Key Points
Question
After accounting for selection bias from patient characteristics, what are the real-world cost differences of laparoscopic and open colectomy?
Findings
In our fully adjusted instrumental variable analysis in this population-based study, laparoscopic colectomy was associated with lower Medicare expenditures than open surgery; however, the magnitude of the association was less than what is observed from conventional multiple regression models. Index hospitalization payments were similar between laparoscopic and open surgery; in contrast, payments for readmissions and postacute care were less common after laparoscopic surgery and resulted in lower overall payments.
Meaning
For patients considered candidates for either approach, laparoscopic colectomy is associated with lower Medicare payments, and the mechanism of cost savings comes from reductions in readmission and postacute care payments.
Abstract
Importance
Numerous study findings suggest that the use of laparoscopy is associated with lower health care costs for many operations, including colectomy. The extent to which these differences are due to the laparoscopic approach itself or selection bias from healthier patients undergoing the less invasive procedure is unclear.
Objective
To evaluate the differences in Medicare expenditures for laparoscopic and open colectomy.
Design, Setting, and Participants
A population-based study was conducted of Medicare beneficiaries undergoing laparoscopic or open colectomy between January 1, 2010, and December 31, 2012. The dates of the analysis were November 13 to December 10, 2016. Using instrumental variable methods to account for selection bias, actual Medicare payments after each procedure were evaluated. To identify the mechanisms of potential cost savings, the frequency and amount of physician, readmission, and postacute care payments were evaluated. Several sensitivity analyses were performed restricting the study population by patient demographic or surgeon specialty.
Main Outcomes and Measures
Actual Medicare expenditures up to 1 year after the index operation.
Results
The study population included 428 799 patients (mean [SD] age, 74 [10] years; 57.0% female). When using standard methods, patients undergoing laparoscopic colectomy (vs open) had lower total Medicare expenditures (mean, −$5547; 95% CI, −$5408 to −$5684; P < .01). When using instrumental variable methods, which account for potentially unmeasured patient characteristics, patients undergoing laparoscopic colectomy (vs open) still had lower Medicare expenditures (mean, −$3676; 95% CI, −$2444 to −$4907; P < .01), although the magnitude of the association was reduced. When examining the root causes of the difference in costs between patients who underwent laparoscopic and open colectomy, the key drivers were a reduction in costs from readmissions (mean, −$1102; 95% CI, −$1373 to −$831) and postacute care (mean, −$1446; 95% CI, −$1988 to −$935; P < .01).
Conclusions and Relevance
This population-based study demonstrates the influence of selection bias on cost estimates in comparative effectiveness research. While the use of laparoscopy reduced total episode payments, the source of savings is in the postacute care period, not the index hospitalization.
This population-based study uses instrumental variable analysis to evaluate the differences in Medicare expenditures for laparoscopic and open colectomy.
Introduction
Surgeons in the United States are performing laparoscopic surgery with greater regularity. For colectomy, one of the most common abdominal operations, the use of laparoscopy will increase by another 40% over the next decade, which is driven by many factors. Laparoscopic approaches are safer, with similar outcomes, and may also lead to lower costs due to less time in the hospital and fewer complications.
However, prior studies examining the differences in cost between laparoscopic and open surgery may be misleading. Most existing evidence is flawed due to selection bias from unmeasured patient characteristics (eg, more complicated patients having open surgery and straightforward cases with laparoscopy). Moreover, prior studies are unable to identify the mechanisms by which laparoscopy reduces costs. Emerging techniques, such as instrumental variable analysis, can be used to address selection bias in observational studies but have not been widely applied in health care research. Using natural variation in 1 or more instrumental variables, this method helps minimize confounding from unmeasured patient characteristics.
Using colectomy as a case study, we sought to systematically examine the association between laparoscopic surgery and Medicare expenditures using an instrumental variable approach. We hypothesized that prior studies overestimated the cost-benefit of laparoscopic surgery due to differences in patient selection. If this type of bias is present, the independent association of laparoscopy with Medicare expenditures will be attenuated in the instrumental variable analysis. Because we use the full range of Medicare payments (ie, hospital stay, readmissions, physician services, and postacute care) rather than charges known to be unreliable, we are also able to explore potential drivers of cost reduction associated with laparoscopy.
Methods
Data Source and Study Population
We used national data from the 100% Medicare Provider Analysis and Review files for the period of January 1, 2010, to December 31, 2012. The dates of the analysis were November 13 to December 10, 2016. The Centers for Medicare & Medicaid Services maintains this database using claims submitted by hospitals where Medicare beneficiaries receive care. We collected patient data, including sex, age, race, comorbidities (ie, principal and secondary diagnosis codes), procedural codes, 30-day complications and mortality, and information regarding hospital length of stay. We linked patients’ records to other Medicare files containing claims relevant to the surgical episode of care. These data included durable medical equipment, home health, long-stay hospitalizations, outpatient, and skilled nursing facility claims. We selected patients undergoing colon resection using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes (45.73, 17.33, 17.32, 45.75, 45.76, 17.35, 17.36, 45.74, 17.34, 45.82, 45.83, 45.81, 48.50, 48.51, 48.52, and 48.53). We excluded patients younger than 65 years or older than 99 years. We further excluded patients not continuously enrolled in Medicare Parts A and B at the time of operation and for the year after surgery. This study was deemed exempt from review by the institutional review board at the University of Michigan. Informed consent was not relevant to deidentified data from administrative sources.
