Annals of Surgical Oncology

, Volume 19, Issue 3, pp 722–727

Race and Health Disparities in Patient Refusal of Surgery for Early-Stage Non-Small Cell Lung Cancer: A SEER Cohort Study

Authors

  • Rohtesh S. Mehta
    • Division of General Internal Medicine, Department of MedicineUniversity of Pittsburgh School of Medicine
  • Diana Lenzner
    • Biostatistics FacilityUniversity of Pittsburgh Cancer Institute
    • Division of Hematology/Oncology, Department of Medicine, and Cancer Therapy & Research CenterUniversity of Texas Health Science Center
Healthcare Policy and Outcomes

DOI: 10.1245/s10434-011-2087-3

Cite this article as:
Mehta, R.S., Lenzner, D. & Argiris, A. Ann Surg Oncol (2012) 19: 722. doi:10.1245/s10434-011-2087-3

Abstract

Introduction

Several factors, including race, age, stage, comorbid conditions, social support, and socioeconomic status, have been linked to the likelihood of a patient having surgery for early-stage non-small cell lung cancer (NSCLC). The aim of the present study is to determine the influence of race and health disparities on refusal of recommended potentially curative surgery.

Methods

The Surveillance, Epidemiology, and End Results (SEER) database was used to create a cohort of 62,514 patients diagnosed with stages I and II NSCLC between 1988 and 2002, of whom 51,938 were recommended for surgery. The outcome variable was refusal of recommended surgical treatment, while race was the key predictor variable. Potential confounders were adjusted for in the hierarchical generalized logistic regression analysis.

Results

A majority was White (86%) and underwent surgery (81%). About 2% of Blacks (n = 109), 1.4% of Whites (n = 756), and 2.8% of “other” race individuals (n = 96) refused surgery. In the multivariable adjusted model, Blacks [odds ratio (OR) 1.95, 95% confidence interval (CI) 1.5, 2.3, P < 0.001] and those of “other” race (OR 2.03, 95% CI 1.5, 2.5, P < 0.001) had greater odds of refusing surgery than did Whites. Increasing age, male gender (OR 1.17, P = 0.031), and being unmarried (OR 2.1, P < 0.001) were other factors associated with higher odds of refusal. Significant county variations were also noted in refusal of surgery.

Conclusions

Blacks and “other” races are more likely to refuse recommended surgery for early-stage NSCLC compared with Whites. Future studies should focus on exploring potential reasons for refusal and developing communication interventions.

Lung cancer is the leading cause of cancer death and a major public health problem.1,2 Moreover, the number of prevalent cases is expected to increase from 222,000 in 2010 to 338,000 in 2030.3 More than half of cases of lung cancer are diagnosed at advanced stage, which is generally incurable.4 However, when diagnosed at early stage, non-small cell lung cancer (NSCLC) is potentially curable with treatment including surgery. Overall, about 86% of all lung cancers are of NSCLC histology.5 The observed 5-year survival rates for early-stage (stages I and II) NSCLC are about 54% with surgery and disappointingly 10% without surgery.4

Racial differences exist in surgical treatment for lung cancer. Blacks are significantly less likely to undergo surgery than are Whites for early-stage NSCLC.613 For lung cancer in general, Blacks are more likely to present with poorer performance status, substantial weight loss, and lower socioeconomic status (SES), and are more likely to be unmarried, disabled, unemployed, and have lower education levels.14,15 All of these factors are known to be associated with less likelihood of receiving adequate treatment.16 However, racial differences in surgical treatment of NSCLC persist even after comorbid conditions and socioeconomic factors are accounted for, and even when patients are eligible for surgery.13,17 One of the reasons for this may be due to possibly higher rates of refusal of recommended surgery by Blacks, as suggested by a few regional studies.17,18 Our study objective is to explore potential racial differences in the refusal rates of recommended surgery using a national database. From 1988 onwards, the Surveillance, Epidemiology, and End Results (SEER), which is the largest source of US cancer statistics, records the reason for not performing surgery, such as if it was contraindicated or not recommended, or if the patient actually refused the recommended treatment. We hypothesized that Whites are more likely to undergo surgery and less likely to refuse it compared with Blacks even when adjusting for potentially confounding factors.

