Abstract
BACKGROUND
Surgical resection is the primary treatment for colon cancer, but utilization of laparoscopic approaches varies widely despite demonstrated short- and long-term benefits.
OBJECTIVE
To identify characteristics associated with laparoscopic colon cancer resection and quantify variation based on patient, hospital, and geographic characteristics.
DESIGN
Bayesian cross-classified, multilevel logistic models calculated adjusted odds ratios and confidence intervals for patient, surgeon, hospital, and geographic characteristics and unexplained variability (predicted vs. observed values) using adjusted median odds ratios for hospitals and counties.
SETTINGS
Surveillance, Epidemiology, and End Results-Medicare claims database (2008–2011) supplemented with county-level American Community Survey (2008–2012) demographic data
PATIENTS
10,618 colon cancer resection patients ≥ 66 years old
INTERVENTION
none
MAIN OUTCOME MEASURES
Non-urgent/non-emergent resections for colon cancer patients ≥ 66 years old were classified as laparoscopic or open procedures
RESULTS
Patients resided in 579 counties and utilized 950 hospitals; 47% of patients underwent laparoscopic surgery. Medicare/Medicaid dual enrollment, age ≥ 85 years, higher tumor stage and grade were negatively associated with laparoscopic surgery receipt; proximal tumors and increasing hospital size and surgeon caseload were positively associated. Significant unexplained variability at the hospital (adjusted median odds ratio = 3.31, P < 0.001) and county (adjusted median odds ratio = 1.28, P < 0.05) levels remained after adjustment.
LIMITATIONS
Observational study lacking generalizability to younger patients without Medicare or those with Health Maintenance Organization coverage, dataset did not reflect national hospital studies or hospital volume, unable to account for specific types of comorbidities such as obesity, and broad categories for surgeon caseload.
CONCLUSIONS
Determining sources of hospital-level variation among poor insured patients may help increase laparoscopic resection to maximize health outcomes and reduce cost. See Video Abstract at http://links.lww.com/DCR/AXXX.
Keywords: Colon cancer, Disparity, Geographic variation, Hospital size, Laparoscopic resection utilization, Multilevel analysis, SEER-Medicare
Introduction
Minimally invasive laparoscopic surgery for colon cancer has been used more frequently for some patients after clinical trials demonstrated its safety and oncological similarity to open surgery. Short-term benefits of laparoscopic resection include reduced analgesia,1 fewer complications,1–4 decreased post-operative intensive care/skilled nursing,3,4 and shorter length-of-stay1–6 without increased rates of readmission or emergency visits.7 Several,2,3,5,6,8 but not all,4 studies have documented decreased 30-day mortality. Compared to open resection, laparoscopic colectomy is associated with similar5,6,9–13 or better8 long-term oncological and survival outcomes and cost savings.4,14,15
Despite these benefits, laparoscopic surgery has not been widely adopted for eligible colon cancer patients, and even among those who receive laparoscopic colectomy, unexplained hospital-level variation in outcomes (hospital center effect) exists.16 Previous studies have shown extensive variability in laparoscopic surgery utilization based on characteristics of patients [e.g., age, gender, race/ethnicity, socioeconomic status (SES), payer/insurance type, disease stage, comorbidity], hospitals (e.g., size/volume, teaching hospital status, urban/rural, geographic region)3,5,8,17–21 and surgeon caseload.22 To our knowledge, no studies have examined small area variability (e.g., at the county level) in the use of laparoscopic surgery in the US. Moreover, no studies concurrently examined variation across patients, hospitals, and geography.
This study describes variability in use of laparoscopic (vs. open) surgery for colon cancer among Medicare patients. Identifying the level (patient, hospital, and/or geographic unit) at which variation is occurring and the characteristics associated with underutilization of laparoscopic surgery for colon cancer would help direct interventions aimed at decreasing disparities in availability and access—interventions that might have potential for cost savings and improving healthcare quality.
Materials and Methods
Data Sources
Data were from the most recent release of an existing linkage of the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) database with Medicare claims data (2008–2011), supplemented with county-level demographic data from the 2010 US Census and the 2008–2012 American Community Survey (ACS). SEER is a nationally representative collection of population-based cancer registries collecting demographic, diagnostic, initial treatment, and survival data for 28% of the US population.23 This study included 17 SEER registries: Connecticut, Iowa, New Mexico, Utah, Hawaii, New Jersey, Kentucky, Louisiana, Detroit, San Francisco-Oakland, Atlanta, Seattle-Puget Sound, Rural Georgia, Greater Georgia, Los Angeles County, San Jose-Monterey, and Greater California. The Centers for Medicare and Medicaid Services (CMS) collect claims information for covered healthcare services from the time of an individual’s Medicare enrollment until death. Use of the linked database provides detailed claims-related information such as physician- and hospital-level variables as well as the ascertainment of comorbidities prior to diagnosis which are unavailable through the SEER cancer registry. The Institutional Review Boards approved this study as non-human subjects research exempt from oversight.
