Fax: (216) 296-0025
Pricing of surgeries for colon cancer†
Patient severity and market factors
Article first published online: 8 MAY 2012
Copyright © 2012 American Cancer Society
Volume 118, Issue 23, pages 5741–5748, 1 December 2012
How to Cite
Dor, A., Koroukian, S., Xu, F., Stulberg, J., Delaney, C. and Cooper, G. (2012), Pricing of surgeries for colon cancer. Cancer, 118: 5741–5748. doi: 10.1002/cncr.27573
Preliminary results were presented in part at the AcademyHealth Annual Research Meeting, Seattle, Washington, June 11-14, 2011.
- Issue published online: 19 NOV 2012
- Article first published online: 8 MAY 2012
- Manuscript Accepted: 27 FEB 2012
- Manuscript Revised: 30 JAN 2012
- Manuscript Received: 2 DEC 2011
- colon cancer;
- medical prices
This study examined effects of health maintenance organization (HMO) penetration, hospital competition, and patient severity on the uptake of laparoscopic colectomy and its price relative to open surgery for colon cancer.
The MarketScan Database (data from 2002-2007) was used to identify admissions for privately insured colorectal cancer patients undergoing laparoscopic or open partial colectomy (n = 1035 and n = 6389, respectively). Patient and health plan characteristics were retrieved from these data; HMO market penetration rates and an index of hospital market concentration, the Herfindahl-Hirschman index (HHI), were derived from national databases. Logistic and logarithmic regressions were used to examine the odds of having laparoscopic colectomy, effect of covariates on colectomy prices, and the differential price of laparoscopy.
Adoption of laparoscopy was highly sensitive to market forces, with a 10% increase in HMO penetration leading to a 10.9% increase in the likelihood of undergoing laparoscopic colectomy (adjusted odds ratio = 1.109; 95% confidence interval [CI] = 1.062, 1.158) and a 10% increase in HHI resulting in 6.6% lower likelihood (adjusted odds ratio = 0.936; 95% CI = 0.880, 0.996). Price models indicated that the price of laparoscopy was 7.6% lower than that of open surgery (transformed coefficient = 0.927; 95% CI = 0.895, 0.960). A 10% increase in HMO penetration was associated with 1.6% lower price (transformed coefficient = 0.985; 95% CI = 0.977, 0.992), whereas a 10% increase in HHI was associated with 1.6% higher price (transformed coefficient = 1.016; 95% CI = 1.006, 1.027; P < .001 for all comparisons).
Laparoscopy was significantly associated with lower hospital prices. Moreover, laparoscopic surgery may result in cost savings, while market pressures contribute to its adoption. Cancer 2012. © 2012 American Cancer Society.
Colon cancer is currently the fourth most commonly diagnosed malignancy and second leading cause of cancer death in the United States.1 Surgical resection of the malignant portion of the colon, or partial colectomy, remains the standard of care and offers the primary opportunity for long-term disease-free survival. Approximately 134,000 elective colectomies were performed nationally in 2004.2 Colectomy may be performed using open or laparoscopic techniques, with laparoscopic resection gradually emerging as a frequent alternative to open resection.2-4 The potential benefits of the laparoscopic approach include reductions in intraoperative blood loss, faster postoperative recovery time, and improved pain control.5 A meta-analysis of randomized controlled trials has demonstrated that long-term survival after laparoscopic resection is similar to that after open resection.6
Although the clinical benefits of laparoscopic surgery may have contributed to its diffusion, the role of market forces has not been considered. Moreover, despite a growing emphasis on health care consumerism and policy interest in providing patients and purchasers with pricing data needed to make informed choices,7-9 virtually no information on the relative price paid by insurers to hospitals of laparoscopic versus open surgery exists. Previous research has focused on diagnostic procedures, emphasizing the impact of competitive pressures exerted by managed care on use rates10-14 or on cardiac procedures, examining the impact of hospital competition on pricing.15, 16 In addition, price differences may reflect actual costs associated with the delivery of the procedures.