Patients and Methods

Data

This cohort study utilized the SEER data which covers geographic areas representing 28% of the US population and represents about 25% of Whites and 26% of Blacks.19 The SEER-17 November 2009 Submission (1973–2007 varying) database was obtained using the SEER*Stat software.20

Inclusion/Exclusion Criteria

We included all patients who were diagnosed with malignant neoplasm of lung and bronchus with early-stage (stages I and II) NSCLC histology between 1988 and 2002. Of these patients, a cohort of patients who were recommended for surgery was created, which included patients who underwent surgery as well as those who refused recommended surgery (extracted from SEER unique data point coding). NSCLC histology was identified using ICD-O-3 coding (8000–8040, 8046–9989). American Joint Committee on Cancer (AJCC) 3rd edition (1988+) staging stage was used to identify patients with stages I and II. No specific age limit was set for inclusion criteria; however, patients with unknown age were excluded. No other exclusion criteria were used. The specified study time period was chosen to allow uniformity of data coding throughout the study period.

Variables

The primary outcome of interest was receipt of surgery, which was categorized as (a) surgery performed or (b) recommended but refused by patient. The key predictor variable was race, which was categorized as White, Black, and other (including American Indian/Alaskan Native, Asian/Pacific Islander, all other races, and unknown). Although we included the “other” category in the race variable, this was not included in our primary hypothesis because of lack of clinical information gained from a merged race category. Other covariates included (1) age, categorized as <50, 50–64, 65–79, or 80 years and over, (2) gender, (3) diagnostic confirmation, categorized as either positive laboratory diagnosis or radiological/clinical diagnosis only, (4) marital status, categorized as married or not married (single/never married/separated/divorced/widowed), (5) number of primary tumors, dichotomized as single or multiple, (5) year of diagnosis (continuous variable from 1988 to 2002), and (6) proxy measures (county-level data) of SES such as education (percentage of individuals with less than high-school education in a county), percentage of people with English-language difficulties, and if they lived in a metropolitan or a nonmetropolitan county, to estimate roughly the proximity and availability of healthcare. The classification of metropolitan versus nonmetropolitan counties was based on the SEER “Rural–Urban Continuum Code.”21 As education and income are highly correlated, income was not included in the model.

The following variables were considered for entry into the final analysis model because of their prognostic significance (race, age, gender, marital status, year of diagnosis, and the county-level variables of rural–urban continuum, percentage with less than high-school education, and percentage with language barrier). Covariates that were not significant based on P-values <0.05 were excluded from the final model. The variable “laboratory diagnosis” was removed from consideration, since nearly all patients with positive laboratory diagnosis had surgery (99.99%). The final model contained only significant covariates.

Statistical Analysis

To examine crude differences in receipt of surgical treatment among different races, χ2 tests were used. To examine whether the relationship between race and refusal of surgical treatment was significant after adjusting for other covariates, a hierarchical generalized logistic model was created with the PROC GLIMMIX procedure in SAS 9.2.22 A hierarchical model was chosen due to the county-linked (not individual-level) SES data in the SEER database. This was done by treating county as a random effect to account for potential correlation among individuals within the same county. The estimated probability of an individual refusing surgical treatment conditioned on a set of predictor variables was modeled.

Results

The dataset contained information on 62,514 individuals (Table 1). The majority of individuals were White (86%) and had surgery performed (81%). Fewer than 2% (1.5%) of the patients refused recommended surgery, while 17% did not undergo surgery as it was contraindicated or not recommended. There were roughly similar numbers of males and females, a majority (56%) fell into the 65–79 years age group, almost 60% were married, and about 90% lived in a metropolitan county. County-level data were pulled from 469 different counties. On average, about 19% had less than high-school education, 5.8% were categorized as having “language isolation,” and almost 60% of the counties were nonmetropolitan.
Table 1

Individual-level statistics (%)

 

White n = 53,662

Black n = 5,375

Other n = 3,425

Total n = 62,514

85.8%

8.6%

5.5%

Surgery indicatora

 Surgery performed

82.0

74.3

80.3

81.2

 Surgery refused

1.4

2.0

2.8

1.5

 No surgery recommended

16.2

23.3

16.6

16.8

 Unknown

0.4

0.4

0.3

0.4

Gendera

 Male

52.9

59.2

58.8

53.7

 Female

47.1

40.8

41.2

46.2

Age (years)a

 <50

3.7

7.7

5.3

4.2

 50–64

26.9

37.6

28.3

27.9

 65–79

57.2

47.5

54.2

56.2

 ≥80

12.1

7.2

12.1

11.7

Marital statusa

 Not married

36.9

53.4

29.1

37.9

 Married

60.2

42.3

68.5

59.1

 Unknown

2.9

4.3

2.4

3.0

Laboratory diagnosisb

  