Patient sample
Eligibility included patients aged 66 years and older who underwent non-urgent/non-emergent surgical resection within 90 days of diagnosis for a first primary in-situ or invasive colon cancer diagnosed during 2009–2011, who had both Medicare Parts A and B coverage during this period. Patients with unknown surgery type were excluded. Including patients aged 66+ years allowed for a complete year of claims data (from 2008 records) prior to diagnosis to determine comorbidities. Diagnosis during 2009–2011 was selected because the ICD-9 procedure codes for laparoscopic colon resection became available on October 1, 2008. We excluded patients who were members of a Health Maintenance Organization (HMO) or had incomplete Medicare Parts A and B coverage due to lack of claims data. Similarly, patients identified by autopsy or death certificate were excluded from analysis. We identified urgent/emergent colon cancer patients based ICD-9 codes for bowel obstruction (56089, 5609), perforation (56983), hemorrhage (5693, 5789), peritonitis (5670, 5672, 5678, 5679), or emergency inpatient admission (see Text, Supplemental Digital Content 2, which details the exclusion criteria for urgent, emergent, and emergency procedures).24
Type of surgery
Initial surgery was classified as laparoscopic, open, or converted-to-open (V64.4 and V64.41) based on ICD-9 procedure codes (see Table, Supplemental Digital Content 3, which lists the ICD-9 procedure codes used to identify surgery type) and CPT codes (see Table, Supplemental Digital Content 4, which lists the surgeon CPT codes used to identify surgery type).21,25,26 Surgery type was dichotomized for analysis based on intent-to-treat as laparoscopic, including converted-to-open, (vs. open) surgeries.
Covariates
Covariates were based upon patient, surgeon, hospital, and county characteristics thought to explain variability at different levels.27,28 Patient-level characteristics included: age, sex, race/ethnicity (excluding Alaska Natives due to few cases), number of comorbidities measured by the Elixhauser method29 any use of Medicaid during diagnosis year (proxy measure for low income), American Joint Commission on Cancer (AJCC)30 tumor stage, grade, location, and histology. Tumor location was defined as proximal colon (cecum, ascending), transverse colon (hepatic flexure, transverse colon, splenic flexure), or distal colon (descending colon, sigmoid colon, overlapping lesion, or colon NOS). Surgeon and hospital characteristics obtained from the CMS Healthcare Cost Report and Provider of Service files (2008–2011) included mean annual surgeon caseload of colon cancer resections for patients in this sample, regardless of surgery type or hospital affiliation, hospital size (total number of hospital beds), and hospital teaching status. County-level poverty was based on the percentage of the population living in poverty for each patient’s county of residence at time of diagnosis using ACS data (2008–2012). The county urban-rural variable was based on census data using collapsed categories of the 2003 Rural/Urban Continuum Code. Patients with missing data were excluded listwise.
Statistical analysis
Chi-square tests were used to compare descriptive characteristics by surgery type. Mean annual surgeon caseload was treated as a hospital-level characteristic due to model complexity and based on the skewed distribution of the number of patients nested within surgeons. To account for lack of independence of patients within hospitals and within patient residential counties, we used a Bayesian cross-classified multilevel logistic model, which allows nesting within multiple separate, unrelated groups.31 Covariates were forced into the models to calculate adjusted odds ratios (aOR) and 95% confidence intervals (CI). Deviance Information Criterion was used to assess model fit, with lower values indicating better fit. WinBUGS (version 1.4.3, MRC, UK) was used to perform 5,000 burn-in iterations of the cross-classified Bayesian analysis, with an additional 5,000 iterations retained for parameter estimates.
To facilitate interpretation of the variance across hospitals and counties on a scale directly comparable to odds ratios for other covariates in the model, we calculated the median odds ratio (MOR).32 Based on the random effects variance component (V) from the regression model, can be interpreted as the median value of the ratio of predicted odds of the outcome for two patients randomly selected from different hospitals (or counties) with equivalent covariates. An MOR of 1 indicates no variation in outcome across hospitals (or counties). Due to the 1-tailed testing of heterogeneity measures, 95% credible intervals (similar to confidence intervals using the Frequentist approach) are not reported and 1-tailed p-values are used to indicate statistical significance.33
We also estimated variability in laparoscopic surgery utilization across hospitals and counties with at least 30 patients. Predicted probability of receiving laparoscopic surgery was computed for each patient using model coefficients. These individual predicted probabilities were then averaged for all patients treated at a particular hospital and residing in a particular county to create hospital- and county-level predicted values. These predicted values were compared to the observed values across hospitals and counties.
Results
Data were analyzed for 10,618 patients residing in 579 counties who underwent non-urgent/non-emergent colon cancer resection at 950 hospitals. Most patients were female (57.9%), white (84.0%), not dual enrolled in Medicaid (82.8%), and < 85 years (81.2%) (Table 1). AJCC Stage II tumors were the most common (33.2%). The majority of tumors were moderately differentiated (65.2%), in the proximal colon (55.4%), and non-mucinous adenocarcinoma (87.7%). Surgeons with annual caseloads <10 performed 43.4% of all the surgeries in this analysis. Most patients (63.9%) resided in counties with 10–19% poverty.
Table 1.