Outcomes
Our primary outcome of interest was actual Medicare payments. In contrast to charges, payments reflect actual Medicare spending related to episodes of care. We extracted payment information for all service types for the index hospitalization. We further collected data on payments for each beneficiary up to 1 year after surgery as detailed above. Payments were grouped into categories for subsequent analyses, including total episode, index hospitalization, physician, readmission, and postacute care. Because Medicare recognizes geographic variation in spending, all payments were price standardized using previously described methods. We identified complications with ICD-9-CM codes using predefined categories, including the following: renal failure (code 584), hemorrhage (code 998.1), myocardial infarction (codes 410.00-410.91), pulmonary failure (codes 518.81, 518.4, 518.5, and 518.8), pneumonia (codes 481, 482.0-482.9, 483, 484, 485, and 507.0), surgical site infection (codes 958.3, 998.3, 998.5, 998.59, and 998.51), deep venous thrombosis or pulmonary embolism (codes 415.1, 451.11, 451.19, 451.2, 451.81, and 453.8), and gastrointestinal bleeding (codes 530.82, 531.00-531.21, 531.40, 531.41, 531.60, 531.61, 532.00-532.21, 532.40, 532.41, 532.60, 532.61, 533.00-533.21, 533.40, 533.41, 533.60, 533.61, 534.00-534.21, 534.40, 534.41, 534.60, 534.61, 535.01, 535.11, 535.21, 535.31, 535.41, 535.51, 535.61, and 578.9). These complications represent a subset of ICD-9-CM codes with the highest sensitivity and specificity as previously described. Rates of payments for services and postoperative events reported in the Results section are unadjusted.
Statistical Analysis
Baseline patient characteristics were evaluated in 2 ways. We compared characteristics between patients undergoing laparoscopic and open surgery as well as after stratifying patients by the median regional use of laparoscopy (23.4%). All comparisons were made by calculating standardized differences for continuous and dichotomous variables. These methods have been previously described. An absolute standardized difference (ASD) exceeding 10 is indicative of statistically significant differences between comparison groups.
Our primary objective was to evaluate the independent association of laparoscopy with Medicare payments. For all models, we adjusted for differences in patient illness using hierarchical condition categories. In brief, categories are assembled from data on patient age and medical comorbidities. They were designed and validated for risk adjustment of Medicare payments, and it is the method used by the Centers for Medicare & Medicaid Services. It is considered more accurate than Elixhauser comorbidities, for example, when the outcome of interest is financial data. We further adjusted for year-to-year differences in payment using categorical dummy variables. We also accounted for regional differences in payments not otherwise accounted for by price standardization using categorical dummy variables for each hospital referral region as a fixed effect. Our first analysis sought to evaluate the independent association of laparoscopic colectomy with Medicare payments using multiple linear regression models. Similar analyses have been reported in the literature, and the intent was to replicate existing estimates of treatment association to confirm generalizability of additional findings from our patient population.
We next used an instrumental variable analysis to address selection bias not accounted for in the conventional multivariable analysis. This analysis was necessary because we hypothesized that patients selected for laparoscopic operations may have lower payments based on clinical characteristics, such as fewer comorbidities, smaller tumors, or more favorable anatomy. If true, the financial benefits of laparoscopy would be incorrectly inflated relative to open surgery.
Instrumental variable methods are an econometric technique used to balance measured and unmeasured differences in characteristics between 2 or more comparison groups. The instrumental variables must be correlated with the exposure (laparoscopy) but not associated with the outcome (Medicare payments) except through its correlation with the exposure (eFigure in the Supplement) This latter criterion is referred to as exogeneity. Our instrumental variable was regional use of the laparoscopic approach. We calculated the proportion of colon resections performed laparoscopically for each hospital referral region. To ensure that the instrument was exogenous, we excluded the hospital in which the patient received his or her operation.
Some patients are more likely to receive laparoscopic colectomy because they received care in a region where this approach is more common. Our analysis accounts for this occurrence and was designed to compare Medicare payments between operative approaches for a patient who would be considered a candidate for either approach (ie, the marginal patient). Conceptually, this technique does not provide estimates for patients who are clearly candidates for an open or laparoscopic approach. We evaluated the instrument in several ways. First, we confirmed its correlation with our exposure, the receipt of laparoscopic colectomy (F = 430). An F statistic exceeding 10 is suggestive of a strong instrument. We confirmed that the instrumental variable approach was necessary using the Durbin-Wu-Hausman tests of endogeneity, which were statistically significant for all instrumental variable models. This finding indicates that standard multivariable regression resulted in biased estimates compared with the instrumental variable model. For example, the coefficient for the first-stage residuals in a model with total episode payments as the outcome was −5546.5 (t test, −102.1) (P < .01). Results were similar for all other outcomes. Using multiple linear regression models, we observed a slight, but statistically significant, association between our instrumental variable and total price-adjusted Medicare payments. This observation is interpreted as the association of the instrument working through our primary exposure.
We used a 2-stage, least squares method for our instrumental variable analysis. This method has been described and validated for financial analyses with health care data. Our first-stage model assessed the association between the receipt of laparoscopic colectomy and our instrumental variable. In this model, we also adjusted for hierarchical condition categories, time (categorical dummy variable for each year), and hospital referral region fixed effects with categorical dummy variables for each region. We used the linear prediction from the first-stage model in the second-stage multiple linear regression model to generate estimates of the local average treatment effect for laparoscopic compared with open colectomy. In this setting, the local average treatment effect is the coefficient (β) for the covariate in question. It can be interpreted as the mean US dollar amount change-per-unit change in continuous variables or per the presence of dichotomous variables. For all estimates, we used bootstrapping to generate 95% CIs and the corresponding t statistics. The t statistics were generated from normal-based 95% CIs derived from bootstrapping with 1000 replications; draws were made at the hospital level to deal with clustering at that level. We then used marginal means to generate absolute estimates of Medicare payments based on surgical approach. Several sensitivity analyses were performed in an identical manner restricting the patient population by clinical diagnosis, elective procedure status, or surgeon specialty.