 No laboratory diagnosis

1.9

2.1

1.3

1.8

 Laboratory diagnosis

98.1

97.9

98.7

98.1

Number of primariesa

 One

66.2

69.5

74.9

66.9

 Multiple

33.8

30.5

25.1

33.0

Radiotherapya

 Radiation

19.8

24.7

22.2

20.3

 No radiation

78.2

73.6

76.5

77.7

 Radiation refused

0.9

1.00

1.2

1.0

 Unknown

1.1

0.7

0.1

1.0

Rural/urbana

 Metropolitan county

88.4

96.4

93.5

89.3

 Nonmetropolitan county

11.6

3.6

3.7

10.5

 Unknown

0

0

2.8

0.2

% <high-school education

 Mean, SD

18.7, 7.2

21.5, 6.5

18.9, 6.5

18.9, 7.2

Language barrier

 Mean, SD

5.6, 4.6

6.2, 4.9

8.9, 4.4

5.8, 4.7

SD standard deviation

aChi-square P-value <0.0001

bChi-square P-value <0.05

Crude Differences in Refusal of Surgical Treatment among Different Races

There were significant differences in refusal of surgical treatment among Whites (1.4%, n = 756), Blacks (2.0%, n = 109), and those of “other” race (2.8%, n = 96) (P < 0.001). Blacks had 59% higher odds of refusing surgery compared with Whites (OR 1.59, 95% CI 1.64, 2.52, P-value <0.001), while “other” race individuals had roughly twice the odds of refusal compared with Whites (OR 2.03, 95% CI 1.64, 2.52, P-value <0.001).

Adjusted Differences in Refusal of Surgical Treatment Among Different Races

Data from 50,288 individuals who were recommended surgery from 466 different counties were used. Of these individuals, 49,363 had surgery while 925 refused. Covariates that were joint significant predictors of receipt of surgery were age, race, sex, and marital status. Blacks (OR 1.88, 95% CI 1.50, 2.36, P-value <0.001) and individuals of “other” race (OR 1.95, 95% CI 1.50, 2.52, P-value <0.001) had greater odds of refusing surgical treatment than did Whites. As age group increased, individuals had greater odds of refusing surgery. Males had higher odds of refusing treatment than did females (OR 1.17, 95% CI 1.01, 1.35, P-value 0.032). Married individuals were less likely to refuse treatment than individuals who were not married (OR 0.47, 95% CI 0.41, 0.55, P-value <0.001). As the percentage of individuals with language barrier and percentage of individuals with less than high-school education in the county increased, the odds of refusing surgery also increased (Table 2). There was a county effect on refusal of surgery, signifying that refusal of surgery differed by county even after accounting for the different individuals in each county (P < 0.001).
Table 2

Probability of refusing surgical treatment (n = 50,288)

Effect

Crude

Adjusted

Odds ratio

95% CI

P value

Odds ratio

95% CI

P value

Intercept

       

0.873

Race (baseline = White)

 Black

1.59

1.64

2.52

<0.001

1.88

1.50

2.36

<0.001

 Other

2.03

1.64

2.52

<0.001

1.95

1.50

2.52

<0.001

Age, years (baseline = <50)

 50–64

2.14

1.09

4.21

0.028

2.26

1.14

4.46

0.019

 65–79

4.85

2.50

9.39

<0.001

5.13

2.64

9.95

<0.001

 ≥80

21.65

11.14

42.07

<0.001

21.40

10.96

41.79

<0.001

Sex (baseline = female)

 Male

0.95

0.839

1.08

0.468

1.17

1.01

1.35

0.032

Marital status (baseline = not married)

 Married

0.42

0.37

0.48

<0.001

0.47

0.41

0.55

<0.001

% with language barrier

1.03

1.02

1.05

<0.001

0.96

0.92

0.99

0.011

% with less than high-school education

1.03

1.02

1.03

<0.001

1.03

1.01

1.04

0.001

Covariance parameter estimates (n = 50,288, no. of counties = 466)

Intercept = county, estimate = 0.33, standard error = 0

Discussion

It is known that Blacks are less likely to receive surgical treatment for early-stage NSCLC compared with Whites.613,17 We demonstrate that Blacks are also more likely to refuse recommended potentially curative surgery compared with Whites, which supports our hypothesis. Furthermore, patients from the “other” race category were more likely than Whites to refuse surgery. However, given the composite nature of this category, which included American Indians, Pacific Islanders, Asians, and all other minority groups, the clinical inference and public health policy implication of this finding are limited. Further dissection of this fused category may provide different conclusions. This fact is highlighted by the results of a recent study which showed that non-Hispanic Asian/Pacific Islanders and Hispanics have 2 and 1.6 times the odds of getting cancer-directed surgery for lung cancer compared with non-Hispanic Whites, respectively.13

