of 7,762 laparoscopic appendectomies, cholecystectomies, and inguinal/femoral hernia repairs at 2 health systems found surgeons are significant drivers of variation in surgical supplies and OR time, but have little effect on postoperative length of stay. Ranking surgeons on intraoperative and postoperative utilization may be an important tool for reducing surgical costs.
INTRODUCTION
In the era of value-based payments, understanding and reducing the cost of surgical care is paramount for hospitals to maintain financial viability1. The three most expensive components of a surgical patient’s stay are room and board, time spent in the operating room (OR), and surgical supplies –together accounting for approximately 75% of the total cost to the hospital2. Understanding the variation in each of these components – including the primary drivers – will be useful to clinicians, administrators, and policy makers attempting to reduce surgical costs.
There is a growing body of literature focused on one component of surgical costs – intraoperative disposable supplies – including numerous efforts to reduce variation in supply use across surgeons. The success of these efforts has been modest3. In part, this may reflect a “jumping of the gun,” as interventions cannot be appropriately designed without first understanding the primary drivers of variation. For example, if surgeons only explain a small percentage of the variation in supply costs, then efforts targeting surgeons will have limited effect. A handful of studies have attempted to understand variation in supply costs4–13 but most have focused on a single operation at a single health system and have used limited methods for understanding variation. The inclusion of only one health system is especially limiting as institutional policies and practices may contribute substantially to cost variation. Further, supply costs are only one part of the puzzle, and should be evaluated in the context of other major drivers such as length of stay (LOS) and procedure time.
In this study we report the results of a collaboration between two large health systems in Southern California to understand variation in the cost of common general surgery operations. We used regression methods to analyze the 3 main components of a surgical patients’ stay: intraoperative supply costs, procedure time, and postoperative LOS – hereafter referred to as utilization measures. We addressed 3 questions: First, after adjusting for patient case mix, what proportion of variation in utilization measures is explained by the location of the operation and the surgeon? Second, can outlier surgeons be identified based on their utilization profiles? And third, what are the associations among utilization measures? (e.g. Are higher supply costs associated with shorter postoperative stays?)
METHODS
Data Source, Sample, & Ethics Review
This retrospective review utilized data available in the electronic health records (EHRs) of two distinct academic health systems. Both systems use the Epic EHR. The bioinformatics infrastructure that allowed patient-level queries has previously been described for one of the health systems14. We identified adult (aged 18 years or older) patients undergoing laparoscopic cholecystectomy, laparoscopic appendectomy, and laparoscopic inguinal/femoral hernia repair between November 1st, 2013 and November 30th, 2017. Using current procedural terminology (CPT) codes and booking slip information, we excluded cases where an additional procedure was coded except those that represented a coding phenomenon (i.e. when diagnostic laparoscopy was always coded in addition to laparoscopic cholecystectomy) or, for laparoscopic cholecystectomy, if a concurrent intraoperative cholangiogram (IOC) was coded. IOC was routinely performed at one institution but not the other and excluding these cases would artificially reduce the variation identified at the location-level. Robotic cases were also excluded as relevant costs (i.e. acquisition, maintenance, and instrument costs) were not sufficiently itemized15. Finally, we limited our analysis to surgeons who performed the procedure at least 5 times during the study time period. Institutional review board approval was obtained at each health system prior to starting the research including a waiver of informed consent for patient chart review (IRB#16–001327, Pro00047831).
Utilization Measures
Following the terminology outlined by the Second Panel on Cost Effectiveness in Health and Medicine, the financial perspective of this analysis was that of the healthcare sector – and specifically that of the hospital16. This analysis focused on the actual cost of a hospital delivering a service, an amount that is distinct from that paid by a third party payor (i.e., patient, insurer).
Outcomes of interest included the hospital cost of intraoperative disposable supplies, procedure time, and postoperative LOS. A detailed review of assigning and analyzing intraoperative disposable supply costs was reported elsewhere17. In brief, this outcome captured the cost of disposable supplies and implants used or wasted in the OR. This does not include reusable instruments or capital costs. Within each health system, a single price was assigned to each item over the study time period to account for price changes that occur when supplies are reordered or contracts are renegotiated. This adjustment prevented a surgeon for being penalized if they happen to operate the day after a price increase. Price information was not routinely provided to surgeons in this study. Procedure time is measured in minutes and represents skin-to-skin time. This is distinct from room time that may be influenced by factors outside of the surgeons’ control, such as post anesthesia care unit holds. Postoperative LOS was measured in hours from when the patient left the OR to when the nurse closed the patient encounter (i.e. when the patient left the room/facility). This is different from when the discharge order was signed. Extreme LOS outliers (>10 days for cholecystectomy and appendectomy, >2 days for hernia repair) were excluded because of their leverage exerted on regression coefficients and also because they were suspected to represent cases that substantially deviate from the norm which may unfairly penalize a surgeon.