All statistical analyses were performed using a software program (Stata, version 14; StataCorp LP). We used a 2-sided approach at the 5% significance level for all hypothesis testing.
Results
Differences in Patient Characteristics
We first compared characteristics for patients receiving laparoscopic or open colectomy. While patients were similar in age, we observed statistically significant differences in specific comorbidities, operative diagnoses, and the proportion of elective procedures (Table 1). For example, patients undergoing open colectomy were more likely to have congestive heart failure (10.6% vs 6.4%; ASD, 14.9) and less likely to have an elective procedure (48.8% vs 79.8%; ASD, 47.6). The mean unadjusted total episode payments were also statistically significantly higher for open surgery ($25 778) compared with laparoscopic operations ($16 255) (ASD, 33.1).
Table 1. Patient Characteristics by Type of Procedure and Regional Use of Laparoscopy.
| Variable | Type of Procedure | Regional Use of Laparoscopya | ||||
|---|---|---|---|---|---|---|
| Laparoscopic Surgery (n = 133 528) |
Open Surgery (n = 295 271) |
Absolute Standardized Differenceb | <Median (n = 207 976) |
>Median (n = 220 823) |
Absolute Standardized Differenceb | |
| Age, y | ||||||
| Mean (SD) | 74 (10) | 73 (8) | 6.8 | 73 (9) | 74 (9) | 3.6 |
| Median (IQR) | 75 (14) | 73 (12) | 74 (13) | 74 (13) | ||
| Race, No. (%) | ||||||
| White | 115 575 (86.6) | 252 772 (85.6) | 8.6 | 181 815 (87.4) | 186 529 (84.5) | 2.9 |
| Black | 11 706 (8.8) | 29 698 (10.1) | 6.1 | 18 148 (8.7) | 23 256 (10.5) | 4.4 |
| Comorbidities | ||||||
| Mean (SD) | 2 (2) | 3 (2) | 28.1 | 2 (2) | 2 (2) | 3.8 |
| Median (IQR) | 2 (2) | 2 (3) | 2 (2) | 2 (2) | ||
| Hospital length of stay, median (IQR), d | 5 (3) | 8 (7) | 47.6 | 7 (7) | 7 (7) | 2.2 |
| Specific comorbidities, No. (%) | ||||||
| Congestive heart failure | 8591 (6.4) | 31 235 (10.6) | 14.9 | 19 063 (9.2) | 20 763 (9.4) | 1.0 |
| Pulmonary circulatory disease | 2427 (1.8) | 8151 (2.8) | 6.3 | 5182 (2.5) | 5396 (2.4) | 1.0 |
| Type 2 diabetes | 26 497 (19.8) | 55 351 (18.7) | 2.7 | 39 173 (18.8) | 42 675 (19.3) | 1.2 |
| Type 2 diabetes with complications | 2857 (2.1) | 6630 (2.2) | 1.0 | 4646 (2.2) | 4841 (2.2) | 1.0 |
| Liver disease | 1852 (1.4) | 4310 (1.5) | 1.0 | 3013 (1.4) | 3149 (1.4) | 1.0 |
| Renal failure | 8550 (6.4) | 26 522 (9.0) | 9.7 | 17 094 (8.2) | 17 978 (8.1) | 1.2 |
| Metastatic cancer | 13 916 (10.4) | 47 544 (16.1) | 16.8 | 29 276 (14.1) | 32 184 (14.6) | 1.4 |
| Obesity | 11 184 (8.4) | 22 888 (7.8) | 3.3 | 17 488 (8.4) | 16 584 (7.5) | 2.2 |
| Depression | 8835 (6.6) | 19 524 (6.6) | 1.8 | 14 236 (6.8) | 14 123 (6.4) | 1.0 |
| Operative diagnosis, No. (%) | ||||||
| Cancer | 63 278 (47.4) | 119 908 (40.6) | 13.7 | 86 820 (41.7) | 96 366 (43.6) | 3.8 |
| Diverticular disease | 18 456 (13.8) | 76 613 (25.9) | 30.7 | 44 703 (21.5) | 50 371 (22.8) | 3.1 |
| Inflammatory bowel disease | 1821 (1.4) | 11 235 (3.8) | 15.4 | 7104 (3.4) | 5949 (2.7) | 4.1 |
| Elective procedure, No. (%) | 106 565 (79.8) | 144 061 (48.8) | 47.6 | 130 032 (62.5) | 120 590 (54.6) | 8.5 |
| Postoperative events, No. (%) | ||||||
| Complications | 17 915 (13.4) | 90 842 (30.8) | 36.9 | 50 361 (24.2) | 58 396 (26.4) | 4.4 |
| Death within 30 d | 2193 (1.6) | 23 074 (7.8) | 28.3 | 10 926 (5.3) | 14 341 (6.5) | 5.1 |
| Unadjusted total episode payments, US $ | 16 255 | 25 778 | 33.1 | 21 706 | 23 855 | 6.7 |
Abbreviation: IQR, interquartile range.
Regional use of laparoscopy, the instrumental variable, is the proportion of colectomies performed laparoscopically within a hospital referral region in a given year, excluding the hospital in which the specific beneficiary had his or her operation.
See the Statistical Analysis subsection of the Methods section for a detailed explanation. An absolute standardized difference exceeding 10 suggests statistically significant imbalance of a covariate.
