Objectives To evaluate associations among hospital volume, costs, and length of stay (LOS) and to assess whether reduced hospital cost of care adversely affected quality of care.
Design Four-year, nationwide, population-based study.
Setting Data were obtained from claims submitted to the South Korean National Health Insurance database.
Patients We identified 48 938 patients at 274 hospitals who had undergone gastric resection from January 1, 2002, through December 31, 2005. Hospital volumes were divided into quartiles.
Main Outcome Measures Patient demographics and socioeconomic and clinical variables were investigated as factors that might affect costs and LOS.
Results Independent predictors of higher costs and longer LOS included older age, increased Charlson score, and hospitals with fewer beds. After adjusting for relevant factors, an inverse relationship between volume and costs or LOS was found such that higher-volume hospitals had the lowest procedure costs and LOS. Results showed no association between hospital cost and quality of care.
Conclusions Higher hospital volume is predictive of lower costs and LOS for patients undergoing gastric resection. By referring these patients to high-volume centers, we may improve quality of care and reduce costs. Furthermore, high-quality care can be maintained when costs are lowered due to high volume.
The incidence of gastric cancer is decreasing worldwide; however, it remains a common cancer type in developed countries.1 In South Korea, gastric cancer is the most common type of cancer and is second only to lung cancer in cancer mortality.2 Thus, almost 16 000 gastrectomies are performed annually, at a cost in 2008 of more than $70 000 as payment for insurance-covered services. Furthermore, gastric cancer treatment is responsible for a large portion of medical expenses, accounting for 14% of all cancer-related payments for insurance-covered services.3
As the number of gastrectomies increases, interest is growing regarding hospital factors that result in better outcomes. Although the number of resections performed at a hospital cannot be a direct indicator of outcome, the need for monitoring surgical quality in hospitals with lower procedure volume has arisen because a previous report4 from the United States demonstrated that a lower number of operations at a hospital is associated with lower surgical quality. In fact, a high possibility of qualitative differences in outcomes exists during 1 year between low-volume and high-volume hospitals, defined as those that have performed 10 and 200 gastrointestinal operations, respectively. All patients desire high-quality service in a hospital. Two ways to enable this outcome would be for medical institutions to invest dedicated efforts toward improving service quality and to provide evidence-based information to aid patients in choosing high-quality hospitals with lower costs.5
Assessment of outcomes is critical for evaluating medical services for patients with cancer.6,7 Outcome end points help prevent decisions made regarding reputation only, allowing consumers to choose more qualified health care professionals, which, over time, will lead to positive changes in the overall quality of care provided as improvements are implemented. Thus, outcome assessment can reveal the quality of services provided by health care professionals and help determine the effectiveness of treatment. For these reasons, some countries use outcome indicators for cancer operations as an criterion for evaluation of the quality of care provided by medical institutions.8 Although the primary impetus for this effort is to reduce mortality rates, many also assume that high-volume hospitals will provide surgical procedures at lower cost than low-volume hospitals due to more favorable economies of scale.9,10 Although previous studies4,11-14 have examined the relationship between hospital volume and costs, only a few have examined this relationship for gastrectomy, and none have examined volume and cost data for gastrectomy in an Asian population, to our knowledge. Other studies have examined the relationship between volume and length of stay (LOS). High volume in hospitals is generally associated with shorter LOS, but this relationship is based on relatively old data or information obtained from individual states, making it difficult to generalize.4,12-14 In this study, we used the National Health Insurance (NHI), a South Korean national administrative database, to examine the relationship between hospital volume and total cost and LOS, after adjusting for demographic characteristics and hospital-level variability of patients undergoing gastrectomy. We also assessed the extent to which resource use is associated with 1 measure of quality of care for gastrectomy, the hospital standard mortality ratio (HSMR; observed mortality to expected mortality).
Using NHI claims data, which cover almost the entire South Korean population of 48 431 000, we identified patients who had undergone gastrectomy from January 1, 2002, through December 31, 2005. The government of South Korea had launched a compulsory NHI program for the entire population in 1989, so the program has unique claims data for medical services. These data included only the payments for insurance-covered services. The NHI data also include patient sociodemographic information, such as sex, age, monthly NHI premium, the residential area of all health services facilities visited, comorbid diseases, and other health-related information, such as specific surgical procedures performed and course of admission.