Covariates
Covariates were included from 3 levels – location, surgeon, and the patient.
The data set included operations performed at 6 locations within the 2 health systems – 4 locations at UCLA (a main OR and an ambulatory surgery center [ASC] at each of the Ronald Reagan and Santa Monica sites) and 2 (a main OR and an ASC) at Cedars Sinai Medical Center (CSMC). We have previously demonstrated that supply costs vary between facilities and settings even within the same health system17.
Patient-level case mix variables included the following: patient age, body mass index (BMI), the American Association of Anesthesiologists physical status classification score (ASA), sex (male, female), race/ethnicity (Hispanic, Non-Hispanic White, Non-Hispanic Black, Non-Hispanic Other), and Elixhauser comorbidity index. Race/ethnicity was included because of differences across the various locations as well as potential effects on outcomes because of differences in pathology (e.g., gallbladder disease is more common among those of Hispanic ethnicity) and as a potential proxy for socioeconomic status. The Elxihauser comorbidity index was generated using ICD10 codes abstracted from the patient encounter and the STATA command Elixhauser18. For laparoscopic cholecystectomy we further included a variable for indication (biliary colic/elective indications, biliary obstruction [i.e. choledocholithiasis, pancreatitis], and acute cholecystitis) using ICD10 codes.
Analysis
All statistical analyses were performed in STATA v15.1 using two-sided tests and an alpha of 0.05. Multiple mixed-effects regression models were fit for each procedure and each utilization measure. All models included covariates as described above. Non-linear relationships and collinearity were addressed through a combination of quadratic terms and de-meaning (or centering). To address our aims, models varied in their hierarchical structure as described below.
To estimate how much variation of each outcome was explained by the location and surgeon, we fit crossed random-effects models. Crossed refers to combinations existing across levels (locations, surgeons) instead of in hierarchies. An additive crossed-effects model and a crossed-effects model with random interaction were run to assess the need for a random intercept for each combination of surgeon and location. Intraclass correlations were calculated by dividing the variance of the random intercept at a given level (location, surgeon) by the sum of all variances across levels including the residual error. To provide context to the magnitude of the intraclass correlations we also estimated the amount of variation explained by patient-level covariates. We compared a a null model that included random effects for the surgeon and location only to a model that included random effects for the surgeon and location along with all patient-level covariates.
After calculating the proportion of variance at each level, we then focused on understanding variation at the surgeon level. We repeated our full model using location fixed effects instead of random effects. This simplified the model while also controlling for unmeasured facility-level factors (i.e. institutional policies, purchasing contracts, nurse staffing) that may influence each outcome. Following estimation, best linear unbiased predictions of the surgeon random effects and standard errors were calculated. Estimates and 95% confidence intervals that were entirely above or below 0 represented “outlier” surgeons with risk-adjusted averages higher or lower than the overall average for that outcome (e.g., LOS).
Most variables had complete or near-complete (<2% missing) data; the only exception was BMI for appendectomy, which was missing in 5.5% of cases. Given the low rates of missingness, all analyses were therefore done on a complete case basis.
RESULTS
Sample Characteristics
The sample included 7,762 operations, including 4,089 laparoscopic cholecystectomies, 2,489 laparoscopic appendectomies, and 1,184 laparoscopic hernia repairs (Table 1). The number of surgeons performing each operation ranged from 20 for laparoscopic hernia repair to 63 for laparoscopic cholecystectomy.