However, once patients were stratified using our instrumental variable (regional use of laparoscopy), there were no longer any statistically significant differences in baseline patient characteristics (Table 1). This observation demonstrates the ability of the instrumental variable to balance patient characteristics.
Total Episode and Index Hospitalization Payments
Next, we compared estimates of Medicare expenditures for laparoscopic and open colectomy using conventional multivariable regression (Table 2). In these analyses, the use of laparoscopy was associated with a reduction in total Medicare payments by a mean of −$5547 (95% CI, −$5408 to −$5684; P < .01) compared with the open approach. Results were similar when examining Medicare expenditures for the index hospitalization, physicians, readmissions, and postacute care. For example, laparoscopy reduced payments for the index hospitalization by −$3419 (95% CI, −$3724 to −$3151; P < .01).
Table 2. Comparison of Payment Estimates for Laparoscopic vs Open Colectomy Using Conventional Multivariable and Instrumental Variable Models.
| Variable | US $ | P Value | ||
|---|---|---|---|---|
| Laparoscopic Surgery | Open Surgery | Average (95% CI) Treatment Effect | ||
| Total Episode Payments | ||||
| Multiple linear regression | 17 161 | 26 062 | −5547 (−5408 to −5684) | <.01 |
| Instrumental variable analysis | 20 452 | 24 573 | −3676 (−2444 to −4907) | <.01 |
| Index Hospitalization Payments | ||||
| Multiple linear regression | 13 016 | 18 502 | −3419 (−3724 to −3151) | <.01 |
| Instrumental variable analysis | 15 077 | 17 570 | −335 (−842 to 254) | .38 |
| Physician Payments | ||||
| Multiple linear regression | 2019 | 2798 | −402 (−356 to −461) | <.01 |
| Instrumental variable analysis | 2249 | 2694 | −292 (−180 to 304) | <.01 |
| Readmission Payments | ||||
| Multiple linear regression | 9284 | 10 429 | −840 (−623 to −1016) | <.01 |
| Instrumental variable analysis | 9791 | 10 292 | −5642 (−3245 to −7986) | <.01 |
| Postacute Care Payments | ||||
| Multiple linear regression | 1204 | 3029 | −1109 (−824 to −1457) | <.01 |
| Instrumental variable analysis | 1850 | 2737 | −1446 (−1988 to −935) | <.01 |
However, when using an instrumental variable analysis, the independent association of laparoscopy with Medicare expenditures was attenuated (Table 3). The mean difference in Medicare expenditures for the total episode of care related to the use of laparoscopy was −$3676 (95% CI, −$2444 to −$4907; P < .01). However, for index hospitalization payments, there was no statistically significant difference between laparoscopic and open approaches (−$335; 95% CI, −$1074 to $405; P = .38).
Table 3. Comparison of the Mean Medicare Payments for Laparoscopic vs Open Colectomy Across Different Patient Demographicsa.
| Variableb | US $ | P Value | ||
|---|---|---|---|---|
| Laparoscopic Surgery | Open Surgery | Average (95% CI) Treatment Effect | ||
| All Patients (N = 428 799) (100%)c | ||||
| Total episode payments | 20 452 | 24 573 | −3676 (−2444 to −4907) | <.01 |
| Index hospitalization payments | 15 077 | 17 570 | −335 (−1074 to 405) | .38 |
| Cancer Operations (n = 183 186) (42.7%) | ||||
| Total episode payments | 20 671 | 22 958 | −3465 (−1879 to −5102) | <.01 |
| Index hospitalization payments | 14 941 | 16 218 | −541 (−1342 to 321) | .12 |
| Benign Operations (n = 245 613) (57.3%) | ||||
| Total episode payments | 20 079 | 25 746 | −3828 (−2901 to −4743) | <.01 |
| Index hospitalization payments | 15 052 | 18 551 | −334 (−998 to 656) | .49 |
| Elective Operations (n = 250 626) (58.4%) | ||||
| Total episode payments | 18 209 | 27 244 | −7661 (−8775 to −6545) | <.01 |
| Index hospitalization payments | 13 693 | 13 948 | −3150 (−2438 to −3887) | <.01 |
| Urgent or Emergent Operations (n = 178 173) (41.6%) | ||||
| Total episode payments | 29 122 | 30 207 | 2042 (−48 to 4132) | .06 |
| Index hospitalization payments | 20 383 | 21 072 | 3709 (−2833 to −4655) | <.01 |
| Operations by Colorectal Surgeons (n = 44 240) (10.3%) | ||||
| Total episode payments | 20 813 | 23 612 | −4184 (−7178 to −902) | .03 |
| Index hospitalization payments | 14 148 | 15 610 | −960 (−2300 to 1259) | .36 |
| Operations by All Other Surgeons (n = 384 559) (89.7%) | ||||
| Total episode payments | 20 416 | 24 610 | −4388 (−3216 to −2421) | <.01 |
| Index hospitalization payments | 15 229 | 17 743 | 167167 (−327 to 616) | .66 |
Estimates are derived from instrumental variable models.
Percentages indicate the percentage of the total patient population represented by each demographic.
The study population had a mean (SD) age of 74 (10) years, and 57.0% were female.
To assess whether our results varied in specific patient demographics, we performed several sensitivity analyses. The mean association of laparoscopy with total episode payments was similar for patients undergoing surgery for cancer and benign diagnoses (Table 3). Laparoscopy reduced total episode costs by a larger amount, on average, for elective operations (−$7661; 95% CI, −$8775 to −$6545; P < .01). In contrast, there was no statistically significant association of laparoscopy with total episode payments for urgent and emergent procedures ($2042; 95% CI, −$48 to $4132; P = .06).