The admission for each patient in our analysis who underwent gastrectomy was identified using appropriate procedure codes from the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM).15,16 Gastrectomy procedures included in the analysis included partial and total gastrectomy.
To ensure that our data included only gastrectomies and to increase the homogeneity of the study sample, we excluded patients whose files did not contain an accompanying cancer code related to the indexed procedure in the primary diagnosis. Secondary diagnostic codes were abstracted to enumerate comorbid conditions according to the Charlson comorbidity index, which has been validated as a favorable instrument to predict clinical outcomes, costs, and use of resources.17,18 The monthly NHI premium is income based and serves as a reasonable proxy for income. The route of admission separated patients into emergency or nonemergency categories based on admission from the emergency or the outpatient department. All variables were coded as being of the categorical or dummy type.
The outcome variables were costs and mean LOS. These variables were used to compare the relative use of resources among hospitals that performed differing volumes of cancer procedures. In this study, costs were defined as the cost per episode for surgery and were adjusted for inflation to 2005 premiums using the NHI Corporation Input Price Indices.19 For the index admission, the LOS was defined as the period from the index procedure to hospital discharge. We also examined HSMRs as an indicator of gastrectomy care quality. These ratios were calculated as the ratio of the actual number of gastrectomy deaths at each hospital to the expected number of deaths, multiplied by 100. The ratio is adjusted for other factors that affect mortality, such as age, sex, duration of hospital stay, admission course, principal diagnosis, and comorbidities.20
The number of procedures performed at each hospital from 2002 to 2005 was determined using a unique hospital identification code. We assessed hospital volume by calculating the total number of each type of procedure performed at a given hospital during the 4-year study period. Because we had access to a nationwide database that included almost the entire South Korean population, we were able to count the actual number of hospital procedures performed at each hospital. For each procedure, the hospitals were first ranked in order of increasing total volume as a continuous variable, after which volume cutoff points were selected to create volume groups with approximately equal numbers of patients. This method is consistent with those used in previous studies.21,22 Hospital volume was stratified into quartiles (very high, high, low, and very low volume).
Descriptive analyses were performed to clarify the distributions of patient demographics, hospital volumes, and economic outcomes. The differences in patient characteristics were compared across hospital volume groups using the χ2 statistic for categorical variables. Bivariate analyses with analysis of covariance (Bonferroni procedure) were performed to determine differences in costs and LOS among the 4 groups. Associations between costs and HSMRs were determined by calculation of correlation coefficients (Pearson r).
We used multiple regression to examine the relationship between hospital volume and economic outcome (a continuous variable) after adjusting for patient demographics and clinical information. For each outcome, a series of multiple regression models examined the degree to which our study variables mediated the volume-to-outcome relationship for hospitals. In the process of regression modeling, we tested all demographic and clinical hospital variables. Final regression models included all the original variables that were significant predictors of the outcome or that improved the model fit. Preliminary analyses showed that the distributions of economic outcome indicators (costs and LOS) were skewed substantially to the right, thus violating the normality assumption of our regression model. Accordingly, we performed a natural log transformation to achieve a more normal distribution. All statistical analyses pertaining to costs and LOS were based on the log-transformed data. P < .05 was deemed to indicate statistical significance. Data analyses were performed using SPSS statistical software, version 12.0 (SAS Institute Inc, Cary, North Carolina).
Patient characteristics and clinical information
Between 2002 and 2005, 48 938 patients underwent gastrectomy at 274 hospitals. Table 1 gives the distribution of patient characteristics by hospital volume.
Patients undergoing gastrectomy had a mean age of 58 years, and 32 902 (67.2%) were men. The 4 groups were similar with respect to age and sex. Patients treated at very low–volume hospitals were the most likely to have the lowest monthly insurance premiums (30.5%; < 28 010 won [$25.06]); patients treated in very high–volume hospitals had the lowest percentage (19.3%) of individuals with the lowest premiums. Similarly, residents of rural areas were more likely to undergo surgery at a lower-volume hospital than were residents of metropolitan areas. Furthermore, patients at higher-volume hospitals were more likely to undergo elective surgery, and patients with a comorbidity index score of greater than 2 tended to be more prevalent. Finally, most patients undergoing gastrectomy at teaching hospitals were in metropolitan areas and private hospitals with more than 699 beds. Significant differences in case mix were found among the hospital volume groups in all case mix variables.