In private markets, hospitals prices are determined by negotiations between hospitals and insurers, and therefore these prices may be susceptible to pressures from both managed care and from competing hospitals.13 To address this in the case of hospital inpatients undergoing partial colectomy, we examined the extent to which the likelihood of having laparoscopy versus open surgery is affected by both sources of competitive pressures, adjusted for patient severity and risk. Similarly, we examined how these competitive pressures affect the price of colectomy and the differential price of laparoscopic surgery. To this end, we analyzed a claims database for privately insured individuals. Unlike Medicare, where prices are administratively determined, prices in the private sector are susceptible to the direct influence of market forces. Moreover, by focusing on private claims, we were able to analyze the actual payment received by the hospital, namely transaction prices,15, 16 rather than reported charges. Although hospital charges are commonly used in the literature, they imprecisely reflect payments made by patients and insurers. The negotiated price paid by insurers may vary across patients, based on a variety of risk-adjustment methods.17
MATERIALS AND METHODS
Data and Sample
The main database used is the MarketScan Commercial Claims and Encounters file; it assembles complete insurance claims for approximately 100 medium-size and large employers previously used in nationally representative population-based studies.18, 19 We extracted claims for hospitalizations for employees and dependents with a diagnosis of colorectal cancer who underwent partial colectomy. In addition to actual payments (transaction prices), variables include type of insurance plan, comorbidities associated with the main diagnosis, descriptors of surgery type and complexity, and demographic characteristics. Race was not reported. Because of strict confidentiality requirements in the data, the identity of employers is not provided. However, for purpose of this study, hospital identities were made available so that hospital characteristics could be linked, provided that the identity of individual hospitals remained confidential. Hospital characteristics were retrieved from the American Hospital Association Annual Survey; the Managed Market Surveyor File from InterStudy provided the health maintenance organization (HMO) penetration in the market area. Pooling all files from 2002 to 2007, we identified a total of 7424 hospital admissions for partial colectomy surgeries, of which 1035 were laparoscopic. This study was approved by the institutional review boards at George Washington University, Washington, District of Columbia, and at Case Western Reserve University, Cleveland, Ohio.
The sample for the study was identified from all hospitalizations with a diagnosis code for colon cancer (International Classification of Diseases, Ninth Revision With Clinical Modification [ICD-9-CM] 153.0-153.4, 153.6-153.9) and a procedure code during the same hospitalization for partial colectomy (ICD-9-CM 45.71, 45.72, 45.73, 45.74, 45.75, 45.76, 45.79, 54.21; Current Procedural Terminology, Fourth Edition [CPT-4] 44140, 44141, 44143, 44144, 44145, 44146, 44147, 44160, 44204, 44205, 44206, 44207, 44208, 44210, 44212). Surgical cases were divided into laparoscopic and open resection according to procedure codes (laparoscopic: ICD-9 54.21, CPT-4 44204-8, 44210-2; all others open). To adjust for patients' burden of illness and preexisting conditions, we used the list of comorbid conditions provided by Elixhauser et al.20 We opted to include the count of these conditions (0, 1, 2, 3+) rather than indicators for the most prevalent conditions. An exception was the presence of metastatic cancer, which entered our models as a separate variable (ICD-9-CM 197.0-197.4, 197.6-199.0).21 In addition, to identify colorectal resections that were performed on an emergency basis, we included diagnosis codes indicating intestinal obstruction (ICD-9-CM 560.81, 560.89, 560.9), peritonitis (ICD-9-CM 567.0, 567.2, 567.8, 567.9) and perforation (ICD-9-CM 569.83).22