Patient, surgeon, hospital, and geographic characteristics by surgery type among SEER-Medicare colon cancer patients, 2009–2011
| Variable Characteristics | Total patients N = 10,618 (%)a |
Laparoscopic n = 5,040 (47%)b |
Open n = 5,578 (53%)b |
P |
|---|---|---|---|---|
| Patient Characteristics | ||||
| Sex** | 0.002 | |||
| Male | 4471 (42.1) | 2202 (49.3) | 2269 (50.7) | |
| Female | 6147 (57.9) | 2838 (46.2) | 3309 (53.8) | |
| Race** | 0.009 | |||
| White | 8924 (84.0) | 4262 (47.8) | 4662 (52.2) | |
| African American | 923 (8.7) | 394 (42.7) | 529 (57.3) | |
| Other/Unknown | 771 (7.3) | 384 (49.8) | 387 (50.2) | |
| Comorbidity 30*** | 0.001 | |||
| 0 | 2452 (23.1) | 1197 (48.8) | 1255 (51.2) | |
| 1 | 3258 (30.7) | 1608 (49.4) | 1650 (50.6) | |
| 2+ | 4908 (46.2) | 2235 (45.5) | 2673 (54.5) | |
| Medicaid (dual eligibility)*** | <0.0001 | |||
| Yes | 1831 (17.2) | 747 (40.8) | 1084 (59.2) | |
| No | 8787 (82.8) | 4293 (48.9) | 4494 (51.1) | |
| Age group*** | <0.0001 | |||
| 66 – 74 | 3970 (37.4) | 1971 (49.6) | 1999 (50.4) | |
| 75 – 84 | 4647 (43.8) | 2201 (47.4) | 2446 (52.6) | |
| 85+ | 2001 (18.8) | 868 (43.4) | 1133 (56.6) | |
| AJCC tumor stage*** | <0.0001 | |||
| 0 | 189 (1.8) | 113 (59.8) | 76 (40.2) | |
| I | 2803 (26.4) | 1562 (55.7) | 1241 (44.3) | |
| II | 3523 (33.2) | 1622 (46.0) | 1901 (54.0) | |
| III | 2813 (26.5) | 1295 (46.0) | 1518 (54.0) | |
| IV | 1070 (10.1) | 348 (32.5) | 722 (67.5) | |
| Unknown | 220 (2.1) | 100 (45.4) | 120 (54.6) | |
| Tumor grade*** | <0.0001 | |||
| Well-differentiated | 907 (8.5) | 481 (53.0) | 426 (47.0) | |
| Moderately differentiated | 6922 (65.2) | 3372 (48.7) | 3550 (51.3) | |
| Poorly differentiated | 2014 (19.0) | 852 (42.3) | 1162 (57.7) | |
| Undifferentiated | 337 (3.2) | 121 (35.9) | 216 (64.1) | |
| Undetermined | 437 (4.1) | 213 (48.7) | 224 (51.3) | |
| Tumor location*** | <0.0001 | |||
| Proximal colon | 5877 (55.4) | 2952 (50.2) | 2925 (47.8) | |
| Transverse colon | 2062 (19.4) | 883 (42.8) | 1179 (57.2) | |
| Distal colon | 2679 (25.2) | 1205 (45.0) | 1474 (55.0) | |
| Tumor histology** | 0.014 | |||
| Mucinous adenocarcinoma | 1023 (9.6) | 485 (47.4) | 538 (52.6) | |
| Other adenocarcinoma | 9309 (87.7) | 4446 (47.8) | 4863 (52.2) | |
| Non-adenocarcinoma | 279 (2.6) | 107 (38.4) | 172 (61.6) | |
| Hospital Characteristics | ||||
| Mean annual surgeon caseload*** | <0.0001 | |||
| < 10 | 4607 (43.4) | 1866 (40.5) | 2741 (59.5) | |
| 10 – 19 | 3086 (29.1) | 1524 (49.4) | 1562 (50.6) | |
| 20 – 29 | 1264 (11.9) | 726 (57.4) | 538 (42.6) | |
| 30+ | 507 (4.8) | 323 (63.7) | 184 (36.3) | |
| Unknown/Missing | 1154 (10.9) | 601 (52.1) | 553 (47.9) | |
| Hospital size (no. beds)*** | <0.0001 | |||
| 1 – 199 | 2494 (23.5) | 970 (38.9) | 1524 (61.1) | |
| 200 – 349 | 3204 (30.2) | 1520 (47.4) | 1684 (52.6) | |
| 350 – 499 | 2397 (22.6) | 1190 (49.6) | 1207 (50.4) | |
| 500+ | 2523 (23.8) | 1360 (53.9) | 1163 (46.1) | |
| Teaching hospital*** | <0.0001 | |||
| Yes | 5095 (48.0) | 2554 (50.1) | 2541 (49.9) | |
| No | 4801 (45.2) | 2104 (43.8) | 2697 (56.2) | |
| Unknown | 722 (6.8) | 382 (52.9) | 340 (47.1) | |
| Geographic Characteristics | ||||
| County-level poverty rate*** | <0.0001 | |||
| < 10% | 2154 (20.3) | 1072 (49.8) | 1082 (50.2) | |
| 10 – 19% | 6789 (63.9) | 3300 (48.6) | 3489 (51.4) | |
| ≥ 20% | 1675 (15.8) | 668 (39.9) | 1007 (60.1) | |
| Urban/Rural*** | <0.0001 | |||
| Big Metro | 5450 (51.3) | 2726 (50.0) | 2724 (50.0) | |
| Metro | 3078 (29.0) | 1541 (50.1) | 1537 (49.9) | |
| Urban | 680 (6.4) | 256 (37.6) | 424 (62.4) | |
| Less Urban | 1148 (10.8) | 409 (35.6) | 739 (64.4) | |
| Rural | 262 (2.5) | 108 (41.2) | 154 (58.8) | |
| SEER Registry*** | <0.0001 | |||
| Atlanta | 267 (2.5) | 137 (51.3) | 130 (48.7) | |
| Connecticut | 634 (6.0) | 381 (60.1) | 253 (39.9) | |
| Detroit | 611 (5.8) | 260 (42.6) | 351 (57.4) | |
| Greater California | 1740 (16.4) | 925 (53.2) | 815 (46.8) | |
| Greater Georgia | 904 (8.5) | 447 (49.4) | 457 (50.6) | |
| Rural Georgia | 23 (0.2) | 12 (52.2) | 11 (47.8) | |
| Hawaii | 110 (1.0) | 43 (39.1) | 67 (60.9) | |
| Iowa | 897 (8.5) | 337 (37.6) | 560 (62.4) | |
| Kentucky | 938 (8.8) | 320 (34.1) | 618 (65.9) | |
| Los Angeles | 695 (6.5) | 338 (48.6) | 357 (51.4) | |
| Louisiana | 755 (7.1) | 303 (40.1) | 452 (60.9) | |
| New Jersey | 1589 (15.0) | 759 (47.8) | 830 (52.2) | |
| New Mexico | 225 (2.1) | 119 (52.9) | 106 (47.1) | |
| San Francisco | 361 (3.4) | 232 (64.3) | 129 (35.7) | |
| San Jose | 210 (2.0) | 103 (49.0) | 107 (51.0) | |
| Seattle | 482 (4.5) | 210 (43.6) | 272 (56.4) | |
| Utah | 177 (1.7) | 114 (64.4) | 63 (35.6) |
Column percentages presented
Row percentages presented
P ≤ 0.05
P ≤ 0.01
P ≤ 0.001
Among the 93.7% of patients for whom surgeon specialty was available, 98.6% had a single surgeon. Patients primarily saw general surgeons (72.6%), colorectal surgeons (22.1%) or surgical oncologists (2.0%).
