SEARCH

SEARCH BY CITATION

Keywords:

  • disparities;
  • cancer;
  • Medicaid;
  • uninsured

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES

BACKGROUND:

Early research demonstrated that patients' length-of-stay and inpatient costs varied according to their health insurance status. The authors of the current report studied a population-based sample of privately insured, Medicaid-insured, and uninsured inpatients ages 21 to 64 years who underwent surgical resection for either nonsmall cell lung cancer (NSCLC) (n = 781) or colorectal cancer (CRC) (n = 8190) or who underwent mastectomy (n = 6201) to compare length of stay and inpatient costs by insurance status.

METHODS:

Data for this study were derived from all civilian, acute-care hospitals in Virginia from 1999 to 2005. Hierarchical generalized linear models were used to estimate the relation between the explanatory variables and lengths of stay and costs. All analyses controlled for patient characteristics and hospital random effects.

RESULTS:

Medicaid-insured patients with NSCLC had longer lengths of stay (39% or 2.64 days longer) and higher inpatient costs (20% or $2479 higher costs). Uninsured and Medicaid-insured patients with CRC had longer lengths of stay and higher inpatient costs. In contrast, uninsured patients with breast cancer had 11% shorter lengths of stay and 12% lower inpatient costs than privately insured patients. Medicaid-insured patients had 10% lower inpatient costs than privately insured patients. Differences were no longer statistically significant when reconstruction was added to the models.

CONCLUSIONS:

Health insurance affected the need for health care and the amount of health care received. Uninsured and Medicaid-insured patients with lung cancer and colon cancer who underwent resection had longer lengths of stay and higher inpatient costs than privately insured patients, but they had shorter lengths of stay when reconstruction was not provided. Among the patients with breast cancer, patients and/or providers economized on discretionary procedures. Cancer 2012. © 2012 American Cancer Society.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES

As the United States grapples with medical care cost containment and simultaneously seeks ways to provide care for its estimated 52 million uninsured,1 health care costs continue to rise, and disparities in health outcomes are widespread. Studies have tested the hypothesis that, in the absence of a third-party payer, hospitals and/or patients economize on unneeded procedures. Most of those studies have demonstrated that inpatient charges and lengths of stay are shortest for the uninsured followed by the Medicaid-insured relative to charges and lengths of stay for privately insured patients.2-5 Nationwide, hospital stays are shorter for uninsured patients.6 Such evidence is sometimes interpreted as support for the belief that hospitals offer services in response to financial incentives rather than services based solely on patient need.7 Inpatient mortality often is greater for uninsured and Medicaid-insured patients relative to their privately insured counterparts,4, 5, 8 suggesting that these patients may be more severely ill and/or may receive less than optimal care.

The negative relation between uninsurance and Medicaid insurance and inpatient receipt of healthcare has been demonstrated across several populations, including newborns,2 children with asthma,9 adult patients with acute myocardial infarction,8 and patients with human immunodeficiency virus/acquired immunodeficiency syndrome.3, 10, 11 In a review of the literature, Hadley12 observed that the uninsured receive fewer preventive and diagnostic services, less therapeutic care, and tend to be more severely ill than their insured counterparts. The uninsured also are less likely to receive medical care when they develop symptoms.13, 14

In the current study, we assessed differences in inpatient use and length of stay between privately insured, uninsured, and Medicaid-insured young adult patients (ages 21 to 64) who were hospitalized for surgical treatment of nonsmall cell lung cancer (NSCLC), colorectal cancer (CRC), or breast cancer. Cancer is a prevalent and expensive disease to treat where insurance-related disparities are known to exist.15-19 Resection generally is not considered elective and is associated with expensive inpatient procedures. However, in patients with breast cancer, reconstruction after resection may be considered discretionary. Therefore, our selection of surgical procedures included nondiscretionary and discretionary services. The roles of individual characteristics, disease severity, and hospital characteristics were considered in determining how uninsured and publicly insured, low-income patients were treated relative to patients for whom reimbursements were more lucrative.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES

Data Source

Inpatient treatment information was extracted from the Virginia Health Information (VHI) discharge database, which contained discharge abstracts on all Virginia civilian hospital admissions that exceeded 23 hours. Abstracts included patient information, International Classification of Diseases (ICD) version 9 diagnosis and procedure codes, payer information, dates of admission and discharge, and charge information. To identify NSCLC, we linked the Virginia Cancer Registry and VHI data using deterministic and probabilistic matching techniques. The following histology codes according to ICD for Oncology, third edition (ICD-O-3) were used to identify patients with NSCLC: 8010, 8012, 8013, 8020, 8046, 8050-8052, 8140, 8260, 8310, 8430, 8480, 8481, 8490, 8560, 8240, 8246, 8249, 8070-8078, 8250-8255, 8570-8575. Both data sets contained Social Security number (SSN), date of birth, sex, and zip code. Among the matched records for patients with NSCLC, 87% matched exactly on SSN, date of birth, and sex. Patients with breast cancer and CRC were identified when a diagnostic code for cancer was associated with a cancer surgical code in the discharge claims data. Patients with breast cancer were identified using ICD Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis codes 174.0 through 174.9 and V10.3 and ICD-9-CM procedure codes 85.41 through 85.48. Patients with CRC were identified using ICD-9-CM diagnosis codes 153.0 through 154.3, 154.8, V10.05, and V10.06 and ICD-9-CM procedure codes 45.4XX, 45.5XX, 45.7XX, 45.8XX, 48.4X, 48.5XX, and 48.6XX.

The VHI and American Hospital Association survey supplied industry information on hospitals, including tax status, teaching status, staffed beds, ownership, and inpatient and outpatient operating expenses. It has been demonstrated that these characteristics influence charges and length of stay for hospitalized patients.2, 9-11

Study Sample

Excluding patients with unknown or missing race information (n = 992) or unknown sex or male breast cancer (n = 5), there were 17,732 patients with cancer ages 21 to 64 years who were diagnosed between January 1, 1999 and December 31, 2005 and who underwent either resection for NSCLC or CRC or total or modified radical mastectomy for breast cancer within 12 months of diagnosis. From this sample, we excluded those who had insurance other than private or military insurance, those with Medicaid coverage, and those who were uninsured (n = 1563), leaving a total sample size of 15,172. Medicare patients were not included because they qualified for Medicare as part of Social Security Disability Insurance and may have had conditions that interfered with surgery and recovery. The remaining sample size was 781 patients with NSCLC, 8190 patients with CRC, and 6201 patients with breast cancer. We chose these 3 cancer sites because they are among the most common cancers, and resection is indicated for nearly all CRC and breast cancer tumors. Furthermore, the resection procedure is somewhat similar for CRC and breast cancer across physicians and hospitals. Breast cancer surgery offers a contrast to NSCLC and CRC resection because it often involves reconstruction, which is not necessary for removal of the cancer.

Variables

Outcomes

The outcomes of interest were length of stay and inpatient costs. To calculate the cost-to-charge ratio for each hospital, we divided each hospital's total inpatient and outpatient operating expenses by its inpatient and outpatient charges. Separate hospital-specific, cost-to-charge ratios were estimated for each year during the study period. We converted charges to costs by multiplying charges by the VHI hospital-specific, cost-to-charge ratio for the year in which the inpatient visit occurred. Costs were adjusted for inflation using the Medical Consumer Price Index for the Washington-Baltimore, Maryland, District of Columbia, Virginia, and West Virginia20 area over the period from 1999 to 2006.

Patient variables

We included race and age in all models. Race was categorized as white, African American, or other. Age at the time of surgery was included as a continuous variable. Sex was included in the NSCLC and CRC models. To estimate patient comorbidity burden, we used the Deyo et al21 adaptation of the Charlson Comorbidity Index,22 which has been used to predict the probability and extent of cancer treatment.23, 24 Comorbidities were counted and classified into 3 groups (0, 1, and ≥2 comorbidities) using all inpatient claims for the patient within the year before surgery. Insurance source was extracted from VHI discharge abstracts. Inpatient mortality was included in all estimations.