Table 1:
Descriptive data
| Appendectomy | Cholecystectomy | Hernia Repair | |||||
|---|---|---|---|---|---|---|---|
| Sample Size | Patients | 2489 | 4089 | 1184 | |||
| Surgeons | 50 | 63 | 20 | ||||
| Patient Age (Years) | 36 | (27–50) | 52 | (37–67) | 57 | (45–66) | |
| BMI | 24.7 | (22.1–28.2) | 27.6 | (24.1–32.1) | 24.6 | (22.8–27.0) | |
| ASA Score | 1 | 1023 | 41.2% | 562 | 13.8% | 389 | 32.9% |
| 2 | 1214 | 48.8% | 2255 | 55.2% | 641 | 54.1% | |
| 3 | 240 | 9.7% | 1165 | 28.5% | 154 | 13.0% | |
| 4+ | 9 | 0.4% | 100 | 2.4% | 0 | 0.0% | |
| Female | 1199 | 48.2% | 2589 | 63.3% | 119 | 10.1% | |
| Race | Non-Hispanic White | 1520 | 61.4% | 2067 | 50.9% | 909 | 77.7% |
| Non-Hispanic Black | 126 | 5.1% | 353 | 8.7% | 49 | 4.2% | |
| Non-Hispanic Other | 425 | 17.2% | 577 | 14.2% | 120 | 10.3% | |
| Hispanic | 403 | 16.3% | 1067 | 26.3% | 92 | 7.9% | |
| Elixhauser Score | 0 | 1675 | 67.3% | 1722 | 42.4% | 782 | 66.0% |
| 1 | 535 | 21.5% | 1052 | 25.9% | 296 | 25.0% | |
| 2 | 167 | 6.7% | 638 | 15.7% | 80 | 6.8% | |
| 3+ | 112 | 4.5% | 677 | 16.7% | 26 | 2.2% | |
| Procedure Minutes | 54 | (42–72) | 83 | (61–110) | 75 | (61–95) | |
| Postoperative LOS (Hours) | 20.4 | (14.0–33.2) | 23.9 | (8.5–46.1) | 4.3 | (2.5–12.6) | |
ASA = American Society of Anesthesiologists Physical Status; BMI = Body Mass Index; LOS = Length of Stay
All continuous data are presented as median (IQR)
<1% of data was missing for any variable except for 5.5% of operations were missing BMI for appendectomy; the few missing data points explain why variable samples do not always add to the total sample
Patient demographics and case mix factors varied across procedures as expected. For example, patients undergoing appendectomy were typically younger than those undergoing cholecystectomy and most hernia repairs were performed in males while cholecystectomies were predominantly performed in females. For cholecystectomy, 50.5% (2,043/4,042, 47 missing) of operations were for elective indications, 39.2% (1,583/4,042) were for acute cholecystitis, and 10.3% (416/4,042) were for biliary obstruction (i.e. pancreatitis, choledocholithiasis).
Variation in utilization measures explained by surgeon and location
The variation of each utilization measure explained by the surgeon and location are illustrated in Figure 1. Full regression specifications are provided in Appendix 1. The variation explained by surgeon ranged across procedure types from 8.9% to 38.2% for supply cost, from 15.1% to 54.6% for procedure time, and from 1.3% to 7.0% for postoperative LOS. The variation explained by location ranged from 12.1% to 26.3% for supply cost, from 0.2% to 2.5% for procedure time, and from 0.0% to 31.8% for postoperative LOS.
Figure 1: Variation in resource utilization explained by surgeon and location.
LOS = Length of Stay
Notes: Variation explained at each level (surgeon, location) was calculated as the variance of the random intercept for that level (surgeon, location) divided by the sum of variance for the surgeon, location, and residual, after patient case-mix adjustment.
For context, the amount of variation explained by observable patient-level covariates was generally less than 10% (Appendix 2). The only exception was for LOS for cholecystectomy, where patient-level covariates explained 28.8% of the residual variation.
Surgeon profiles & association between utilization measures
Surgeon profiles for each procedure and each utilization measure are included in Figure 2 (appendectomy) and Appendices 3 and 4 (cholecystectomy and hernia repair). Figure 2a ranks surgeons based on their risk-adjusted average supply cost, with green bars representing low outliers (i.e. surgeons with average risk-adjusted supply costs significantly lower than the average) and red bars representing high outliers. The rank assigned to a surgeon in Figure 2a is maintained for Figures 2b and 2c in order to illustrate the relationship between utilization measures. For example, the far left surgeon in Figure 2a is a low outlier for supply costs, is neither a high or low outlier for procedure time, and is a high outlier for postoperative LOS. Conversely, the far right surgeon in Figure 2a is a high outlier for supply costs, and is neither a high or low outlier for procedure time or postoperative LOS.
Figure 2: Surgeon utilization profiles for laparoscopic appendectomy.