Increased hospital case volume was associated with improvement in 2 outcome measures (Figure 1). After adjusting for case mix, results showed a statistically significant association between volume and mean cost for total and partial gastrectomy (P < .001). The mean costs for both procedures were significantly higher at very low–volume hospitals than at very high–volume hospitals ($5011 vs $3609 and $3532 vs $2578, respectively; P < .001).
Table 2 gives the regression estimates of the cost differences for the 2 procedures by hospital volume, adjusting for patient and hospital characteristics. Low-volume, high-volume, and very high–volume hospitals performed gastrectomies at lower cost than did very low–volume hospitals (P < .001). Although the addition of patient demographics, clinical characteristics, and hospital characteristics partly explained the association between hospital volume and costs, very high–volume hospitals maintained their association with decreased costs in the fully adjusted regression model.
Regression analysis revealed that independent predictors of higher costs included male sex, older age, metropolitan residence, emergency admission, Charlson score greater than or equal to 2, private ownership of hospital, metropolitan hospital location, teaching hospitals, and hospitals with fewer than 500 beds (Table 3). In the cost regression model, high-volume and very high–volume hospitals showed significant effects. The explanatory power of the regression model was 62.1%, after adjusting for all included variables.
The case mix–adjusted LOS was reduced in the very high–volume group compared with the very low–volume group according to procedure (Figure 1). The mean LOS was shorter for partial gastrectomy (13.63 days) than for total gastrectomy (24.59 days). In general, patients who underwent total gastrectomy had longer mean LOS than those who underwent partial gastrectomy. Statistically significant associations between volume and LOS were found for both procedures (P < .001).
Table 2 gives the regression estimates of LOS differences for the 2 procedures according to hospital volume, after adjustment for patient and hospital characteristics. Patients in low-volume, high-volume, and very high–volume hospitals had significantly lower LOS than did patients in very low–volume hospitals (P < .001). Although the addition of patient demographics, clinical characteristics, and hospital characteristics partly explained the association between hospital volume and LOS, very high–volume hospitals maintained their association with decreased LOS in the fully adjusted regression model.
Multiple regression analysis revealed that independent predictors of longer LOS included female sex, greater age, low insurance premium, nonmetropolitan residence, routine admission, Charlson score greater than or equal to 2, public ownership of hospital, metropolitan hospital location, nonteaching hospitals, and hospitals with fewer than 500 beds (Table 3). In the LOS regression model, hospital volume consistently showed statistically significant effects. The explanatory power of the regression model was 54.6% after adjusting for all included variables.
We examined the relationship between the mean cost of hospital treatment and HSMR to determine whether a systematic relationship between cost and quality of care exists. No association was found between cost and HSMR (correlation coefficient, 0.02; P = .80; Figure 2).
Our results suggest that although gastrectomy is performed with a high degree of safety within the South Korean health care system, certain hospital factors, such as volume, are associated with inverse economic outcomes. This study serves as a reminder that hospital volume is an important predictor of economic outcomes, consistent with previously published data.13,23,24 Specifically, patients treated at very high–volume hospitals had lower costs and shorter LOS compared with those treated at very low–volume hospitals after case mix adjustment. Patient characteristics (eg, sex, monthly insurance premium, and residential area), clinical characteristics (eg, admission course and Charlson score), and hospital characteristics (eg, ownership, number of beds, and teaching status) were all associated with cost and LOS.
We also explored whether hospital care cost was associated with better outcomes. We found no correlation between costs and 1 measure of gastrectomy care quality, namely, hospital standardized mortality rates. These findings are consistent with a recent study25 that used data from national databases to examine hospital cost and quality of care for 2 common diagnoses (congestive heart failure and pneumonia) and found that mortality rates in higher-volume hospitals were low, suggesting that lower-cost hospital care is not necessarily lower in quality. When considered with our current findings that costs and LOS were lower in higher-volume hospitals, these data suggest that high-quality care can be delivered at lower costs in high-volume hospitals. However, Chen et al25 targeted internal medicine procedures, in which the bulk of the care, examinations, and medication do not contribute much to costs. By contrast, our study targeted surgical operations, which are relatively resource intensive and costly, providing strong evidence that high-quality care can be delivered at low cost.