We included variables for cancer stage using American Joint Committee on Cancer (AJCC) criteria. Stage was categorized as early (AJCC stage 0 or I), advanced with no metastases (AJCC stage II or III), or distant metastases (AJCC stage IV). Nearly one-third of patients with CRC and breast cancer could not be linked to the registry (36% and 34%, respectively). Stage for patients who could not be linked to the Virginia Cancer Registry was derived using the method developed by Cooper et al.25 This method uses all principal and secondary diagnostic codes during the 3 months before or after diagnosis that indicate regional or distant spread of tumor. Based on the most advanced stage documented, the disease stage is imputed. This method overestimates local stage disease and underestimates regional and distant stage disease. In a sensitivity analysis, we re-estimated our models using only the sample that could be linked to the registry.

According to the procedure described by Diez Roux et al,26 we constructed a summary measure of socioeconomic status for all census zip codes in Virginia (as opposed to restricting the variable to zip codes in which a patient resided) using data from the 2000 US Census. This measure comprised 3 measures of household wealth and income (median household income; value of housing unit; proportion of households with interest, dividend, or rental income); 2 measures of education achievement (proportion of adult residents completing high school and proportion completing college); and the proportion of employed residents with management, professional, and related occupations. The mean and standard deviations were calculated for each of the 6 variables,27 and a z-score was constructed for each zip code by subtracting the mean of all Virginia zip codes and dividing by the standard deviation for each variable. The summary measure was the summation of the 6 z-scores; it ranged from −12.2 to 17.5 and was rescaled to be between 0 and 100, with 0 indicating the lowest socioeconomic zip code and 100 indicating the highest. We entered the score in all estimations as quartiles (<25, 25-49, 50-74, and 75-100). Finally, travel distance was computed from the patient's address to the hospital's address. The driving distance from a patient's zip code center to the hospital's zip code center was calculated using Google Maps. Distance was grouped into 3 categories: ≤20 miles, 21 to 60 miles, and >60 miles.

Earlier research suggested that uninsured and Medicaid-insured patients were less likely to undergo reconstruction after mastectomy, which would lead to shorter lengths of stay and lower inpatient costs.28 Therefore, we included a variable for reconstruction in the breast cancer models. Reconstruction was defined using the following ICD codes: 85.50 to 85.54, 85.7, 85.70 to 85.76, 85.79, 85.8, 85.82 to 85.85, 85.87, 85.89, 85.93, 85.95, 86.60, 86.70, 86.71, 86.72, 86.74, and 86.75. Because some women undergo a reconstruction after the inpatient admission for mastectomy, we searched inpatient claims for a reconstruction claim up to 12 months after the hospitalization during which the patient underwent mastectomy. Ten percent of women underwent reconstruction during a separate hospital admission.

Hospital variables

Hospital variables included ownership (private for-profit, private nonprofit, and government owned), teaching status, and the number of beds (≤100, 101-500, and >500).

Statistical Analyses

Patient characteristics, lengths of stay, and costs were analyzed descriptively by cancer site. Statistically significant differences in the means of continuous, normally distributed variables were tested using t tests. Statistical significance of categorical differences in sex, race, distance, and stage were determined using chi-square tests. The significance of differences in the length of stay and cost was determined using the chi-square test from a generalized gamma distribution to correct for skewness in the distribution.

Hierarchical generalized linear models were used to estimate the relation between the explanatory variables and lengths of stay and costs. To avoid undesirable issues of transformation of estimates, generalized models with a gamma distribution and log link were used. These models estimate the log of expected length of stay and inpatient costs. To account for clustering of patients within hospitals and to estimate the intraclass correlations among patients within each hospital, hospital-level random intercepts were included in the model. This model permits the separation of within-hospital and between-hospital variations after adjustment for patient characteristics. All models were stratified by cancer site.

To ease the interpretation of the nonlinear, regression-based coefficients, we report the marginal effect of health insurance status on length of stay and inpatient cost using the parameter estimates from the regression models.29-31 From each of the stratified models, we evaluated the marginal effect of insurance status on length of stay and cost of resection as the difference between the estimated length of stay or cost assuming that all patients were privately insured, had Medicaid coverage, or were uninsured. We used the bootstrap method to calculate 95% confidence intervals for these estimates and their differences from privately insured patients. One thousand stratified random samples were drawn from the original data with replacement, and each sample was the same size as the original distribution. The stratified models were fit on each of these samples, and the estimated differences were calculated. Nonparametric 95% confidence intervals were constructed using the percentile method for each of the estimates of interest. All analyses were conducted using SAS statistical software (version 9.2; SAS Institute Inc., Cary, NC).32