Mechanism of Cost Savings From Laparoscopy
Patients undergoing laparoscopic colectomy (vs open) were less likely to have a readmission to the hospital (10.7% vs 17.4%, P < .01), which translated into lower Medicare expenditures. These readmissions resulted in large Medicare expenditures for patients who underwent laparoscopic ($9791) or open ($10 292) surgery (Table 4). Nonetheless, because readmissions were more common after open surgery, patients who underwent laparoscopy had lower readmission payments per case (−$1102; 95% CI, −$1373 to −$831; P < .01).
Table 4. Main Associations and Sensitivity Analyses to Determine the Mechanism of Cost Savings From the Use of Laparoscopic vs Open Colectomya.
| Variable | Laparoscopic Surgery | Open Surgery | Average (95% CI) Treatment Effect, US $ | P Value | ||
|---|---|---|---|---|---|---|
| No./Total No. (%) of Patients With Payment | Payment, US $ | No./Total No. (%) of Patients With Payment | Payment, US $ | |||
| All Patients | ||||||
| Physician payments when present | 90 896/133 528 (68.1) | 3385 | 199 413/295 270 (67.5) | 3943 | 288 (151 to 438) | .02 |
| Mean overall | NA | 2249 | NA | 2694 | −292 (−218 to −359) | <.01 |
| Readmission payments when present | 13 211/133 528 (9.9) | 9791 | 48 942/295 270 (16.6) | 10 292 | −5642 (−4571 to −6674) | <.01 |
| Mean overall | NA | 1275 | NA | 1571 | −1102 (−1373 to −831) | <.01 |
| Postacute care payments when present | 48 569/133 528 (36.4) | 4584 | 143 074/295 270 (48.5) | 5972 | 262 (−398 to 813) | .49 |
| Mean overall | NA | 1850 | NA | 2737 | −1446 (−1988 to −935) | <.01 |
| Cancer Operations | ||||||
| Physician payments when present | 44 519/63 278 (70.4) | 3573 | 86 523/119 908 (72.2) | 3885 | 231 (−17 to 488) | .07 |
| Mean overall | NA | 2484 | NA | 2842 | −828 (−674 to −1004) | <.01 |
| Readmission payments when present | 6504/63 278 (10.3) | 9605 | 17 950/119 908 (15.0) | 9779 | −3166 (−2775 to −3567) | <.01 |
| Mean overall | NA | 1216 | NA | 1347 | −777 (−540 to −1065) | <.01 |
| Postacute care payments when present | 27 572/63 278 (43.6) | 4407 | 62 583/119 908 (52.2) | 5002 | 501 (−341 to 1329) | .17 |
| Mean overall | NA | 2029 | NA | 2567 | −1318 (−1009 to −1649) | <.01 |
| Benign Operations | ||||||
| Physician payments when present | 46 377/70 250 (66.0) | 3213 | 112 890/175 363 (64.4) | 3998 | 422 (344 to 514) | <.01 |
| Mean overall | NA | 2058 | NA | 2598 | −636 (−444 to −819) | <.01 |
| Readmission payments when present | 6707/70 250 (9.5) | 9888 | 30 992/175 363 (17.7) | 10 608 | −7310 (−5632 to −9165) | <.01 |
| Mean overall | NA | 1301 | NA | 1735 | −1405 (−1143 to −1754) | <.01 |
| Postacute care payments when present | 20 997/70 250 (29.9) | 4682 | 80 491/175 363 (45.9) | 6441 | 41 (−38 to 110) | .93 |
| Mean overall | NA | 1669 | NA | 2860 | −1452 (−833 to −2018) | <.01 |
| Elective Operations | ||||||
| Physician payments when present | 71 976/106 565 (67.5) | 2987 | 96 444/144 068 (66.9) | 2996 | −605 (−433 to −789) | <.01 |
| Mean overall | NA | 1979 | NA | 2039 | 1202 (845 to 1597) | <.01 |
| Readmission payments when present | 9470/106 565 (8.9) | 9579 | 20 692/144 068 (14.4) | 9823 | −4532 (−3892 to −5176) | <.01 |
| Mean overall | NA | 1137 | NA | 1201 | −1145 (−943 to −1376) | <.01 |
| Postacute care payments when present | 35 960/106 565 (33.7) | 3634 | 63 500/144 068 (44.1) | 3785 | −1378 (−1167 to −1545) | <.01 |
| Mean overall | NA | 1399 | NA | 1547 | −2162 (−1764 to −2529) | <.01 |
| Urgent or Emergent Operations | ||||||
| Physician payments when present | 18 920/26 963 (70.2) | 3623 | 102 969/151 210 (68.1) | 3790 | 96 (−21 to 110) | .65 |
| Mean overall | NA | 3287 | NA | 3329 | −215 (−631 to 278) | .30 |
| Readmission payments when present | 3741/26 963 (13.9) | 2632 | 28 250/151 210 (18.7) | 2555 | −856 (−2310 to 1432) | .55 |
| Mean overall | NA | 1815 | NA | 1927 | −1017 (−893 to −1157) | <.01 |
| Postacute care payments when present | 12 609/26 963 (46.8) | 4485 | 79 574/151 210 (52.6) | 4583 | −15 (−540 to 550) | .97 |
| Mean overall | NA | 3636 | NA | 3877 | −381 (−883 to 598) | .29 |
| Operations Performed by Colorectal Surgeons | ||||||
| Physician payments when present | 21 400/21 461 (99.7) | 3 315 | 22 738/22 779 (99.8) | 3786 | −445 (−1230 to 476) | .18 |
| Mean overall | NA | 3306 | NA | 3778 | −452 (−1504 to 713) | .16 |
| Readmission payments when present | 2079/21 461 (9.7) | 9313 | 3787/22 779 (16.6) | 9495 | −445 (−2183 to 1239) | .87 |
| Mean overall | NA | 1170 | NA | 1327 | −550 (−1381 to 287) | .28 |
| Postacute care payments when present | 11 098/21 461 (51.7) | 3785 | 15 540/22 779 (68.2) | 4541 | 1357 (−1006 to 3120) | .17 |
| Mean overall | NA | 2188 | NA | 2896 | −2220 (−1780 to −2699) | <.01 |
| Operations Performed by All Other Surgeons | ||||||
| Physician payments when present | 69 496/112 067 (62.0) | 3406 | 176 675/272 492 (64.8) | 3963 | 210 (−238 to 455) | .10 |
| Mean overall | NA | 2091 | NA | 2586 | −1690 (−1345 to 1967) | <.01 |
| Readmission payments when present | 11 132/112 067 (9.9) | 9875 | 45 159/272 492 (16.6) | 10 360 | −5550 (−6275 to −4897) | <.01 |
| Mean overall | NA | 1292 | NA | 1592 | −1080 (−793 to 1230) | <.01 |
| Postacute care payments when present | 37 471/112 067 (33.4) | 4766 | 127 534/272 492 (46.8) | 5959 | 1107 (654 to 1685) | <.01 |
| Mean overall | NA | 1802 | NA | 2717 | −1783 (−1508 to 2054) | <.01 |
Abbreviation: NA, not applicable.