In our analysis of the association among hospital volume, costs, and LOS, we found that after adjustment for various patient, clinical, and hospital factors, very high–volume hospitals have lower costs and shorter LOS than do very low–volume hospitals. This finding may be explained by the learning effect theory, which states that increasing the number of treatments can lead to a reduction in costs and LOS for the treatment in question due to improved efficiency in early medical decisions and that hospitals with high volumes have fewer complications, leading to shortened LOS.13,22,26 However, in addition to these factors, per standard practice in the South Korean medical system, high-volume hospitals have a tendency to switch hospital care to home care through early discharge to increase bed turnover rate, which appears to have contributed to the decreased LOS in high-volume hospitals, which, in turn, lowers costs.
Multivariate regression was used to rank variables according to their strength of association with the outcome. Greater age, increased Charlson score, and hospitals with fewer beds were significantly associated with higher costs and longer LOS. Initially, we expected costs and LOS to be low for populations with low socioeconomic status, but this study showed that such populations are associated with decreased costs and increased LOS. This finding may be explained by the fact that gastric cancer is more common in the population with the lowest socioeconomic status and that most of those individuals undergo operations in public hospitals.
It is a commonly held belief that hospital ownership (ie, public or private) is a major factor in determining the type of treatment and in the finding that treatment costs are lower at public hospitals. We show herein that costs are lower and LOS is longer at public hospitals. The longer LOS may reflect attempts to maintain hospital income by increasing bed occupation; public hospitals have a greater probability of longer hospitalization for the same operation. Furthermore, higher costs and longer LOS are found in highly competitive metropolitan hospitals. This finding may be due to larger staffs and more services, both of which require large capital expenditure.27
Hospitals with similar costs or LOS may differ in the ways that resources are used during hospitalization, and intensity of care is known to differ between teaching and nonteaching hospitals.28 In general, patients treated at teaching hospitals tend to receive more diagnostic tests and to spend more time in the intensive care unit.24 Thus, teaching hospitals have higher costs than do nonteaching hospitals. In addition, large portions of teaching hospitals belong to tertiary hospitals. Therefore, it is likely that patients are discharged immediately once acute care is completed to maximize benefits, which explains the relatively short LOS in teaching hospitals.
Gastric cancer is not an acute disease and generally does not require emergency surgery. However, emergency surgery is necessary when significant hemorrhage in the cancerous lesion or peritonitis due to gastric perforation occurs. For this reason, the average LOS was first estimated to be long in patients who received treatment after emergency admission. However, LOS was shorter than expected in this population. Our findings suggest that this result may occur because patients had undergone all examinations in the emergency department. In most cases, although patients received treatment for gastric cancer, emergency admissions occurred because of intestinal infectious diseases, malnutrition, and hypertensive disease.
Costs increased and LOS decreased over time. This finding seems to occur because of the introduction of newer drugs and improved surgical and diagnostic techniques, which are more expensive but result in shorter LOS.24,29
Our study had some limitations. First, we used an administrative claims database that lacked information regarding cancer-specific clinical severity, such as disease stage and tumor size; thus, some residual confounding as a result of these covariates is possible.30,31 However, we minimized this problem by only including patients who received major gastrectomy; thus, the study participants were likely to have similar cancer stages. Second, because the study was based on limited cross-sectional research using data from 2002 to 2005, the generalizability of the observed relationships is limited. Therefore, it is necessary to confirm our findings with further studies across different periods and in-depth serial research over time. Despite these limitations, our results have important implications from a health policy perspective. To our knowledge, our study is the first population-level description of economic outcomes and of the factors, including volume, that affect costs and LOS after gastrectomy. In addition, this study provides a basis for cost-effective improvements in surgical intervention in gastric cancer treatment. Finally, our data provide useful information regarding costs and treatment quality for patients selecting a hospital for gastrectomy.