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES

Descriptive Findings

Table 1 reports patient characteristics by cancer site and insurance status. Patients with private health insurance were more likely to be white; to have earlier stage disease, fewer comorbid conditions, lower inpatient mortality; and to reside in zip codes with a higher socioeconomic score than uninsured or Medicaid-insured patients, although these differences were not uniformly statistically significant in all 3 cancer sites. With the exception of patients with NSCLC, privately insured patients traveled shorter distances to the hospital for surgery. In the CRC sample, the percentage of inpatient mortality was greater among uninsured and Medicaid-insured patients; and, in the breast cancer sample, a lower percentage of uninsured and Medicaid-insured patients underwent reconstruction.

Table 1. Distribution of Patient Characteristics by Cancer Site and Insurance Statusa
 No. of Patients (%)
CharacteristicPrivate InsuranceUninsuredMedicaid
  • Abbreviations: CRC, colorectal cancer; NSCLC, nonsmall cell lung cancer; SD, standard deviation; SES, socioeconomic status.

  • a

    Hospital information was missing for 3 observations. Statistically significant differences compared with privately insured individuals were determined using the chi-square test for categorical variables and t tests for continuous variables.

  • b

    P < .01.

  • c

    P < .05.

  • d

    P < .10.

NSCLC   
 Total557 (100)121 (100)103 (100)
 Men315 (56.55)67 (55.37)44 (42.72)c
 Race bb
  White484 (86.89)91 (75.21)70 (67.96)
  African American66 (11.85)30 (24.79)32 (31.07)
  Other7 (1.26)0 (0)1 (0.97)
 Early stage disease92 (16.52)26 (21.49)c19 (18.45)c
 No. of comorbidities  b
  0170 (30.52)39 (32.23)16 (15.53)
  1250 (44.88)48 (39.67)39 (37.86)
  ≥2137 (24.60)34 (28.10)48 (46.60)
 Distance traveled, miles   
  <20375 (67.32)78 (64.46)77 (74.76)
  21-60138 (24.78)35 (28.93)20 (19.42)
  ≥6035 (6.28)6 (4.96)6 (5.83)
 Inpatient mortality70 (12.57)15 (12.40)16 (15.53)
 Socioeconomic quartile cc
  First quartile: Lowest SES17 (3.05)3 (2.48)6 (5.83)
  Second quartile414 (74.33)105 (86.78)86 (83.50)
  Third quartile117 (21.01)12 (9.92)10 (9.71)
  Fourth quartile: Highest SES9 (1.62)1 (0.83)1 (0.97)
 Age: Mean±SD, y55.9±6.3853.71±6.8855.45±6.19
CRC   
 Total7079 (100)730 (100)381 (100)
 Men3693 (52.17)385 (52.74)153 (40.16)b
 Race bb
  White5483 (77.45)421 (57.67)217 (56.96)
  African American1349 (19.06)259 (35.48)150 (39.37)
  Other247 (3.49)50 (6.85)14 (3.67)
 Early stage disease3968 (56.05)370 (50.68)c179 (46.98)c
 No. of comorbidities bb
  05014 (70.83)472 (64.66)211 (55.38)
  11231 (17.39)128 (17.53)82 (21.52)
  ≥2834 (11.78)130 (17.81)88 (23.10)
 Distance traveled, miles bd
  <204769 (67.49)436 (59.73)251 (66.05)
  21-601784 (25.25)205 (28.08)88 (23.16)
  ≥60410 (5.80)78 (10.68)32 (8.42)
 Inpatient mortality69 (0.97)17 (2.33)b16 (4.20)b
 Socioeconomic quartile bb
  First quartile: Lowest SES456 (6.44)49 (6.71)47 (12.34)
  Second quartile4718 (66.65)591 (80.96)297 (77.95)
  Third quartile1757 (24.82)86 (11.78)36 (9.45)
  Fourth quartile: Highest SES148 (2.09)4 (0.55)1 (0.26)
 Age: Mean±SD, y53.94±7.7152.24±8.9452.18±8.94c
Breast cancer   
 Total5523 (100)300 (100)378 (100)
 Race bb
  White4340 (78.58)169 (56.33)203 (53.70)
  African American959 (17.36)102 (34.00)158 (41.80)
  Other224 (4.06)29 (9.67)17 (4.50)
 Early stage disease4779 (86.53)232(77.33)b302(79.89)b
 No. of comorbidities bb
  04539 (82.18)189 (63.00)234 (61.90)
  1669 (12.11)70 (23.33)80 (21.16)
  ≥2315 (5.70)41 (13.67)64 (16.93)
 Distance traveled, miles bc
  <203650 (66.11)192 (64.00)237 (62.70)
  21-601394 (25.25)64 (21.33)90 (23.81)
  ≥61398 (7.21)42 (14.00)42 (11.11)
 Inpatient mortality2 (0.04)0 (0)0 (0)
 Reconstruction2822 (51.10)55 (18.33)b81 (21.43)b
 Socioeconomic quartile bb
  First quartile: Lowest SES343 (6.21)9 (3.00)35 (9.26)
  Second quartile3389 (61.36)248 (82.67)307 (81.22)
  Third quartile1658 (30.02)41 (13.67)36 (9.52)
  Fourth quartile: Highest SES133 (2.41)2 (0.67)0 (0)
 Age: Mean±SD, y50.55±8.2750.11±9.1249.74±9.16d