Estimates are derived from instrumental variable models comparing episode payments for laparoscopic vs open colectomy. Rates of physician payments, readmission payments, and postacute care payments are unadjusted.
Patients undergoing laparoscopic colectomy (vs open) were also less likely to have payments for postacute care (36.4% vs 48.5%, P < .01). As a result, postacute care costs were lower, on average, for patients undergoing laparoscopic surgery (−$1446; 95% CI, −$1988 to −$935; P < .01). Furthermore, laparoscopy did not lead to marked differences in Medicare payments for physician services compared with open surgery. Trends in Medicare payments for physicians, readmissions, and postacute care were consistent across various patient demographics.
Finally, we sought to determine how complications modify cost savings associated with the use of laparoscopic surgery. Complications were more common after open surgery (31%) compared with laparoscopic surgery (16.3%) (P < .01). Complications after either approach resulted in statistically significantly higher Medicare expenditures (Table 5). However, even for patients without a documented complication, laparoscopy (vs open surgery) was associated with lower overall episode payments (−$3707; 95% CI, −$4664 to −$2751; P < .01). Among patients with at least one complication, laparoscopy was also associated with lower overall episode costs relative to open surgery (−$3041; 95% CI, −$5029 to −$874; P < .01).
Table 5. Influence of Complications on the Mean Episode Payments for Laparoscopic vs Open Colectomy Using the Instrumental Variable Models.
| Variable | Laparoscopic Surgery | Open Surgery | Average (95% CI) Treatment Effect, US $ | P Value | ||
|---|---|---|---|---|---|---|
| No./Total No. (%) of Patients With Payment | Payment, US $ | No./Total No. (%) of Patients With Payment | Payment, US $ | |||
| Patients Without a Documented Complication (n = 320 042) (74.6%) | ||||||
| Total episode payments | NA | 17 129 | NA | 19 748 | −3707 (−4664 to −2751) | <.01 |
| Index hospitalization payments | NA | 12 762 | NA | 14 269 | −979 (−544 to −1493) | .09 |
| Physician payments when present | 78 177/115 613 (67.6) | 2934 | 137 714/204 429 (67.4) | 3271 | 151 (−127 to 346) | .10 |
| Mean overall | NA | 1946 | NA | 2232 | −725 (−231 to 1292) | <.01 |
| Readmission payments when present | 10 158/115 613 (8.8) | 9112 | 30 057/204 429 (14.7) | 9398 | −2913 (−4325 to −1501) | <.01 |
| Mean overall | NA | 1049 | NA | 1244 | −531 (−466 to 604) | <.01 |
| Postacute care payments when present | 39 613/115 613 (34.3) | 3559 | 95 724/204 429 (46.8) | 4433 | −208 (−754 to 389) | .47 |
| Mean overall | NA | 1370 | NA | 2002 | −1471 (−1103 to 1761) | <.01 |
| Patients With Any Complication (n = 108 757) (25.4%) | ||||||
| Total episode payments | NA | 34 603 | NA | 36 848 | −3041 (−5029 to −874) | <.01 |
| Index hospitalization payments | NA | 26 002 | NA | 24 841 | 2100 (1865 to 2459) | <.01 |
| Physician payments when present | 12 719/17 915 (71.0) | 5295 | 61 699/90 842 (67.9) | 5618 | 797 (532 to 1027) | <.01 |
| Mean overall | NA | 3569 | NA | 3589 | −946 (−2305 to 1258) | .22 |
| Readmission payments when present | 3053/17 915 (17.0) | 9390 | 18 885/90 842 (20.8) | 13 821 | −10 647 (−13 371 to −7923) | <.01 |
| Mean overall | NA | 2240 | NA | 2403 | −2818 (−2344 to −3324) | <.01 |
| Postacute care payments when present | 8956/17 915 (50.0) | 7818 | 47 350/90 842 (52.1) | 8786 | 340 (−995 to 1576) | .66 |
| Mean overall | NA | 3951 | NA | 4583 | −1377 (−1143 to −1616) | <.01 |
Abbreviation: NA, not applicable.