Laparoscopic surgical use
Overall, laparoscopic surgery was used for 47% of all resections, including 484 converted surgeries. All covariates differed significantly by surgery type. Laparoscopic surgery rates were similar among men (49.3%) and women (46.2%). Among Whites, 47.8% of resections were laparoscopic, with 42.7% among Blacks and 49.8% among Other races. Laparoscopic resection rates were highest for patients with only 1 comorbidity (49.4%). Among Medicaid patients, 40.8% of resections were performed laparoscopically, compared to 48.9% among non-Medicaid patients. As age increased, laparoscopic surgery rates decreased, from 49.6% among patients aged 66–74 to 43.4% among patients aged 85+. As tumor stage increased, laparoscopic-resection rates decreased. As the number of hospital beds increased, so did the rate of laparoscopic surgery, from 38.9% in hospitals with fewer than 200 beds to 53.9% in hospitals with 500+ beds. Patients living in counties with >20% poverty rates had the lowest rates of laparoscopic resection (39.9%), as did patients in less urban areas (35.6%). Rates also differed by SEER area.
Adjusted Model
When adjusted, patients were 25% less likely to receive laparoscopic surgery if they were aged 85+ (vs. 66–74) and 41% more likely if they did not receive Medicaid (vs. Medicaid) (Table 2). Males were 14% more likely than females to receive laparoscopic resection. Race and number of comorbidities did not predict laparoscopic resection. Patients with proximal (vs. distal) tumors were 36% more likely to receive laparoscopic surgery, and patients with more advanced disease (stages II and III-IV each vs. stage 0-I) and more aggressive tumors (grade 3 and 4 vs. grade 1) were significantly less likely to undergo laparoscopic surgery. As hospital size increased, so did the odds of receiving laparoscopic surgery. Hospital teaching status was non-significant. Patients from counties with more than 20% of the population living in poverty (vs. <10%) were 35% less likely to receive laparoscopic surgery.
Table 2.
Adjusted odds of receiving non-urgent laparoscopic surgery for colon cancer
| Included Variables | aOR (95% CI)a |
|---|---|
| Individual-level Characteristics | |
| Sex | |
| Male | 1.14 (1.04 – 1.26) |
| Female | Ref |
| Race | |
| White | Ref |
| African American | 0.90 (0.75 – 1.07) |
| Other | 1.21 (0.98 – 1.48) |
| Comorbidity (Elixhauser) | |
| 0 | Ref |
| 1 | 1.05 (0.92 – 1.19) |
| 2+ | 0.91 (0.80 – 1.02) |
| Medicaid (dual eligibility) | |
| Yes | Ref |
| No | 1.41 (1.23 – 1.61) |
| Age group | |
| 66 – 74 | Ref |
| 75 – 84 | 0.92 (0.82 – 1.02) |
| 85+ | 0.75 (0.66 – 0.86) |
| AJCC tumor stage | |
| 0 – I | Ref |
| II | 0.66 (0.58 – 0.74) |
| III – IV | 0.58 (0.51 – 0.65) |
| Tumor grade | |
| 1 – Undifferentiated | Ref |
| 2 – Poorly differentiated | 0.88 (0.73 – 1.05) |
| 3 – Moderately differentiated | 0.70 (0.57 – 0.86) |
| 4 – Well differentiated | 0.58 (0.42 – 0.81) |
| Tumor location | |
| Proximal | 1.36 (1.21 – 1.53) |
| Transverse | 0.92 (0.80 – 1.06) |
| Distal | Ref |
| Tumor histology | |
| Mucinous adenocarcinoma | Ref |
| Other-adenocarcinoma | 1.01 (0.87 – 1.18) |
| Non-adenocarcinoma | 0.80 (0.57 – 1.14) |
| Hospital Characteristics | |
| Mean annual surgeon caseload | |
| 1 – 9 | 0.40 (0.31 – 0.52) |
| 10 – 19 | 0.66 (0.51 – 0.86) |
| 20 – 29 | 0.75 (0.56 – 0.99) |
| 30+ | Ref |
| Hospital size (no. beds) | |
| 1 – 199 | Ref |
| 200 – 249 | 1.47 (1.12 – 1.91) |
| 350 – 499 | 1.56 (1.14 – 2.14) |
| 500+ | 1.87 (1.30 – 2.69) |
| Teaching hospital | |
| Yes | Ref |
| No | 0.97 (0.77 – 1.21) |
| Geographic Characteristics | |
| % of population living in poverty | |
| 0 – 9 | Ref |
| 10 – 19 | 0.92 (0.74 – 1.15) |
| 20+ | 0.65 (0.50 – 0.86) |
aOR: adjusted odds ratio; CI: confidence interval
Multi-Level Variation
Due to missing data, 226 patients were excluded from cross-classified modeling (n=10,392). In unadjusted models, significant variation in laparoscopic surgery utilization existed at hospital (MOR = 3.31, V = 1.58, p < 0.001) and county (MOR = 1.32, V = 0.09, p < .05) levels (Table 3). When adjusted, hospital variation (aMOR = 3.31, V = 1.58, p < 0.001) remained larger than county-level variation (aMOR = 1.28, V = 0.07, p < .05).