Inpatient Lengths of Stay and Costs

Table 2 reports the unadjusted mean lengths of stay and inpatient costs for resections by cancer site and patient insurance status. Privately insured patients with NSCLC had the shortest length of stay (6.57 days) relative to uninsured patients (7.42 days) and Medicaid-insured patients (9.29 days). The difference between privately insured and Medicaid-insured patients was statistically significant (P < .01). Both uninsured and Medicaid-insured patients with CRC had longer lengths of stay than privately insured patients (9.9 days and 13.6 days, respectively, vs 7.9 days; P < .01). Among the patients with breast cancer, the unadjusted lengths of stay were similar between privately insured and uninsured patients, but they were slightly longer for Medicaid-insured patients (P < .05).

Table 2. Unadjusted Inpatient Length of Stay and Costs by Cancer Site and Patient Insurance Source
 Unadjusted Mean ± SDa
VariableNSCLC, N = 781CRC, N = 8190Breast Cancer, N = 6201
  • Abbreviations: CRC, colorectal cancer; d, days; LOS, length of stay; NSCLC, nonsmall cell lung cancer; SD, standard deviation.

  • a

    The statistical significance of mean differences from privately insured determined using the chi-square test from a generalized gamma distribution.

  • b

    P < .05.

  • c

    P < .10.

  • d

    P < .01.

LOS, d   
 Uninsured7.42 ± 5.279.90 ± 7.56b2.14 ± 1.86
 Medicaid9.29 ± 9.05b13.61 ± 14.57b2.51 ± 3.02b
 Private6.57 ± 5.157.90 ± 7.032.28 ± 1.78
Total cost, $   
 Uninsured13,125 ± 10,90919,053 ± 16,865b8093 ± 5474
 Medicaid14,644 ± 14,261c24,319 ± 26,559d7858 ± 5325
 Private12,174 ± 12,10914,487 ± 14,9018184 ± 4752

Because length of stay and costs are closely related, the patterns and statistical significance of inpatient costs by insurance source were similar to those observed for length of stay. The main difference between the 2 sets of estimates was that costs for patients with breast cancer were not statistically different by health insurance status.

Table 3 reports adjusted coefficients. Predictions of length of stay and costs using parameter estimates generated from the regression models also are reported. Despite the control variables, these estimates closely resemble the unadjusted means (Table 2). Medicaid-insured patients with NSCLC had 39% longer lengths of stay than privately insured patients (P < .05), which translates into an adjusted average length of stay for a Medicaid-insured patient of 9.27 days, whereas the average length of stay for a privately insured patient was 6.63 days. Similarly, lengths of stay were 20% and 62% longer for uninsured and Medicaid-insured patients with CRC, respectively, relative to privately insured patients (P < .01) or 9.51 days and 12.87 days versus 7.97 days, respectively. In contrast, in the breast cancer cohort, the length of stay was 11% shorter for uninsured women compared with privately insured women. When we controlled for whether the patient underwent reconstruction, the differences between uninsured and privately insured women was no longer statistically significant. However, in this analysis, Medicaid-insured women had slightly longer lengths of stay than privately insured women (P < .05).