Readmissions were less common after laparoscopic surgery (vs open) for patients with (9.1% vs 15.4%, P < .01) and without (17.4% vs 21.3%, P < .01) a complication. For patients with vs without complications, undergoing a laparoscopic procedure statistically significantly reduced Medicare expenditures associated with readmissions, although the magnitude of the benefit was greater for patients with complications (−$10 647; 95% CI, −$13 371 to −$7923; P < .01) vs those without complications (−$2913; 95% CI, −$4325 to −$1501; P < .01).
Likewise, Medicare expenditures for postacute care were less common with the laparoscopic (33.7%) vs open (47.2%) (P < .01) approach in patients without a complication. However, in patients with at least one complication, payment for postacute care services occurred at a similar rate between open (52%) and laparoscopic (50%) procedures (P = .18). When postacute care payments were present, laparoscopic surgery did not statistically significantly reduce Medicare expenditures for patients with or without a complication.
Discussion
Among Medicare beneficiaries undergoing colectomy in this population-based study, the use of laparoscopy was associated with lower overall expenditures compared with open surgery. However, the magnitude of this benefit was less than what is described in prior studies. This finding highlights the importance of fully accounting for differences in patient characteristics (eg, selection bias, with healthier patients undergoing laparoscopic surgery or open surgical procedures being more extensive) when evaluating different surgical procedures. This study also characterized the mechanisms by which laparoscopy reduces Medicare payments. Specifically, the benefits of laparoscopy are largely due to lower expenditures for complications, readmissions, and postacute care.
Complications were an important driver of higher costs. They were more common after open surgery, which resulted in more payments for readmissions and postacute care, which also increased costs. Nonetheless, total episode payments almost doubled when patients had a complication, regardless of operative approach. Whether the patient underwent open or laparoscopic surgery, complications resulted in marked increases in payments for readmissions and postacute care. Laparoscopic surgical procedures were, on average, less expensive. However, the relative cost-benefit of laparoscopy over open surgery was attenuated when complications occurred.
When interpreting results of this study, it is important to consider how it differs from the existing literature. In clinical trials, patients are randomized to a particular treatment. Whether they receive a new intervention or the standard of care, they were enrolled in the trial because they are likely to benefit from the intervention, which creates selection bias. Generalizability is further limited by the inclusion of only experienced surgeons. Experienced surgeons are generally thought to have better results overall, which may artificially inflate the association of a new approach or technology. In contrast, observational studies are inclusive of diverse practice settings (surgeons with varying experience) and reflect the real-world application of new and existing technologies. Prior studies assert that laparoscopic colectomy results in lower episode payments, reductions in health care use, and faster return to work for patients. However, these studies must be interpreted within the context of their limitations. Patients undergoing laparoscopic surgery tend to be healthier and less obese and have more favorable anatomy. It is plausible that they would have fewer complications or consume less resources with any operative approach. Our study is novel because we are able to account for this selection bias using an instrumental variable approach. This work also shows that methods previously used for economic analyses can help answer difficult questions in surgical practice.
Limitations and Implications
There are several limitations to this study. Because we use Medicare data, our results may not be generalizable to all patients. However, colon surgery is more common in elderly populations, and we would not expect the delivery of laparoscopic or open surgery to differ in younger cohorts. Selection bias is a limitation of studies using administrative data. However, our study was designed to explicitly address this issue. It is plausible that our instrumental variable is a surrogate for hospital or surgeon quality. We addressed this limitation in showing that the instrument is not a statistically significant predictor of Medicare payments. The most common limitation of instrumental variable analyses relates to study design. Defining a suitable instrumental variable is difficult. Herein, we confirmed first that the instrument met the criteria outlined in our flow diagram (eFigure in the Supplement). Next, we confirmed mathematically that it conformed to the standards of exogeneity.
This work has broad implications for comparative effectiveness studies using cost as an outcome. Many studies, in particular single-center analyses, use imperfect cost measures to generate conclusions. Charges (ie, what the hospital requests from the insurer as reimbursement) are misleading because they are often higher than actual payments. They are also not generalizable because they reflect individual hospitals’ or health systems’ negotiations with payers. Costs derived from ratios of costs to charges can be problematic as well. Although these figures tend to be more accurate than charges, they do not represent what was actually paid for services. Generalizability is also an issue because expected payments vary among hospitals and across different payers. By using actual Medicare payments, this study provides more accurate and generalizable estimates on which we can compare the financial implications of an evolving surgical technology.
This study also has direct clinical implications for insurers in addition to surgeons and hospitals that perform colon surgery. The cost-benefit of laparoscopy is accounted for in the time after the index hospitalization. This period should be a focus of payer-based incentive programs and quality improvement initiatives that aim to reduce costs for patients undergoing colectomy. Because complications are uniformly expensive after either approach, these efforts must be informed by individual surgeons’ comfort or experience with laparoscopy. After all, certain surgeons actually have higher complication rates with the minimally invasive approach. Nonetheless, most hospitals in the United States that perform colectomy or other major abdominal operations already use minimally invasive techniques. Incentive programs would, as a result, be generalizable to most practice settings.