Notes: Green bars represent surgeons with point estimates and 95% confidence intervals that are below average (i.e. better) for that utilization measure, blue represent utilization metrics that are neither above or below average because the 95% confidence interval crosses 0, and red bars represent metrics that are above average (i.e. worse). Surgeons were rank ordered in panel 2a based on the magnitude of their deviation (i.e. random intercept); these ranks were maintained in panels 2b and 2c in order to help demonstrate the association between outliers in one domain versus the others. For example, the first surgeon (far left) had mean risk-adjusted supply costs much lower than average (low outlier), procedure time that was not higher or lower than average, but postoperative length of stay that was a high outlier.
The association between utilization measures for laparoscopic appendectomy is included in Figure 3. Each point represents a surgeon, with the x and y axis representing their average risk-adjusted deviation for two of the utilization measures. Among 9 different comparisons (bivariate combinations of supply costs, procedure time, and postoperative LOS, for 3 different operations) only one was significant - a positive correlation (ρ=0.49, p=0.03) between higher supply costs and longer procedure times for hernia repair (Appendix 5). The remaining comparisons were not significant.
Figure 3:
Scatter plots comparing surgeon-level deviations in supply costs, procedure time, and postoperative length of stay for laparoscopic appendectomy
DISCUSSION
In this multi-health system evaluation of cost drivers for 3 common general surgery operations, surgeons accounted for a significant portion of variation in surgical supply costs and procedure time, with a much smaller influence on postoperative LOS. Surgeons with outlier utilization for one component (e.g, supply costs) were generally not outliers on other components, suggesting that increased utilization in one cost component does not “make up” for other components.
This analysis provides a number of insights for understanding ways to control surgical costs. First, location matters when evaluating supply costs. Not just the health system or the facility within a health system, but also the setting (i.e. main OR vs. ASC) in which a procedure is performed is associated with costs. Differences in supply costs between facilities and settings have been shown even when the same distributors and purchasing department are involved17. For example, one location may use bundled packs while the other locations require supplies to be assembled individually, preference cards can vary – even for the same surgeon – across settings, and certain items may be stocked (or simply more conveniently located) in one setting but not the other. Analyses that evaluate supply costs must stratify by facility and setting to create a complete picture.
Second, the surgeon “effect” on procedure time is as strong as, if not stronger than, the effect of surgeons on supply costs. The total cost of running an OR as of 2014 was $37/minute19. The marginal cost – the value of adding or removing a minute of OR time – is likely smaller, with activity-based costing estimates around $10/minute20. As a result, surgeons with procedure times within a couple minutes of average will likely have a negligible effect on costs, but some of the outliers in this study were more substantial. At the extremes, for appendectomy, one surgeon had risk-adjusted procedure times 23 minutes shorter than average (n=108 operations), and another had procedure times 21 minutes longer than average (n=49 operations). Bringing this second surgeon to the average would have cumulatively saved 1,029 minutes in OR time. This difference may have a real impact on hospital finances – not just the added labor costs, but also potentially loss of revenue from operations that could have been performed (i.e. opportunity costs). There are certainly justifiable reasons for high outliers – such as the surgeon’s learning curve or resident teaching – but these profiles can quickly identify surgeons that may require further investigation. Conversely, what strategies or techniques does the low outlier surgeon employ in order to achieve such short operative times?
Third, targeting surgeons to reduce LOS is likely not a high-yield strategy for cost containment in these three common, but relatively simple, procedures. There was certainly variation in LOS but this variation was randomly distributed across surgeons after controlling for patient case mix. Fundamentally, this suggests a different driving force. Whereas supplies and procedure time are inherently under the surgeon’s control, postoperative stay may be more influenced by patient and disease characteristics than by a particular surgeon’s management strategy. Further evidence for this is the relatively larger fraction of variance explained by patient factors for postoperative LOS compared to the other outcomes.
Fourth, providing surgeons with a utilization profile across all 3 measures simultaneously may be a much more powerful tool to motivate change behavior than looking at one in isolation. A number of commercial software systems have come to market profiling surgeons on supply costs alone. Indeed, when we started this collaboration, surgical supplies were our focus, but surgeons quickly noted that supply costs may be inextricably linked to other utilization measures such as procedure time and LOS. This analysis shows that, in general, this is not true – being a high outlier in one domain does not portend a low outlier in another – and also provides a graphical way of illustrating this. Surgeons may be more amenable to change if they see that they are a high outlier of supply utilization and yet do not “compensate” for this utilization in other areas.