Correspondence: Jong Hyock Park, MD, MPH, PhD, Division of Cancer Policy and Management, National Cancer Control Research Institute, National Cancer Center, 809 Madu 1-dong, Ilsandong-gu, Goyang-si, Gyeonggi-do 410-769, South Korea (whitemiso@ncc.re.kr).
Accepted for Publication: February 7, 2011.
Published Online: April 18, 2011. doi:10.1001/archsurg.2011.81
Author Contributions:Study concept and design: J. A. Lee, Park, Y. Kim, and S. I. Lee. Acquisition of data: J. A. Lee and Park. Analysis and interpretation of data: J. A. Lee, Park, E. J. Lee, and S. Y. Kim. Drafting of the manuscript: J. A. Lee, E. J. Lee, S. Y. Kim, Y. Kim, and S. I. Lee. Critical revision of the manuscript for important intellectual content: J. A. Lee, Park, and S. Y. Kim. Statistical analysis: J. A. Lee, E. J. Lee, and S. Y. Kim. Obtained funding: Park. Administrative, technical, and material support: J. A. Lee, Park, and S. I. Lee. Study supervision: J. A. Lee, Park, S. Y. Kim, and Y. Kim.
Financial Disclosure: None reported.
Funding/Support: The study was supported by grants 0910191 and 0910192 from the National Cancer Center.
Additional Information: The English in this document has been checked by at least 2 professional editors, both native speakers of English. For certification, see http://www.textcheck.com/certificate/3FkO2S.
1.Parkin DM, ed, Whelan SL, ed, Ferlay J, ed, Teppo L, ed, Thomas DB, ed. Cancer Incidence in Five Continents. Vol 8. Lyon, France: International Agency for Research on Cancer Scientific Publications; 2002. IARC publication 155
2.Shin H-R, Won Y-J, Jung K-W,
et al; Members of the Regional Cancer Registries. Nationwide cancer incidence in Korea, 1999~2001; first result using the national cancer incidence database.
Cancer Res Treat. 2005;37(6):325-33119956367
PubMedGoogle ScholarCrossref 4.Gordon TA, Bowman HM, Bass EB,
et al. Complex gastrointestinal surgery: impact of provider experience on clinical and economic outcomes.
J Am Coll Surg. 1999;189(1):46-5610401740
PubMedGoogle ScholarCrossref 5.Marks RB, Totten JW. The effects of mortality cues on consumers' ratings of hospital attributes.
J Health Care Mark. 1990;10(3):4-1210107468
PubMedGoogle Scholar 6.Glance LG, Li Y, Osler TM, Dick A, Mukamel DB. Impact of patient volume on the mortality rate of adult intensive care unit patients.
Crit Care Med. 2006;34(7):1925-193416715030
PubMedGoogle ScholarCrossref 8.Iezzoni LI, Ash AS, Shwartz M, Landon BE, Mackiernan YD. Predicting in-hospital deaths from coronary artery bypass graft surgery: do different severity measures give different predictions?
Med Care. 1998;36(1):28-399431329
PubMedGoogle ScholarCrossref 9.Birkmeyer JD, Lucas FL, Wennberg DE. Potential benefits of regionalizing major surgery in Medicare patients.
Eff Clin Pract. 1999;2(6):277-28310788026
PubMedGoogle Scholar 10.Phillips KA, Luft HS. The policy implications of using hospital and physician volumes as “indicators” of quality of care in a changing health care environment.
Int J Qual Health Care. 1997;9(5):341-3489394202
PubMedGoogle ScholarCrossref 11.Smith DL, Elting LS, Learn PA, Raut CP, Mansfield PF. Factors influencing the volume-outcome relationship in gastrectomies: a population-based study.
Ann Surg Oncol. 2007;14(6):1846-185217406947
PubMedGoogle ScholarCrossref 12.Gordon TA, Burleyson GP, Tielsch JM, Cameron JL. The effects of regionalization on cost and outcome for one general high-risk surgical procedure.
Ann Surg. 1995;221(1):43-497826160
PubMedGoogle ScholarCrossref 13.Swisher SG, Deford L, Merriman KW,
et al. Effect of operative volume on morbidity, mortality, and hospital use after esophagectomy for cancer.