Conclusions
Using actual Medicare payments, this study confirmed that laparoscopic colectomy is overall less expensive than open surgery. However, it is novel because estimates from the instrumental variable models account for selection bias from measured and unmeasured patient characteristics. As a result, this study demonstrated that cost savings of minimally invasive surgery are derived from differences in postacute care payments. These results highlight the potential for payers and clinicians to incentivize the use of laparoscopic approaches to reduce resource use after colectomy.
eFigure. Conceptual Diagram Illustrating the Relationship Between Instrumental Variables, Treatments and/or Risk Factors, and Outcomes
References
- 1.Rea JD, Cone MM, Diggs BS, Deveney KE, Lu KC, Herzig DO. Utilization of laparoscopic colectomy in the United States before and after the Clinical Outcomes of Surgical Therapy Study Group trial. Ann Surg. 2011;254(2):281-288. [DOI] [PubMed] [Google Scholar]
- 2.Kemp JA, Finlayson SR. Nationwide trends in laparoscopic colectomy from 2000 to 2004. Surg Endosc. 2008;22(5):1181-1187. [DOI] [PubMed] [Google Scholar]
- 3.Tsui C, Klein R, Garabrant M. Minimally invasive surgery: national trends in adoption and future directions for hospital strategy. Surg Endosc. 2013;27(7):2253-2257. [DOI] [PubMed] [Google Scholar]
- 4.Crawshaw BP, Chien HL, Augestad KM, Delaney CP. Effect of laparoscopic surgery on health care utilization and costs in patients who undergo colectomy. JAMA Surg. 2015;150(5):410-415. [DOI] [PubMed] [Google Scholar]
- 5.Delaney CP, Kiran RP, Senagore AJ, Brady K, Fazio VW. Case-matched comparison of clinical and financial outcome after laparoscopic or open colorectal surgery. Ann Surg. 2003;238(1):67-72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Epstein AJ, Groeneveld PW, Harhay MO, Yang F, Polsky D. Impact of minimally invasive surgery on medical spending and employee absenteeism. JAMA Surg. 2013;148(7):641-647. [DOI] [PubMed] [Google Scholar]
- 7.Schwenk W, Haase O, Neudecker J, Müller JM. Short term benefits for laparoscopic colorectal resection. Cochrane Database Syst Rev. 2005;(3):CD003145. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Nelson H, Sargent DJ, Wieand HS, et al. ; Clinical Outcomes of Surgical Therapy Study Group . A comparison of laparoscopically assisted and open colectomy for colon cancer. N Engl J Med. 2004;350(20):2050-2059. [DOI] [PubMed] [Google Scholar]
- 9.Bagshaw PF, Allardyce RA, Frampton CM, et al. ; Australasian Laparoscopic Colon Cancer Study Group . Long-term outcomes of the Australasian randomized clinical trial comparing laparoscopic and conventional open surgical treatments for colon cancer: the Australasian Laparoscopic Colon Cancer Study trial. Ann Surg. 2012;256(6):915-919. [DOI] [PubMed] [Google Scholar]
- 10.Fleshman J, Sargent DJ, Green E, et al. ; Clinical Outcomes of Surgical Therapy Study Group . Laparoscopic colectomy for cancer is not inferior to open surgery based on 5-year data from the COST Study Group trial. Ann Surg. 2007;246(4):655-662. [DOI] [PubMed] [Google Scholar]
- 11.Gervaz P, Inan I, Perneger T, Schiffer E, Morel P. A prospective, randomized, single-blind comparison of laparoscopic versus open sigmoid colectomy for diverticulitis. Ann Surg. 2010;252(1):3-8. [DOI] [PubMed] [Google Scholar]
- 12.Weeks JC, Nelson H, Gelber S, Sargent D, Schroeder G; Clinical Outcomes of Surgical Therapy (COST) Study Group . Short-term quality-of-life outcomes following laparoscopic-assisted colectomy vs open colectomy for colon cancer: a randomized trial. JAMA. 2002;287(3):321-328. [DOI] [PubMed] [Google Scholar]
- 13.Sheetz KH, Norton EC, Birkmeyer JD, Dimick JB. Provider experience and the comparative safety of laparoscopic and open colectomy. Health Serv Res. 2017;52(1):56-73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Gottlieb DJ, Zhou W, Song Y, Andrews KG, Skinner JS, Sutherland JM. Prices don’t drive regional Medicare spending variations. Health Aff (Millwood). 2010;29(3):537-543. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Iezzoni LI, Daley J, Heeren T, et al. Identifying complications of care using administrative data. Med Care. 1994;32(7):700-715. [DOI] [PubMed] [Google Scholar]
- 16.Xian Y, Holloway RG, Chan PS, et al. Association between stroke center hospitalization for acute ischemic stroke and mortality. JAMA. 2011;305(4):373-380. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Wagner TH, Upadhyay A, Cowgill E, et al. Risk adjustment tools for learning health systems: a comparison of DxCG and CMS-HCC V21. Health Serv Res. 2016;51(5):2002-2019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Pope GC, Kautter J, Ellis RP, et al. Risk adjustment of Medicare capitation payments using the CMS-HCC model. Health Care Financ Rev. 2004;25(4):119-141. [PMC free article] [PubMed] [Google Scholar]
- 19.Kautter J, Pope GC, Ingber M, et al. The HHS-HCC risk adjustment model for individual and small group markets under the Affordable Care Act. Medicare Medicaid Res Rev. 2014;4(3):mmrr2014-004-03-a03. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Newhouse JP, McClellan M. Econometrics in outcomes research: the use of instrumental variables. Annu Rev Public Health. 1998;19:17-34. [DOI] [PubMed] [Google Scholar]
- 21.Terza JV, Bradford WD, Dismuke CE. The use of linear instrumental variables methods in health services research and health economics: a cautionary note. Health Serv Res. 2008;43(3):1102-1120. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Bonjer HJ, Deijen CL, Abis GA, et al. ; COLOR II Study Group . A randomized trial of laparoscopic versus open surgery for rectal cancer. N Engl J Med. 2015;372(14):1324-1332. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eFigure. Conceptual Diagram Illustrating the Relationship Between Instrumental Variables, Treatments and/or Risk Factors, and Outcomes