Table 3.
Unadjusted and adjusteda hospital and county variability in laparoscopic surgery utilization in SEER-Medicare colon cancer patients
| Unadjusted | Adjusted | |||
|---|---|---|---|---|
|
|
||||
| Hospital | County | Hospital | County | |
|
|
||||
| Variance | 1.58 (P < 0.001) | 0.09 (P < 0.05) | 1.58 (P < 0.001) | 0.07 (P < 0.05) |
| MORb | 3.31 | 1.32 | 3.31 | 1.28 |
Adjusted for sex, race, comorbidity, Medicaid, age, AJCC tumor stage, tumor grade, tumor location, tumor histology, mean annual surgeon caseload, hospital size, hospital teaching status, and county-level percentage of the population living in poverty
MOR: Median odds ratio
P-values based on 1-tailed tests
Among the 85 hospitals with at least 30 incident colon cancer patients, the observed percentage who received laparoscopic surgery ranged from 4.4 to 94.0 (Table 4) while the predicted rate ranged from 8.1 to 91.3. Six hospitals had observed percentage of laparoscopic surgery use below the lower bound of the 1-tailed 95% credible interval of the predicted percentage (laparoscopic surgery use was lower than expected), while eight were above the upper bound (use was higher than expected).
Table 4.
Observed and predicted percentage of patients who received laparoscopic surgery for colon cancer, by hospitals and counties with 30+ patients
| Min | Mean | Median | Max | Number of units with observed % below 95% CI† of predicted |
Number of units with observed % above 95% CI† of predicted |
||
|---|---|---|---|---|---|---|---|
| Hospital (n=85) | 6 | 8 | |||||
| Observed | 4.4 | 52.0 | 53.1 | 94.0 | |||
| Predicted | 8.1 | 51.8 | 53.0 | 91.3 | |||
| County (n=76) | 4 | 11 | |||||
| Observed | 0.0 | 51.1 | 51.3 | 82.5 | |||
| Predicted | 10.4 | 50.4 | 52.3 | 76.7 |
CI: credible interval, 1-tailed
Among the 76 counties with at least 30 new colon cancer patients, the observed percentage who received laparoscopic surgery ranged from 0.0 to 82.5 while the predicted rate ranged from 10.4 to 76.7. Four counties had observed percentage of laparoscopic surgery use below the lower bound of the 95% credible interval of the predicted percentage (use was lower than expected), while 11 were above the upper bound (use was higher than expected).
Discussion
We examined the role of patient, hospital and geographic variability in non-urgent/non-emergent laparoscopic resection for colon cancer. Use varied across hospitals, but less so across counties. When comparing observed and predicted values among hospitals and counties with at least 30 patients, patients in 8.9% of hospitals and 13.1% of counties received laparoscopic surgery below expected rates. At the patient level, sex, dual Medicare/Medicaid enrollment, older age, and tumor characteristics (stage, grade, proximal location) were significant predictors of laparoscopic surgery. Annual surgeon caseload, hospital size, and high county poverty were also significant.
Study limitations include those typical of observational studies, including an inability to account for factors not reflected in the database, such as patient preferences and other factors which may impact utilization of laparoscopic approaches. Although we used the most recent release of the SEER-Medicare data set, there is an inherent time lag in using large linked data. As utilization continues to increase, more recent data may show higher proportions of colon cancer resections being performed laparoscopically. Additionally, this study lacks generalizability to patients without Medicare, including younger patients or those with HMO coverage. We also had to exclude the Alaska Native SEER registry due to low cell counts. SEER includes a nationally representative sample of cancer patients, but not hospitals. Colon cancer disproportionately affects older people,34 so this study applies to the majority of colon cancer patients. HMO market penetration has been associated with increased likelihood of laparoscopic surgery use as well as lower overall costs,14 so our exclusion of patients with HMO coverage may underestimate laparoscopic use. However, Medicare Parts A and B provide complete claims data and therefore eliminate potential confounding due to different reimbursement plans. We were unable to examine the influence of specific comorbidities (e.g., obesity) that are difficult to identify using claims data and have been associated with increased odds of conversion to open surgery.35,36 Surgeon caseload was included as a hospital-level variable rather than being nested within hospital due to small numbers of patients for many providers and model complexity.
Our study extends knowledge by examining county-level geographic variation and by utilizing a Bayesian cross-classified multilevel model to quantify the variation in laparoscopic surgery use separately across hospitals and counties. Previously, geographic variation was examined at the census region21 and hospital referral region26 in the US. Our model accounted for little of the county variation in our study, indicating that additional county-level factors are at work. A hot-spot analysis in Ontario, Canada found geographic variability at the neighborhood level (defined as 3-character postal forwarding sortation areas) associated with presence of minimally invasive surgery fellowship training facilities; 14% of the population resided in cold spots, which is similar to our finding of underutilization at the county level.37 The ability to find geographic variation is related to the number of units examined. Smaller geographic divisions, such as census tracts, would have provided greater variability and may have resulted in different findings, but we were unable to utilize census tracts due to many tracts having fewer than five patients. Other studies found greater variability in colon cancer complications and 30-day mortality at the census-tract versus hospital level,27,28 but the current study found greater variation at the hospital versus county level, possibly due to the larger geographic unit (county) of measurement.