Fifth, and finally, using procedure time and LOS – rather than estimating a cost for these utilization measures - may be more accurate and approachable for clinicians. Commercial software systems can generate a total “cost” for a patient’s stay and can identify surgeons that have overall costs higher than others. However, the validity of these costs is only as good as the underlying accounting methodology. Many of these systems rely on cost-to-charge ratios or other blunt accounting methods that may produce inaccurate estimates for surgical patients21. Procedure time and LOS are measures that are in units the clinician can easily appreciate, are valid, and can be weighed against each other based on finances that are specific to that institution and that procedure.
This study has a number of limitations. First, while this is the first study to evaluate intraoperative and postoperative utilization at more than one health system, we were still limited to 2 organizations in southern California and 3 different procedures. The variation at each level will undoubtedly vary by institution, specialty, and procedure. Second, the risk-adjustment variables available to us were limited, relying largely on ICD10 codes. If a clinical registry, such as NSQIP, would add disposable supply costs, it is likely these utilization profiles could be generated across hospitals and surgeons using more robust risk-adjustment methods. Third, our analysis was limited to common laparoscopic operations with short postoperative stays that may have limited our ability to detect differences in this utilization measure across surgeons. Lastly, we only had information for the inpatient stay. Readmissions are an important driver of cost in many surgical procedures, but they are also quite rare and detecting differences across surgeons would require large surgeon-specific samples which may not be possible.
CONCLUSION
Surgeons are significant drivers of variation in surgical supplies and OR time, and much less so for postoperative LOS. Intraoperative and postoperative utilization profiles can be generated for individual surgeons and are important tools for those interested in reducing surgical costs.
Funding/Support:
Christopher Childers was funded by AHRQ#F32HS025079. Ron D. Hays received support from the University of California, Los Angeles (UCLA) and Charles R. Drew University (CDU), Resource Centers for Minority Aging Research Center for Health Improvement of Minority Elderly (RCMAR/CHIME) under NIH/NIA Grant P30-AG021684.
Appendices
Appendix 1:
Regression specifications
| Appendectomy | Cholecystectomy | Hernia | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Fixed Effects Parameters | ||||||||||
| Cost ($) | Procedure Time (Minutes) | LOS (Hours) | Cost ($) | Procedure Time (Minutes) | LOS (Hours) | Cost ($) | Procedure Time (Minutes) | LOS (Hours) | ||
| Sample (3) | 2333 | 2332 | 2292 | 3986 | 3985 | 3888 | 1170 | 1169 | 1158 | |
| Age(2) | 0.44 | 0.189*** | 0.201*** | 0.698** | 0.166*** | 0.196*** | 0.219 | 0.095 | 0.014 | |
| Age2(2) | 0.012 | −0.004* | 0.005* | 0.015 | 0 | 0.010*** | −0.191** | 0 | 0.001 | |
| BMI(2) | 6.553*** | 0.465*** | −0.053 | 0.613 | 0.698*** | −0.329*** | 0.565 | −0.008 | 0.014 | |
| BMI2(2) | −0.002*** | −0.000*** | 0 | −0.013 | −0.001 | 0.010* | 0 | 0 | 0 | |
| ASA | 1 | Reference | ||||||||
| 2 | −11.826 | −0.451 | 0.271 | 8.12 | 1.845 | 1.686 | 20.845 | 2.099 | 0.437 | |