We also included binary indicators for type of insurance offered by the employer. Preferred provider organizations (PPOs) were the dominant form of insurance in our data, accounting for approximately 75% of all colectomy-related hospital stays in our data. The next group was point-of-service (POS) plans. Traditional, closed-panel HMOs formed the last and relatively least common category. Hospital characteristics included type of ownership (public nonfederal, for-profit, and nonhospital), teaching status, and a binary indicator of multihospital system affiliation. HMO penetration is the percent of insured population covered by HMOs in the area; Consistent with previous studies,23-25 we used the rates defined at the county level rather than the metropolitan area to allow for broader geographical representation. The Herfindahl-Hirschman Index (HHI) is the measure of hospital concentration often used in antitrust litigation as a measure of anticompetitiveness in the market.13, 15, 16, 26 It is defined by the share of hospital admissions squared, summed over all hospitals in the market area, with a higher value indicating less competition. Index values of 1 indicate a local monopoly, whereas values approaching 0 indicate perfect competition. For purposes of this research, we constructed this variable for hospital referral regions, which represent regional health care markets for tertiary medical care.27, 28 The rapid diffusion rate of laparoscopic surgery relative to all colectomy shown in Figure 1 closely matches results from other national databases.2, 3
Outcome variables included use of laparoscopic surgeries compared with open surgery, and price (payments made by insurers to hospitals) for these surgeries. Retrospective data analysis for laparoscopy included multivariate logistic regressions on the odds of having laparoscopy and generalized linear models for the natural log of prices. Log transformations are the standard way of specifying continuous health care expenditure and cost outcomes, as a way of smoothing their distributions in statistical estimation.29 To obtain robust standard errors in the log price models, we estimated the generalized linear model with a log link function and gamma distribution (GENMOD procedure, SAS version 9.2) using maximum-likelihood.29 To obtain relative price ratios, regression coefficients were back-transformed to the linear scale. For categorical variables, a transformed coefficient of 0.90 implies a 10% lower price relative to the reference category. For continuous covariates, such as age, transformed coefficients provide the percent effect on price of a unit change in the covariate.
Earlier work on the association between HMO penetration and colorectal cancer screening examined nonlinearities by stratifying penetration into low, moderate, and high categories.25 It was found that the likelihood of receiving colonoscopy following a fecal occult blood test or flexible sigmoidoscopy was significantly higher for Medicare patients residing in areas with high HMO penetration compared with those residing in areas with moderate and low penetration. To examine if similar nonlinearities apply to our data, we performed an additional sensitivity analysis categorizing HMO penetration into these categories. Similarly, we conducted a sensitivity analysis for the association between low, medium, and high categories of the hospital concentration index. For both variables, we set low, medium, and high to be equally distributed in the data. This yielded the categories of <9%, 9% to 25%, and >25% for HMO penetration, and <0.08, 0.08 to 0.15, and >0.15 for HHI.
Because patients undergoing laparoscopic procedures may be inherently different from those undergoing open colectomy, we repeated our analysis by using the propensity score matching approach and the nearest-neighbor method30-32 to obtain a 1:3 matching for laparoscopic versus open colectomy.33 This method has been shown to alleviate the bias due to systematic differences between the 2 treatment groups.34 In addition, we performed parallel analyses that excluded emergent and metastatic cases, which occurred disproportionately in the open colectomy group. Because our findings did not change in any meaningful way with these adjustments, we present the results from the full sample analysis.
Our study population included 7424 patients who underwent partial colectomy, of whom 1035 (13.96%) underwent laparoscopic colectomy (Table 1). The average price for all colectomy-related hospital stays in our data was about $20,819, well in line with estimates of colectomy-based hospital-administered prices (rates for diagnosis-related groups) recently reported for Medicare.35, 36 The average unadjusted price (and standard deviation) for colectomy was moderately higher for open than for laparoscopic procedures, at $21,257 ($12,605) and $18,113 ($11,830), but differences were statistically significant (P < .001). Mean age and the proportion of men were similar between the open and laparoscopic colectomy groups (54.8 vs 55.3 years of age, P = .098; and 51.2% vs 51.7%, P = .7683, respectively; Table 1). With regard to clinical presentation, the distribution of comorbidities did not vary differ significantly between the open and laparoscopic colectomy groups; however, patients with metastatic cancer and patients undergoing emergency resection were overrepresented in the open surgery group relative to the laparoscopy group (16.7% vs 6.5%, and 12.4% vs 5.6%, respectively, P < .001 for all comparisons).