J Thorac Cardiovasc Surg. 2000;119(6):1126-113210838528
PubMedGoogle ScholarCrossref 14.Balcom JH IV, Rattner DW, Warshaw AL, Chang Y, Fernandez-del Castillo C. Ten-year experience with 733 pancreatic resections: changing indications, older patients, and decreasing length of hospitalization.
Arch Surg. 2001;136(4):391-39811296108
PubMedGoogle ScholarCrossref 15.World Health Organization. International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). Geneva, Switzerland: World Health Organization; 1977
16.Korean Hospital Association. Fee Schedule of National Health Insurance [in Korean]. Seoul, South Korea: Korean Hospital Association; 2009
17.Charlson M, Szatrowski TP, Peterson J, Gold J. Validation of a combined comorbidity index.
J Clin Epidemiol. 1994;47(11):1245-12517722560
PubMedGoogle ScholarCrossref 18.Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation.
J Chronic Dis. 1987;40(5):373-3833558716
PubMedGoogle ScholarCrossref 20.Jarman B, Bottle A, Aylin P, Browne M. Monitoring changes in hospital standardised mortality ratios.
BMJ. 2005;330(7487):32915705689
PubMedGoogle ScholarCrossref 21.Neighbors CJ, Rogers ML, Shenassa ED, Sciamanna CN, Clark MA, Novak SP. Ethnic/racial disparities in hospital procedure volume for lung resection for lung cancer.
Med Care. 2007;45(7):655-66317571014
PubMedGoogle ScholarCrossref 22.Begg CB, Cramer LD, Hoskins WJ, Brennan MF. Impact of hospital volume on operative mortality for major cancer surgery.
JAMA. 1998;280(20):1747-17519842949
PubMedGoogle ScholarCrossref 23.Dimick JB, Cattaneo SM, Lipsett PA, Pronovost PJ, Heitmiller RF. Hospital volume is related to clinical and economic outcomes of esophageal resection in Maryland.
Ann Thorac Surg. 2001;72(2):334-34111515862
PubMedGoogle ScholarCrossref 24.Goodney PP, Stukel TA, Lucas FL, Finlayson EVA, Birkmeyer JD. Hospital volume, length of stay, and readmission rates in high-risk surgery.
Ann Surg. 2003;238(2):161-16712894006
PubMedGoogle Scholar 25.Chen LM, Jha AK, Guterman S, Ridgway AB, Orav EJ, Epstein AM. Hospital cost of care, quality of care, and readmission rates: penny wise and pound foolish?
Arch Intern Med. 2010;170(4):340-34620177036
PubMedGoogle ScholarCrossref 26.Hillner BE, Smith TJ, Desch CE. Hospital and physician volume or specialization and outcomes in cancer treatment: importance in quality of cancer care.
J Clin Oncol. 2000;18(11):2327-234010829054
PubMedGoogle Scholar 27.Xirasagar S, Lin H-C. Cost convergence between public and for-profit hospitals under prospective payment and high competition in Taiwan.
Health Serv Res. 2004;39(6, pt 2):2101-211615544646
PubMedGoogle ScholarCrossref 28.Rosenthal GE, Harper DL, Quinn LM, Cooper GS. Severity-adjusted mortality and length of stay in teaching and nonteaching hospitals: results of a regional study.
JAMA. 1997;278(6):485-4909256223
PubMedGoogle ScholarCrossref 29.Hay JW. Hospital cost drivers: an evaluation of 1998-2001 state-level data.
Am J Manag Care. 2003;9(spec 1):SP13-SP2412817612
PubMedGoogle Scholar 30.Birkmeyer JD, Siewers AE, Finlayson EVA,
et al. Hospital volume and surgical mortality in the United States.
N Engl J Med. 2002;346(15):1128-113711948273
PubMedGoogle ScholarCrossref 31.Aranda MA, McGory M, Sekeris E, Maggard M, Ko C, Zingmond DS. Do racial/ethnic disparities exist in the utilization of high-volume surgeons for women with ovarian cancer?
Gynecol Oncol. 2008;111(2):166-17218829086
PubMedGoogle ScholarCrossref