| 3 | 31.738 | 3.966 | 8.566** | 40.759** | 3.247 | 8.870*** | 17.443 | 3.101 | 1.372* | |
| 4 | −38.41 | 5.069 | 67.272*** | 99.757*** | 8.44 | 36.205*** | NA | |||
| 5 | NA | 207.932 | −32.287 | −44.808 | NA | |||||
| Elixhauser Score | 10.592 | 0.942 | 5.005*** | 2.352 | 2.010* | 4.660*** | −36.117 | −2.278 | −0.421 | |
| Elixhauser Score2 | −1.388 | 0.257 | 0.421 | −0.33 | −0.22 | 0.754*** | 17.565 | 1.719* | 0.559** | |
| Sex | Female | Reference | ||||||||
| Male | −16.823 | −0.84 | 2.611* | −30.441*** | −10.684*** | −2.583* | 47.623 | −5.227* | 0.834 | |
| Race | Non-Hispanic White | Reference | ||||||||
| Non-Hispanic Black | −20.726 | 3.471 | 3.449 | 17.351 | 2.929 | 4.205* | −91.671 | −0.218 | 0.896 | |
| Non-Hispanic Other | −41.92 | −0.966 | 1.107 | 5.219 | 1.051 | 1.642 | 100.525 | 1.637 | 0.387 | |
| Hispanic | −14.197 | 1.181 | 1.952 | 11.101 | 4.958*** | 5.560*** | 52.321 | 3.045 | 0.178 | |
| Constant | 1326.599*** | 61.652*** | 22.746*** | 601.668*** | 88.337*** | 12.044** | 1506.562*** | 91.017*** | 6.862*** | |
| Random Effects Parameters (Variance Estimates) | ||||||||||
| Location | 69217.6 | 10.1 | 0.0 | 11541.7 | 2.9 | 47.8 | 168938.5 | 33.7 | 13.7 | |
| Surgeon | 26892.2 | 89.3 | 12.7 | 36366.8 | 435.0 | 70.1 | 119060.7 | 732.2 | 2.1 | |
| Residual | 207608.8 | 490.9 | 948.6 | 47179.8 | 1243.4 | 889.8 | 354277.2 | 572.6 | 27.2 | |
ASA = American Society of Anesthesiologists; BMI = Body Mass Index; LOS = Length of Stay
p < 0.05,
p<0.01,
p<0.001
Cholecystectomy models also included a 3-category variable for indication; having an indication of biliary obstruction or acute cholecystitis were independent predictors of higher OR costs, longer procedure time, and longer length of stay compared to undergoing surgery for elective indications
BMI and age were de-meaned prior to including quadratic terms in order to prevent collinearity
Sample for LOS smaller than for cost and procedure time due to exclusion of extreme outliers (>10 days for appendectomy and cholecystectomy and >2 days for hernia repair)
Appendix 2:
Proportional Reduction in Residual Variance Explained by Patient Factors
| Residual Variance | ||||
|---|---|---|---|---|
| Null Model | Full Model | Proportional Reduction in Residual Variance | ||
| Appendectomy | Cost | 216663.80 | 207608.80 | 4.2% |
| Procedure Time | 515.75 | 490.86 | 4.8% | |
| LOS | 1031.21 | 948.56 | 8.0% | |
| Cholecystectomy | Cost | 49524.63 | 47179.83 | 4.7% |
| Procedure Time | 1373.60 | 1243.36 | 9.5% | |
| LOS | 1250.41 | 889.78 | 28.8% | |
| Hernia Repair | Cost | 364885.10 | 354277.20 | 2.9% |
| Procedure Time | 588.62 | 572.58 | 2.7% | |
| LOS | 28.21 | 27.23 | 3.5% | |
Notes: Null model includes random effects for surgeon and location only; full model includes all patient-level covariates.
Appendix 3:
Surgeon Utilization Profiles for Laparoscopic Cholecystectomy
Appendix 4: Surgeon Utilization Profiles for Laparoscopic Groin Hernia Repair.
** No profile was generated for postoperative LOS because surgeons explained 0% of the variation in postoperative LOS for this procedure after including location fixed effects**
Appendix 5:
Correlation between supply costs and procedure time for laparoscopic groin hernia repair
Footnotes
This manuscript will be presented at the American College of Surgeons 105th Annual Clinical Congress, HSR I, Scientific Forum, San Francisco, CA, October 2019.