|Hospital price ($)||18,113||(11,830)||21,257||(12,605)||20,819||(12,546)|
|Patient level severity|
|Total||N = 1035||N = 6389||N = 7424|
The proportion of patients who were treated in teaching and in nonprofit hospitals was higher in the laparoscopic than the open colectomy group (23.4% vs 17.5%, P < .001; and 75.3% vs 71.8%, P = .04, respectively). The mean HMO penetration rate was higher in the laparoscopic group than in the open colectomy group (23.11% vs 19.36%) whereas the HHI index was moderately lower (0.13 and 0.15, respectively; P < .001 for both comparisons).
Table 2 presents odds of undergoing laparoscopic colectomy rather than open colectomy. For every 10 years of increase in age, the likelihood of laparoscopic colectomy increased by nearly 11.1% (adjusted odds ratio [AOR] = 1.118, 95% confidence interval [CI] = 1.015, 1.231). On the other hand, presenting with 3 or more comorbid conditions, metastatic cancer, and emergency colectomy were each associated with a significantly lower likelihood of laparoscopy. After adjusting for patient demographics and clinical presentation, those being treated in a teaching hospital were 40.8% more likely than others to undergo laparoscopic colectomy (AOR = 1.408, 95% CI = 1.188, 1.669). Similarly, an increase in the HMO penetration rate by a 10% increment was associated with nearly 10.9% increased likelihood of laparoscopy (AOR = 1.109, 95% CI = 1.062, 1.158). A 0.10 increase in HHI was associated with lower likelihood of undergoing laparoscopic colectomy, meaning that less competition was associated with less laparoscopy (AOR = 0.936, 95% CI = 0.880, 0.996). With regard to trends in the study period from 2002 to 2007, patients were increasingly and significantly more likely to undergo laparoscopic rather than open colectomy in later years. By 2007, patients were approximately 12.5 times more likely to undergo laparoscopy compared with the base year 2002 (AOR = 12.510, 95% CI = 7.993, 19.578). This matches findings from Australia4 over the same period, and findings from 2 US studies spanning 2000 to 20043 and 2007.2
|Variable||Parameter Estimate||Standard Error||Adjusted Odds Ratio||95% Confidence Interval|
|Age (10 year)||0.111a||0.049||1.118||(1.015, 1.231)|
|Patient level severity|
|Elixhauser 1||−0.170a||0.082||0.843||(0.719, 0.990)|
|Elixhauser 2||−0.172||0.115||0.842||(0.672, 1.054)|
|Elixhauser 3+||−0.541b||0.166||0.582||(0.421, 0.805)|
|Plan: HMO||−0.092||0.111||0.912||(0.734, 1.132)|
|Plan: POS||0.008||0.101||1.008||(0.826, 1.229)|
|HMO rate × 10||0.103c||0.022||1.109||(1.062, 1.158)|
|HHI × 10||−0.066a||0.031||0.936||(0.880, 0.996)|
Table 3 presents the results of the multivariable regression analysis on the pricing of colectomy procedures. Adjusting for patient demographics, clinical presentation, hospital characteristics, and attributes of the market structure, the price of laparoscopic colectomy was approximately 7.6% lower than the price of open colectomy (transformed coefficient = 0.927, P < .001). The price of colectomy procedures was significantly higher for sicker patients and those treated in teaching hospitals, but lower for patients who received care in nonprofit hospitals and who had insurance types of HMO or POS. With regard to market structure, an incremental 10% increase in the HMO penetration rate was associated with 1.6% lower price (transformed coefficient = 0.985, P < .001); however, an incremental 0.10 increase in HHl was associated with 1.6% increase in the average price (P = .0015). Finally, the price of colectomy procedures was 17.5% lower in 2007 than in 2002 (transformed coefficient = 0.840, P < .001).