A previous study found surgeon caseload was positively associated with laparoscopic surgery.22 We also found that as surgeon caseload decreased, so did the odds of receiving laparoscopic resection. Less than 5% of patients in our study had surgery performed by surgeons with annual caseloads of 30+ (based on the study population) whereas 43% had surgeons with annual caseloads of less than 10. These patients were 60% less likely to undergo laparoscopic surgery compared to patients seeing surgeons with the highest caseload.
Our study supports prior research indicating that laparoscopic surgery is influenced by patient- and hospital-level factors.3,5,8,17–21 However, our study goes beyond previous studies to show that our proxy for individual-level low SES and county-level poverty (<10% vs. ≥20%) were significantly associated with laparoscopic surgery. One possible explanation for these differences in utilization associated with poverty may be differential access to providers.38 Even at high volume hospitals, socioeconomic disparities in utilization of minimally invasive colorectal surgery persist after controlling for factors such as comorbidities.39
Larger hospital size was also significant in the adjusted model. Because we excluded urgent/emergent surgeries, we did not use critical access designation in the current study. Only 2.7% (n=281) of included patients had surgeries performed at critical access hospitals. Rather, we used total number of beds as an indicator of differences in hospital size. With over 23% of surgeries performed at hospitals with fewer than 200 beds, improving access to laparoscopic colon cancer resection could have significant cost savings and improvements in local healthcare quality. A previous study found near-universal access to laparoscopic equipment at hospitals filing claims with Medicare,26 so hospital-level variation is not likely due to equipment access. Although larger hospitals were more likely to perform laparoscopic surgery to treat their colon cancer patients, none of the variables included in our model accounted for the variation across hospital, as indicated by the lack of change between the unadjusted and adjusted models.
More research is needed to identify modifiable factors with potential to correct this underutilization. Studies have suggested decision aids may help reduce regional variation in surgery type,26 and practice guidelines have advocated for surgeon training, particularly at small, rural hospitals.32 Additionally, reimbursement through CMS value-based care models could incentivize use of laparoscopic surgery for colon cancer, which may reduce variability across hospitals. A possible solution to unwarranted variation (variation unexplained by disease, patient preference, or evidence-based care) in access to healthcare is to increase shared patient-physician decision making.40 Patient education about initial screening referrals, selecting a treating physician, hospital quality, surgical options, and deciding where (which hospital) to have non-urgent/non-emergent colon cancer surgery are likely more important than the county of residence, although county variation was also significant. For example, patients should be educated about the potential health and cost benefits of laparoscopic resection and given referrals to surgeons who perform it on a regular basis or to hospitals with higher laparoscopic resection rates so patients can make informed treatment decisions based on their preferences and cost considerations. However, certain patients, particularly poor and/or rural patients enrolled in Medicare/Medicaid, may have fewer options about where to have surgery due to their financial and/or transportation constraints or to hospitals’ policies for treating patients based on their health insurance status.
In summary, significant variability exists across hospitals, counties, and patients regarding the use of laparoscopic surgery for non-urgent/non-emergent colon cancer surgery; however, there is only limited county-level variability. Determining the sources of hospital-level variation and decreasing disparities in utilization at the patient level, particularly patients on Medicaid, are areas in which increased laparoscopic-surgery utilization could maximize patient outcomes and reduce healthcare costs.
Supplementary Material
Supplemental Digital Content 2. Text that details the exclusion criteria for urgent, emergent, and emergency procedures pdf
Supplemental Digital Content 3. Table which lists the ICD-9 procedure codes used to identify surgery type pdf
Supplemental Digital Content 4. Table, which lists the surgeon CPT codes used to identify surgery type pdf
Acknowledgments
We thank the Alvin J. Siteman Cancer Center at Barnes-Jewish Hospital and Washington University School of Medicine in St. Louis, Missouri, for the use of the Health Behavior, Communication, and Outreach Core and gratefully acknowledge James Struthers at Washington University School of Medicine for his data management and programming services. The authors acknowledge the efforts of the Applied Research Program, NCI; the Office of Research, Development and Information, CMS; Information Management Services (IMS), Inc.; and the Surveillance, Epidemiology, and End Results (SEER) Program tumor registries in the creation of the SEER-Medicare database.
Source of Funding: This work was supported in part by grants from the National Institutes of Health National Cancer Institute (NCI): P30 CA091842 (Eberlein), K07 CA178331 (Lian), R21 CA169807 (Lian), R56 AG049503 (Schootman) and R01 CA137750 (Schootman). Dr. Davidson was supported in part through grants HL-38180, DK-56260, and Digestive Disease Research Core Center DK-52574. Dr. Eberth was supported in part by a Mentored Research Scholar Grant (MRSG-15-148-01-CPHPS) from the American Cancer Society. Dr. Jeffe was supported in part by the NCI Cancer Center Support Grant to the Siteman Cancer Center (P30 CA091842).