Table 3. Total Inpatient Costs and Length of Stay by Cancer Site: Log-Linear Regressiona
 NSCLC, N = 781CRC, N = 8190Breast Cancer, N = 6201
VariableCoefficient ± SEEstimate (95% CI)bCoefficient ± SEEstimate (95% CI)bCoefficient ± SEEstimate (95% CI)bCoefficient ± SE: Reconstructionc
  • Abbreviations: CI, confidence interval; LOS, length of stay; NSCLC, nonsmall cell lung cancer; SE, standard error; d, days.

  • a

    All estimations include control variables for patient sex, age, race (white, African American, other), number of comorbid conditions (0, 1, ≥2) distance traveled to hospital (<20 miles, 21-60 miles, ≥61 miles), cancer stage (early, late) and socioeconomic score quartile. Estimations also include the following hospital characteristics: ownership (private, for-profit, government, private nonprofit) and teaching status. All estimations include hospital random intercepts. Reference categories are private insurance, white race, ≥2 comorbid conditions, socioeconomic score in the fourth quartile, distance traveled ≥61 miles, in-hospital mortality, and whether the hospital was a teaching hospital or a private nonprofit hospital.

  • b

    Predictive margins are listed with bootstrap 95% CIs.

  • c

    In addition to the variables that were included in all other equations, this model included a variable indicating whether the patient underwent reconstruction.

  • d

    P < .01.

  • e

    P < .05.

  • f

    P < .10.

LOS, d       
 Uninsured0.12 ± 0.087.47 (6.54-8.46)0.18 ± 0.03d9.51 (9.03-10.08)−0.12 ± 0.05e2.05 (1.85-2.25)−0.01 ± 0.05
 Medicaid insured0.33 ± 0.09d9.27 (7.46-11.28)0.48 ± 0.04d12.87 (11.51-14.36)−0.001 ± 0.042.30 (2.09-2.52)0.11 ± 0.04e
 PrivateReferent6.63 (6.18-7.11)Referent7.97 (7.81-8.14)Referent2.30 (2.26-2.35)Referent
Total cost, $       
 Uninsured0.05 ± 0.0912,904 (11,268-14,760)0.18 ± 0.04d17,594 (16,661-18,562)−0.13 ± 0.03d7271 (6778-7763)0.02 ± 0.03
 Medicaid insured0.18 ± 0.10f14,711 (12,499-17,312)0.41 ± 0.05d22,194 (19,768-25,025)−0.11 ± 0.03d7439 (6993-7881)0.004 ± 0.02
 PrivateReferent12,232 (11,342-13,180)Referent14,698 (14,382-15,047)Referent8267 (8140-8399)Referent

Inpatient costs were 20% higher for Medicaid patients with NSCLC relative to privately insured patients with NSCLC (P < .10) or $14,711 versus $12,232, respectively. Inpatient costs also were higher for uninsured (20%) and Medicaid-insured (51%) patients with CRC (P < .01). This translated into costs of $17,594 for uninsured patients and $22,194 for Medicaid-insured patients versus $14,698 for privately insured patients. Inpatient costs for patients with breast cancer were 10% and 12% lower for uninsured and Medicaid-insured women, respectively, relative to privately insured women (P < .05). When a variable for reconstruction was added to the estimations, the difference in inpatient costs was no longer statistically significant. We repeated all analyses in Table 3 using only the patients who had a record linked to the registry and observed that the estimates were unchanged.

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES

The lack of health insurance and Medicaid insurance appears to affect both the amount of health care patients need when treated for cancer and the amount of health care received. In the case of NSCLC and CRC resections, which include few discretionary procedures, uninsured and Medicaid-insured patients have longer lengths of stay and higher inpatient costs relative to privately insured patients. This finding may reflect greater disease severity in low-income patients who may be less likely to adhere to preventive measures (eg, smoking cessation, diet), lifestyle modification (eg, exercise), or screening (eg, mammograms, colonoscopies) that also would lead to improved health. In a closer examination of inpatient costs, we observed that uninsured and Medicaid-insured patients with NSCLC and CRC had longer stays in the intensive care unit, more laboratory tests, and greater pharmaceutical costs relative to privately insured patients (results not shown)—all of which suggests that they were in worse health overall.