|Variable||Parameter Estimate||Standard Error||Transformed Estimated||95% Confidence Intervals|
|Age (10 year)||−0.005||0.008||0.995||(0.979, 1.011)|
|Patient level severity|
|Elixhauser 1||0.038c||0.014||1.038c||(1.010, 1.067)|
|Elixhauser 2||0.182c||0.019||1.199c||(1.155, 1.246)|
|Elixhauser 3+||0.282c||0.025||1.325c||(1.262, 1.392)|
|Metastasis||0.178 c||0.017||1.195c||(1.156, 1.235)|
|Plan: HMO||−0.055b||0.020||0.947c||(0.910, 0.985)|
|Plan: POS||−0.081c||0.018||0.922c||(0.890, 0.955)|
|HMO rate × 10||−0.016c||0.004||0.985c||(0.977, 0.992)|
|HHI × 10||0.016b||0.005||1.016c||(1.006, 1.027)|
To test for nonlinearity in the market variables HMO penetration and HHI, we repeated our analysis by stratifying each into the categories of Low, Moderate, and High, based on their distribution in our data (Table 4). “Low” constituted the reference category for both variables. Results from our sensitivity analysis were consistent with the results from our linear model, With respect to the likelihood of undergoing laparoscopic versus open colectomy, moderate- and high-level HMO penetration was associated with 26% and 35% greater likelihood to undergo laparoscopic versus open colectomy, respectively (AOR = 1.26, 95% CI = 1.05, 1.51; and AOR = 1.35, 95% CI = 1.12, 1.64). Conversely, increased HHI was consistently associated with a significantly lower likelihood of laparoscopic rather than open procedure (AOR = 0.72, 95% CI = 0.61, 0.86; and AOR = 0.74, 95% CI = 0.62, 0.88), respectively, for moderate and high levels of HHI.
|Use of Laparoscopy: Odds Ratios||Colectomy Price: Transformed Coefficients|
|Moderate (9%-25%) vs low||1.26a (1.05, 1.51)||0.90c (0.90, 0.96)|
|High (>25%) vs low||1.35b (1.12, 1.64)||0.90c (0.90, 0.96)|
|High vs moderate||1.07d (0.91, 1.26)||1.00d (0.97, 1.03)|
|Hospital concentration (HHI)|
|Moderate (0.08-0.15)||0.72c (0.61, 0.86)||1.05c (1.02, 1.09)|
|High (>0.15)||0.74c (0.62, 0.88)||1.06c (1.03, 1.10)|
|High vs moderate||1.02e (0.86, 1.22)||1.01e (0.98, 1.04)|
Regarding price, consistent with the linear model, increased HMO penetration remains significantly associated with lower price (9% lower price in areas with each of moderate and high levels of HMO penetration, P < .001 for both comparisons). For HHI, moderate and high levels of HHI were associated with 5% and 6% higher price, respectively (P < .001 for all of the above associations). Overall, although the effects of HMO penetration and HHI are greater at their respective highest categories, their incremental changes from moderate to high are small, suggesting that the threshold for significance of market intensity occurs at the middle range.
Using claims data of privately insured individuals undergoing colectomy, we show that the price of rapidly diffusing laparoscopic surgery was 7.6% lower than the price of open surgery, adjusting for patient characteristics, hospital characteristics, and market structure. Although we focus on colon cancer, our results have broader implications, suggesting that novel procedures may result in cost savings for patients and insurers.
Our finding of a positive association between HMO penetration and laparoscopy use is in agreement with an earlier study on the use of colonoscopy,25 a diagnostic procedure for colorectal cancer, whereas our finding contrasts with studies of magnetic resonance imaging, which found that high HMO penetration was negatively associated with use10, 23; it is generally assumed that HMOs tend to limit adoption of expensive technologies such as magnetic resonance imaging in an attempt to reduce overall costs.23, 37 However, this hypothesized negative association between HMO penetration and technology adoption rests on the assumption that novel technologies are cost-augmenting. Conversely, in the case of technologies that result in lower purchasing costs, cost-conscious HMOs may be more likely to adopt them, thus a positive association between HMOs and technology adoption would be expected. Given the lower price found for laparoscopic versus open surgery, the positive association between HMO penetration and laparoscopy use is consistent with the latter case. Previous research has also shown that HMO penetration results in discounted hospital prices of cardiac procedures such as angioplasty16 or bypass surgery,15 but similar empirical research for other medical procedures does not appear to exist. We show that the same general result applies to colon cancer surgeries.