References
- 1.Nathan H, Dimick JB. Medicare’s Shift to Mandatory Alternative Payment Models: Why Surgeons Should Care. JAMA Surg. 2017;152(2):125–126. [DOI] [PubMed] [Google Scholar]
- 2.Stey AM, Brook RH, Needleman J, et al. Hospital costs by cost center of inpatient hospitalization for medicare patients undergoing major abdominal surgery. J Am Coll Surg. 2015;220(2):207–217 e211. [DOI] [PubMed] [Google Scholar]
- 3.Childers CP, Showen A, Nuckols T, Maggard-Gibbons M. Interventions to Reduce Intraoperative Costs: A Systematic Review. Ann Surg. 2018;268(1):48–57. [DOI] [PubMed] [Google Scholar]
- 4.Adkins HH, Hardacker TJ, Ceppa EP. Examining variation in cost based on surgeon choices for elective laparoscopic cholecystectomy. Surgical endoscopy. 2016;30(7):2679–2684. [DOI] [PubMed] [Google Scholar]
- 5.Benoit RM, Cohen JK. The relationship between quality and costs: factors that affect the hospital costs of radical prostatectomy. Prostate cancer and prostatic diseases. 2001;4(4):213–216. [DOI] [PubMed] [Google Scholar]
- 6.Brauer DG, Hawkins WG, Strasberg SM, et al. Cost variation in a laparoscopic cholecystectomy and the association with outcomes across a single health system: implications for standardization and improved resource utilization. HPB : the official journal of the International Hepato Pancreato Biliary Association. 2015;17(12):1113–1118. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Chung WC, Fan PL, Chiu HC, Yang CY, Huang KL, Tzeng DS. Operating room cost for coronary artery bypass graft procedures: does experience or severity of illness matter? Journal of evaluation in clinical practice. 2010;16(6):1063–1070. [DOI] [PubMed] [Google Scholar]
- 8.Kazberouk A, Sagy I, Novack V, McGuire K. Understanding the Extent and Drivers of Interphysician Cost Variation for Spine Procedures. Spine. 2016;41(13):1111–1117. [DOI] [PubMed] [Google Scholar]
- 9.Meier JD, Duval M, Wilkes J, et al. Surgeon dependent variation in adenotonsillectomy costs in children. Otolaryngology--head and neck surgery : official journal of American Academy of Otolaryngology-Head and Neck Surgery. 2014;150(5):887–892. [DOI] [PubMed] [Google Scholar]
- 10.Rosenbaum BP, Modic MT, Krishnaney AA. Value in Single-level Lumbar Discectomy: Surgical Disposable Item Cost and Relationship to Patient-reported Outcomes. Clin Spine Surg. 2017;30(9):E1227–E1232. [DOI] [PubMed] [Google Scholar]
- 11.Sjogren PP, Gale C, Henrichsen J, et al. Variation in costs among surgeons and hospitals in Pediatric tympanostomy tube placement. The Laryngoscope. 2016;126(8):1935–1939. [DOI] [PubMed] [Google Scholar]
- 12.Thomas A, Alt J, Gale C, et al. Surgeon and hospital cost variability for septoplasty and inferior turbinate reduction. International forum of allergy & rhinology. 2016;6(10):1069–1074. [DOI] [PubMed] [Google Scholar]
- 13.Zygourakis CC, Valencia V, Boscardin C, et al. Predictors of Variation in Neurosurgical Supply Costs and Outcomes Across 4904 Surgeries at a Single Institution. World neurosurgery. 2016;96:177–183. [DOI] [PubMed] [Google Scholar]
- 14.Hofer IS, Gabel E, Pfeffer M, Mahbouba M, Mahajan A. A Systematic Approach to Creation of a Perioperative Data Warehouse. Anesth Analg. 2016;122(6):1880–1884. [DOI] [PubMed] [Google Scholar]
- 15.Childers CP, Maggard-Gibbons M. Estimation of the Acquisition and Operating Costs for Robotic Surgery. JAMA. 2018;320(8):835–836. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Sanders GD, Neumann PJ, Basu A, et al. Recommendations for Conduct, Methodological Practices, and Reporting of Cost-effectiveness Analyses: Second Panel on Cost-Effectiveness in Health and Medicine. JAMA. 2016;316(10):1093–1103. [DOI] [PubMed] [Google Scholar]
- 17.Childers CP, Hofer IS, Cheng DS, Maggard-Gibbons M. Evaluating Surgeons on Intraoperative Disposable Supply Costs: Details Matter. J Gastrointest Surg. 2018. [DOI] [PubMed] [Google Scholar]
- 18.Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005;43(11):1130–1139. [DOI] [PubMed] [Google Scholar]
- 19.Childers CP, Maggard-Gibbons M. Understanding Costs of Care in the Operating Room. JAMA Surg. 2018;153(4):e176233. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Yu YR, Abbas PI, Smith CM, et al. Time-driven activity-based costing to identify opportunities for cost reduction in pediatric appendectomy. J Pediatr Surg. 2016;51(12):1962–1966. [DOI] [PubMed] [Google Scholar]
- 21.Childers CP, Dworsky JQ, Russell MM, Maggard-Gibbons M. Comparison of Cost Center-Specific vs Hospital-wide Cost-to-Charge Ratios for Operating Room Services at Various Hospital Types. JAMA Surg. 2019. [DOI] [PMC free article] [PubMed] [Google Scholar]