Footnotes
Publisher's Disclaimer: Disclaimers: None
Conflicts of Interest: No disclosures were reported
Presentations: Podium presentation at the Conference on Geospatial Approaches to Cancer Control and Population Sciences, NIH Campus, Bethesda, MD (September 12–14, 2016)
Author Contributions:
Study concept and design: All authors
Acquisition of data: Schootman
Statistical analysis: Lian
Drafting of the manuscript: Ratnapradipa, Schootman
Critical revision of the manuscript for important intellectual content: All authors
Final approval of version submitted for publication: All authors
Obtained funding: Schootman
Administrative, technical, or material support: All authors
Study supervision: Schootman
References
- 1.Schwenk W, Haase O, Neudecker J, Muller JM. Short term benefits for laparoscopic colorectal resection. Cochrane Database Syst Rev. 2005:Cd003145. doi: 10.1002/14651858.CD003145.pub2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Delaney CP, Chang E, Senagore AJ, Broder M. Clinical outcomes and resource utilization associated with laparoscopic and open colectomy using a large national database. Ann Surg. 2008;247:819–824. doi: 10.1097/SLA.0b013e31816d950e. [DOI] [PubMed] [Google Scholar]
- 3.Steele SR, Brown TA, Rush RM, Martin MJ. Laparoscopic vs open colectomy for colon cancer: results from a large nationwide population-based analysis. J Gastrointest Surg. 2008;12:583–591. doi: 10.1007/s11605-007-0286-9. [DOI] [PubMed] [Google Scholar]
- 4.Varela JE, Asolati M, Huerta S, Anthony T. Outcomes of laparoscopic and open colectomy at academic centers. Am J Surg. 2008;196:403–406. doi: 10.1016/j.amjsurg.2007.11.030. [DOI] [PubMed] [Google Scholar]
- 5.Vaid S, Tucker J, Bell T, Grim R, Ahuja V. Cost analysis of laparoscopic versus open colectomy in patients with colon cancer: results from a large nationwide population database. Am Surg. 2012;78:635–641. [PubMed] [Google Scholar]
- 6.Zheng Z, Jemal A, Lin CC, Hu CY, Chang GJ. Comparative effectiveness of laparoscopy vs open colectomy among nonmetastatic colon cancer patients: an analysis using the National Cancer Data Base. J Natl Cancer Inst. 2015;107:dju491. doi: 10.1093/jnci/dju491. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Hansen DG, Fox JP, Gross CP, Bruun JS. Hospital readmissions and emergency department visits following laparoscopic and open colon resection for cancer. Dis Colon Rectum. 2013;56:1053–1061. doi: 10.1097/DCR.0b013e318293eabc. [DOI] [PubMed] [Google Scholar]
- 8.Bilimoria KY, Bentrem DJ, Nelson H, et al. Use and outcomes of laparoscopic-assisted colectomy for cancer in the United States. Arch Surg. 2008;143:832–839. doi: 10.1001/archsurg.143.9.832. discussion 839-840. [DOI] [PubMed] [Google Scholar]
- 9.Bonjer HJ, Hop WC, Nelson H, et al. Laparoscopically assisted vs open colectomy for colon cancer: a meta-analysis. Arch Surg. 2007;142:298–303. doi: 10.1001/archsurg.142.3.298. [DOI] [PubMed] [Google Scholar]
- 10.Buunen M, Veldkamp R, Hop WC, et al. Survival after laparoscopic surgery versus open surgery for colon cancer: long-term outcome of a randomised clinical trial. Lancet Oncol. 2009;10:44–52. doi: 10.1016/S1470-2045(08)70310-3. [DOI] [PubMed] [Google Scholar]
- 11.Fleshman J, Sargent DJ, Green E, et al. Laparoscopic colectomy for cancer is not inferior to open surgery based on 5-year data from the COST Study Group trial. Ann Surg. 2007;246:655–662. doi: 10.1097/SLA.0b013e318155a762. discussion 662-654. [DOI] [PubMed] [Google Scholar]
- 12.Jayne DG, Thorpe HC, Copeland J, Quirke P, Brown JM, Guillou PJ. Five-year follow-up of the Medical Research Council CLASICC trial of laparoscopically assisted versus open surgery for colorectal cancer. Br J Surg. 2010;97:1638–1645. doi: 10.1002/bjs.7160. [DOI] [PubMed] [Google Scholar]
- 13.Kuhry E, Schwenk W, Gaupset R, Romild U, Bonjer J. Long-term outcome of laparoscopic surgery for colorectal cancer: A cochrane systematic review of randomised controlled trials. Cancer Treat Rev. 2008;34:498–504. doi: 10.1016/j.ctrv.2008.03.011. [DOI] [PubMed] [Google Scholar]
- 14.Dor A, Koroukian S, Xu F, Stulberg J, Delaney C, Cooper G. Pricing of surgeries for colon cancer: patient severity and market factors. Cancer. 2012;118:5741–5748. doi: 10.1002/cncr.27573. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Jensen CC, Prasad LM, Abcarian H. Cost-effectiveness of laparoscopic vs open resection for colon and rectal cancer. Dis Colon Rectum. 2012;55:1017–1023. doi: 10.1097/DCR.0b013e3182656898. [DOI] [PubMed] [Google Scholar]
- 16.Zheng Z, Hanna N, Onukwugha E, Bikov KA, Mullins CD. Hospital center effect for laparoscopic colectomy among elderly stage I-III colon cancer patients. Ann Surg. 2014;259:924–929. doi: 10.1097/SLA.0b013e31829d0468. [DOI] [PubMed] [Google Scholar]
- 17.Dik VK, Aarts MJ, Van Grevenstein WM, et al. Association between socioeconomic status, surgical treatment and mortality in patients with colorectal cancer. Br J Surg. 2014;101:1173–1182. doi: 10.1002/bjs.9555. [DOI] [PubMed] [Google Scholar]
- 18.Fox J, Gross CP, Longo W, Reddy V. Laparoscopic colectomy for the treatment of cancer has been widely adopted in the United States. Dis Colon Rectum. 2012;55:501–508. doi: 10.1097/DCR.0b013e318249ce5a. [DOI] [PubMed] [Google Scholar]
- 19.Kemp JA, Finlayson SR. Nationwide trends in laparoscopic colectomy from 2000 to 2004. Surg Endosc. 2008;22:1181–1187. doi: 10.1007/s00464-007-9732-8. [DOI] [PubMed] [Google Scholar]
- 20.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;254:281–288. doi: 10.1097/SLA.0b013e3182251aa3. [DOI] [PubMed] [Google Scholar]
- 21.Simorov A, Shaligram A, Shostrom V, Boilesen E, Thompson J, Oleynikov D. Laparoscopic colon resection trends in utilization and rate of conversion to open procedure: a national database review of academic medical centers. Ann Surg. 2012;256:462–468. doi: 10.1097/SLA.0b013e3182657ec5. [DOI] [PubMed] [Google Scholar]
- 22.Damle RN, Macomber CW, Flahive JM, et al. Surgeon volume and elective resection for colon cancer: an analysis of outcomes and use of laparoscopy. J Am College Surg. 2014;218:1223–1230. doi: 10.1016/j.jamcollsurg.2014.01.057. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.National Cancer Institute. [Accessed January 3, 2017];Overview of the SEER Program, Surveillance, Epidemiology, and End Results Program: Turning Cancer Data into Discovery. Available at: https://seer.cancer.gov/about/overview.html.