A standard way of measuring hospital competition, for both research purposes and in antitrust litigation, is through market concentration, as measured by the HHI.13, 15, 16, 26 Although studies on the relation between HHI and use of medical or surgical procedures are scant, economic theory predicts that lack of competition (that is, more concentration) slows innovation and diffusion of new technologies.11 Our findings on HHI and use of colorectal cancer surgeries conform with this prediction. The positive association between HHI and price found for colectomy implies that market concentration increases prices, or equivalently, that hospital competition lowers prices. Prior studies on angioplasty,16 heart bypass surgery,15 and appendectomy38 pricing found that a 30% increase in hospital HHI results in a 2% to 4% increase in transaction prices. Our estimate for colectomy pricing falls within this range. Combined, our results on the association of HMO penetration or HHI with price are consistent in that competitive pressures from either the purchaser side or the provider side tend to drive prices lower, as predicted in standard economic theory.
Our results carry important implications for health policy and hospital financing beyond the particular procedure studied. First, in certain cases, adoption of novel procedures may not necessarily drive prices higher, but rather, it may result in price discounts and real savings. Second, although medical markets are known to be especially complex, market forces can operate efficiently by lowering costs to consumers. Thus, it is important that in an era of reform, the value of maintaining competition is not overlooked. However, with regard to laparoscopy, the benefits of lower prices should be weighed against longer term outcomes such as survival or cancer recurrence. A limitation of our data is that we are unable to follow patients in the long term. Although we attempted to extrapolate policy implications from our study to the market as a whole, our ability to generalize to other novel procedures is unknown.
Another limitation of the study is the use of claims data that are collected primarily for billing purposes, for which clinical detail is limited. Thus, we were unable to ascertain all factors that potentially affect decisions about type of surgical approach, such as tumor size or local stage versus regional stage cancers. To address this, we included previously validated risk adjusters including comorbidities and distant metastases, which we were able to identify in the data, and find that our results are stable when high-risk cases are excluded. Furthermore, we minimized unmeasured bias due to patient selection by using propensity scores. Finally, an advantage of the claims database is that it allowed us to accurately identify payments made by the insurer to the hospital. Although the analysis employed a large national sample of privately insured patients, we note that the results may not apply to the Medicare segment of markets, where prices are administratively determined.
No specific funding was disclosed.
CONFLICT OF INTEREST DISCLOSURE
Drs. Dor, Koroukian, Xu, and Cooper were supported by the National Cancer Institute award R01CA129766 to George Washington University “Pricing of Major Cancer Surgeries: Impact of Insurance, Outcomes, and Severity.”
- 9US Department of Health & Human Services, Centers for Medicare & Medicaid Services. Medicare hospital compare quality of care. http://www.hospitalcompare.hhs.gov. Accessed August 19, 2011.
- 11Competition, imitation and growth with step-by-step innovation. Rev Econ Stud. 2001; 68: 467-492., , , .Direct Link:
- 19http://thomsonreuters.com/content/healthcare/pdf/white_papers/marketscan_white_paper. Published January 2011. Accessed August 24, 2011., . Health Research Data for the Real World: The Thomson Reuters MarketScan® Databases. Pharmaceutical Division, Thomson Reuters.
- 27The Dartmouth Atlas of Health Care. Geographic boundary files: hospital referral region boundaries. http://www.dartmouthatlas.org/tools/downloads.aspx. Accessed June 15, 2011.
- 30Constructing matched sets and strata. In: Observational Studies. 2nd edition. New York, NY: Springer-Verlag; 2002: 294-331..
- 31Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. Am Stat. 1985; 39: 33-38., .
- 33Performing a 1:N case-control match on propensity score. In: Proceedings of the 29th Annual SAS Users Group International Conference; May 9–12, 2004; Montreal, Canada. Paper 165–29..