The racial differences in refusal of surgery found in our study are similar to those of a smaller, single-institutional retrospective cohort study. This study involved 281 patients with stage I or II NSCLC and concluded that Blacks had a higher rate of refusal of recommended surgery compared with Whites (18 vs. 5%, P = 0.002).18 In a multivariable adjusted model, age was the only significant predictor for refusing surgery in addition to Black race. This study had much higher overall refusal rates compared with ours, and had significantly higher proportion of Blacks (35%) compared with ours (9%). This could be related to regional variability, as the study was limited to a single geographic location. One prospective cohort study which included patients from five healthcare settings in North and South Carolina suggested that, in addition to being Black, age, comorbid illnesses, and strong belief that faith alone can cure disease were some of the other factors significantly associated with not receiving surgery.17 Of all the patients who did not undergo surgery, about 10% declined it; however, there were no differences in rates of denial between Blacks and Whites (9.6 vs. 10.8% respectively, P-value 0.9). This may be explained by relatively small sample size compared with our study.

Several factors could explain higher refusal rates in Blacks, such as general distrust of the medical system, belief that exposure to air during surgery can spread the cancer, misconceptions about cancer treatment, skepticism about the diagnosis, belief that faith and prayer alone can cure cancer, lack of satisfaction with physician communication, and false perception of the prognosis.2327 On the other hand, although there are racial differences in patient perception of quality of life with progressive lung cancer, these differences do not explain a patient’s decision about surgery.27

Apart from these patient-related factors, a number of physician-related factors also play a crucial role in a patient’s determination about the treatment. Blacks are more likely to report their physician communication as less participatory; however, when these conversations happen with a racially matched physician, these are reported as more participatory.28 Because of negative perceptions of communication and prognosis by patients, lack of empathic responses and emotional support by physicians, and underrecognition of patients’ general distress and psychological disorders including significant depression, less than half of patients recall their discussion about the goals of lung cancer treatment, which is naturally expected to have a negative impact on their decision-making capability.17,2931

Several weaknesses of our study merit attention. First, this study endures biases associated with all retrospective studies such as information bias, including possible misclassification of race and other variables in the database. In addition, as individual-level data were not available for the SES variables, we used proxy measures based on county attributes, which may not apply to all individuals equally. However, hierarchical models take this into consideration and provide more conservative estimates. Also, being based on a large, national database, we expect the results to be representative of the general population. In addition, we did not have information on health insurance or access to healthcare from our database. However, by categorizing patients by metropolitan or nonmetropolitan county residence, we hoped to approximate healthcare access reasonably. We also recognize that our study did not include various comorbidities which could be accessed through the SEER-linked Medicare database. However, this may not impair our interpretations significantly, as our primary interest is to focus on patients who were recommended surgery by their physicians. In general, people who are recommended surgery can be assumed to have similar functional status as well as comorbidity burden. Last but not least, future studies on the topic should include certain physician factors such as age, gender, race/ethnicity, board specialization, and specialty, and certain hospital factors, which again are available through the SEER-linked Medicare database. However, use of the Medicare database will limit the population to those older than 65 years, which is a population with higher odds of refusing surgery compared with younger patients. Another potential bias is the lack of information in the SEER database pertaining to a patient’s spirituality, family support, and psychological stressors, which are all significant factors contributing to a patient’s decision-making process.

Conclusion

We demonstrate that Blacks and “other” races are more likely to refuse recommended surgery for early-stage NSCLC than are Whites. In addition to race, increasing age, male gender, and being unmarried were other factors significantly associated with refusal of surgery. Future studies should consider including certain physician and hospital characteristics in the analysis as well as comorbidity index using the SEER-linked Medicare databases. Nonetheless, important information about the root cause of denying this potentially curative surgery can only be obtained from prospective studies, which also allow intervening at the fundamental level, specifically patient–physician communication. The interventions could be targeted at both the patient and physician level. Involving a multidisciplinary team including the patient’s family physician, social worker, and possibly a spiritual provider, depending upon the situation, may be helpful to address some of the patient-related factors. On the other hand, communication interventions, such as use of audiovisual recordings of patient encounters and subsequent feedback, may mitigate some of the provider-level barriers. Onco-talk, which is the National Cancer Institute (NCI)-funded communication-oriented training program for oncologists, was started in 2002.32 It would be interesting in future studies to assess whether rates of refusal differ before and after initiation of this program.

Acknowledgment

We wish to thank Charity Moore, PhD, MSPH, Kevin Kraemer, MD, and Gordon Wood, MD for their help in preparation and editing of this paper.

Copyright information

© Society of Surgical Oncology 2011