- 24.Pruitt SL, Davidson NO, Gupta S, Yan Y, Schootman M. Missed opportunities: racial and neighborhood socioeconomic disparities in emergency colorectal cancer diagnosis and surgery. BMC cancer. 2014;14:927. doi: 10.1186/1471-2407-14-927. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Nguyen NT, Nguyen B, Shih A, Smith B, Hohmann S. Use of laparoscopy in general surgical operations at academic centers. Surg Obes Relat Dis. 2013;9:15–20. doi: 10.1016/j.soard.2012.07.002. [DOI] [PubMed] [Google Scholar]
- 26.Reames BN, Sheetz KH, Waits SA, Dimick JB, Regenbogen SE. Geographic variation in use of laparoscopic colectomy for colon cancer. J Clin Oncol. 2014;32:3667–3672. doi: 10.1200/JCO.2014.57.1588. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Schootman M, Lian M, Pruitt SL, et al. Hospital and geographic variability in thirty-day all-cause mortality following colorectal cancer surgery. Health Serv Res. 2014;49:1145–1164. doi: 10.1111/1475-6773.12171a. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Schootman M, Lian M, Pruitt SL, et al. Hospital and geographic variability in two colorectal cancer surgery outcomes: complications and mortality after complications. Ann Surg Oncol. 2014;21:2659–2666. doi: 10.1245/s10434-013-3472-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care. 1998;36:8–27. doi: 10.1097/00005650-199801000-00004. [DOI] [PubMed] [Google Scholar]
- 30.Edge S, Byrd DR, Compton CC, Fritz AG, Greene FL, Trotti A. AJCC Cancer Staging Manual. 7. New York: Springer-Verlag New York; 2010. [Google Scholar]
- 31.Snijders TAB, Bosker RJ. Multilevel analysis. An introduction to basic and advanced multilevel modeling. London, England: Sage Publications; 1999. [Google Scholar]
- 32.Moghadamyeghaneh Z, Carmichael JC, Mills S, Pigazzi A, Nguyen NT, Stamos MJ. Variations in Laparoscopic Colectomy Utilization in the United States. Dis Colon Rectum. 2015;58:950–956. doi: 10.1097/DCR.0000000000000448. [DOI] [PubMed] [Google Scholar]
- 33.Lian M. Statistical significance of geographic heterogeneity measures in spatial epidemiologic studies. Open Journal of Statistics. 2015;5:5. doi: 10.4236/ojs.2015.51006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.National Cancer Institute. [Accessed January 3, 2017];SEER Cancer Statistics Factsheets: Colon and Rectum Cancer. 2015 Available at https://seer.cancer.gov/statfacts/html/colorect.html.
- 35.Bhama AR, Charlton ME, Schmitt MB, Cromwell JW, Byrn JC. Factors associated with conversion from laparoscopic to open colectomy using the National Surgical Quality Improvement Program (NSQIP) database. Colorectal Dis. 2015;17:257–264. doi: 10.1111/codi.12800. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Masoomi H, Moghadamyeghaneh Z, Mills S, Carmichael JC, Pigazzi A, Stamos MJ. Risk factors for conversion of laparoscopic colorectal surgery to open surgery: does conversion worsen outcome? World J Surg. 2015;39:1240–1247. doi: 10.1007/s00268-015-2958-z. [DOI] [PubMed] [Google Scholar]
- 37.Doumouras AG, Saleh F, Eskicioglu C, Amin N, Cadeddu M, Hong D. Neighborhood variation in the utilization of laparoscopy for the treatment of colon cancer. Dis Colon Rectum. 2016;59:781–788. doi: 10.1097/DCR.0000000000000600. [DOI] [PubMed] [Google Scholar]
- 38.Weissman JS, Moy E, Campbell EG, et al. Limits to the safety net: teaching hospital faculty report on their patients' access to care. Health affairs (Project Hope) 2003;22:156–166. doi: 10.1377/hlthaff.22.6.156. [DOI] [PubMed] [Google Scholar]
- 39.Robinson CN, Balentine CJ, Sansgiry S, Berger DH. Disparities in the use of minimally invasive surgery for colorectal disease. J Gastrointest Surg. 2012;16:897–903. doi: 10.1007/s11605-012-1844-3. discussion 903-894. [DOI] [PubMed] [Google Scholar]
- 40.Wennberg JE. Practice variations and health care reform: connecting the dots. Health affairs (Project Hope) 2004:140–144. doi: 10.1377/hlthaff.var.140. Suppl Variation:Var. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplemental Digital Content 2. Text that details the exclusion criteria for urgent, emergent, and emergency procedures pdf
Supplemental Digital Content 3. Table which lists the ICD-9 procedure codes used to identify surgery type pdf
Supplemental Digital Content 4. Table, which lists the surgeon CPT codes used to identify surgery type pdf