In contrast, women with breast cancer were both younger and healthier relative to patients with NSCLC and CRC, and those who were uninsured or Medicaid insured had shorter lengths of stay and lower inpatient costs than privately insured women. Drawing on findings from an earlier study that used the same population and controlled for disease stage and radiation therapy (both of which make reconstruction less likely),28 we observed that, once reconstruction was controlled in the analysis, differences in length of stay and costs were no longer statistically significant for uninsured women, and differences in costs were no longer significant for Medicaid-insured women. Whether these differences reflect a withholding of reconstruction because of payment issues or because uninsured and Medicaid-insured women were more likely to present with advanced stage disease and, thus, were more likely to have contraindications for reconstruction or perhaps are more likely to choose not to undergo reconstruction is unknown.

In a study of inpatient hospital resource use, Canto et al33 observed that Medicaid-insured patients with acute myocardial infarction received fewer reperfusion therapies, underwent fewer invasive cardiac procedures, and had longer hospitalizations than Medicare-insured patients. Although these procedures have a discretionary component, they also may be less likely for patients who are in poor health. Similarly, Hadley et al5 observed consistently shorter lengths of stay among uninsured patients who were hospitalized for high-discretionary/low risk of death procedures, but fewer differences between low-discretionary/high risk of death procedures. Rhee et al,34 in a study of motor vehicle crash patients who were treated at a Level I trauma center, observed no differences in use except in for use of elective long-term care facilities. In contrast, an earlier study35 of over 30,000 hospitalization records indicated that uninsured patients were more likely to receive substandard care and, subsequently, to suffer medical injuries. Hospital characteristics did not account for this variation in care.35 Our study findings are consistent with the notion that hospitals do not economize on nondiscretionary procedures for uninsured and Medicaid-insured cancer patients; and, in fact, hospitals may expend more resources to care for these patients. However, uninsured and Medicaid-insured patients may be less likely to undergo discretionary procedures like breast reconstruction.

Our study has 3 limitations. First, the ability to generalize the results is limited by the focus on a single state. The Virginia Medicaid program currently only covers adults with Medicaid-eligible children who are below 31% of the Federal Poverty Level,36 making it 1 of the more restrictive Medicaid programs in the country. Across-state comparisons are nonetheless limited because each state implements its Medicaid program differently. The second limitation is the absence of information on patient preferences. In the case of breast reconstruction, patients may have chosen to forgo the procedure. Finally, there may be unmeasured differences in health status and behaviors that would make uninsured and Medicaid-insured women poor candidates for reconstruction.

According to a Kaiser Family Foundation report,37 the rate of health care spending has grown faster than inflation and the growth in national income. One hypothesis is that providers, particularly hospitals, may compensate for rising costs by performing more services for financially lucrative patients while economizing on services provided to low-income patients. However, the literature on this subject is inconclusive.5, 9-11 In the case of NSCLC and CRC resections, we observe that hospitals spend more on uninsured and Medicaid-insured patients than they do on privately insured patients. The greater expenses are most likely attributable to differences in health status and health care before hospitalization, which leads to different inpatient experiences and outcomes. A different scenario emerges for patients with breast cancer. In general, these patients are relatively young (on average, aged 50 years) and have fewer comorbid conditions relative to patients with NSCLC and CRC. Furthermore, although reconstruction has many physical and emotional benefits,38 it is considered a discretionary procedure. Our evidence suggests that uninsured and Medicaid-insured women are less likely to undergo reconstruction after mastectomy despite controls for disease stage and comorbid conditions. Taken together, the current findings demonstrate that uninsured and Medicaid-insured patients with cancer who undergo resection have longer lengths of stay and higher inpatient costs than their privately insured counterparts except when there is a discretionary component to the resection. The effect of insurance was consistent across cancer sites despite very different distributions in cancer stage within site.

FUNDING SOURCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES

This research was supported by American Cancer Society grant, RSGI-08-301-01 (an examination of uninsured and insured cancer patients in Virginia; Cathy J. Bradley, principal investigator).

CONFLICT OF INTEREST DISCLOSURES

The authors made no disclosures.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES