|Home | About | Journals | Submit | Contact Us | Français|
To assess patients' experiences with cancer care, ratings of their quality of care, and correlates of these assessments.
For 4,093 patients with lung cancer and 3,685 patients with colorectal cancer in multiple US regions and health care delivery systems, we conducted telephone surveys of patients or their surrogates in English, Spanish, or Chinese at 4 to 7 months after diagnosis. The surveys assessed ratings of the overall quality of cancer care and experiences with three domains of interpersonal care (physician communication, nursing care, and coordination and responsiveness of care).
English-speaking Asian/Pacific Islander patients and Chinese-speaking patients and those in worse health reported significantly worse adjusted experiences with all three domains of interpersonal care, whereas white, black, and Hispanic patients reported generally similar experiences with interpersonal care. The overall quality of cancer care was rated as excellent by 44.4% of patients with lung cancer and 53.0% of patients with colorectal cancer, and these ratings were most strongly correlated with positive experiences with coordination and responsiveness of care (Spearman rank coefficients of 0.49 and 0.42 for lung and colorectal cancer, respectively). After multivariate adjustment, excellent ratings were less common for each cancer among black patients, English-speaking Asian/Pacific Islander patients, Chinese-speaking patients, and patients reporting worse health status (all P ≤ .05).
Patients' reports and ratings of care differed significantly by race, language, and health status. Efforts to improve patients' experiences with cancer care should focus on problems affecting Asian and Pacific Islander patients and those in worse health.
Patient-centered care is an essential component of high-quality health care,1 complementing more technical aspects of care such as the appropriate use of medications and procedures.2 Patients with cancer require clear information about their disease and treatment options, timely attention to symptoms, and effective coordination of care across multiple clinical disciplines. In England, a national survey has been conducted to assess patients' satisfaction with cancer care,3 and in the United States, national surveys of patients' satisfaction with their hospital care have been launched.4 The Institute of Medicine and the President's Cancer Panel have highlighted deficiencies in patients' experiences with cancer care.5,6 However, few studies have analyzed this topic in broadly representative cohorts of US patients with cancer.7
In a prior survey of patients with colorectal cancer in Northern California, problems with coordination of care were strongly associated with worse perceptions of the quality of care.8 African American, Hispanic, and Asian/Pacific Islander patients reported more problems with coordination of care and lower overall ratings of care than white patients. Non–English-speaking patients had particularly low ratings of care, with only half of these patients rating their care as excellent or very good compared with four fifths of English-speaking patients.
To build on this prior study of patients that included one type of cancer in a single region, we assessed patients' perspectives on their quality of care for lung cancer and colorectal cancer in multiple regions and health systems. We were particularly interested in patients' reports of the interpersonal aspects of care, such as their physicians' communication with them and the coordination of their cancer care team. Based on prior research,8 we hypothesized that nonwhite and non–English speaking patients would rate their overall quality of care less highly than white and English-speaking patients, respectively, even after adjusting for socioeconomic factors and health status. We also hypothesized that patients' experiences with coordination of care would be most strongly associated with their overall ratings of care and would account statistically for a substantial proportion of observed differences in ratings of care by race, ethnicity, and language.
To evaluate and improve the care and outcomes of lung and colorectal cancer,9,10 the Cancer Care Outcomes Research and Surveillance (CanCORS) Consortium was established by the National Cancer Institute11 to study these two leading causes of cancer mortality.12 The Consortium includes five geographically defined regions (Alabama, Iowa, Los Angeles County, eight counties of Northern California, and 22 counties of central and eastern North Carolina), five integrated health care delivery systems in the Cancer Research Network,13 and 15 Veterans Health Administration hospitals (Appendix Table A1, online only). By design, only patients with lung cancer were enrolled in Iowa, and only patients with colorectal cancer were enrolled in North Carolina. In 2004, the study sites included nearly 30 million people, representing 10% of the US population. The study protocol was approved by institutional review boards of all participating research groups.
Patients age 21 years and older with newly diagnosed invasive lung or colorectal cancer were enrolled onto the CanCORS study from September 2003 through December 2005, and follow-up surveys were conducted through August 2007. Patients were identified within 3 months of diagnosis from population-based cancer registries at geographic sites and from pathology and cytology records at organizational sites.14 After a treating physician was notified of each patient's eligibility, a letter and brochure were mailed to eligible patients inviting them to participate, followed by at least 10 phone calls to enroll them onto the study. To enhance statistical power for analyses by race and ethnicity, African American, Hispanic, and Asian/Pacific Islander patients were oversampled at several sites. To encourage enrollment of non–English-speaking patients, recruitment materials and patient surveys were translated into Spanish and Chinese. Experienced bilingual interviewers recruited and interviewed patients in these languages (including Mandarin and Cantonese) in California where patients speaking these languages were concentrated. All patients or their surrogates provided informed consent.
The numbers of patients sampled, enrolled, and included in our study cohort are detailed in Figure 1. Response rates among all sampled patients were 49.1% for lung cancer and 53.2% for colorectal cancer based on a standard definition.15 The interviewed patients were representative of the demographic characteristics and cancer stage of all newly diagnosed patients in the study regions.16 The final cohort included 4,093 patients with lung cancer and 3,685 patients with colorectal cancer.
Primary data were collected from patients via computer-assisted telephone interviews at 4 to 7 months after diagnosis. For patients who had died or were too ill to be interviewed, a surrogate (relative or household member) familiar with their cancer care was interviewed. The surveys (available at http://www.cancors.org/public) assessed patients' sociodemographic characteristics (age, marital status, race/ethnicity, language spoken at home, education, and annual income), overall health status (patient survey or surrogate survey for living patients only), comorbid conditions, and experiences and ratings of care.17 Collaborative stage at diagnosis was obtained from medical records or participating cancer registries if medical records were unavailable.18
To assess patients' experiences with the interpersonal aspects of cancer care, 13 questions were adapted from 46 items used in the Consumer Assessment of Healthcare Providers and Systems19 survey and from prior research on patients' experiences with cancer care.8,20 In a principal factor analysis with promax rotation,21 examination of eigenvalues for these 13 questions in our cohort suggested a three-factor solution with clinical face validity (Table 1) that was highly consistent when stratified for patient or surrogate respondents. We grouped items by the factor on which they had the largest standardized regression coefficients (ranging from 0.45 to 0.84) and calculated a summed score for each factor, including physician communication (five items; Cronbach's α coefficient = 0.84 for lung cancer and 0.81 for colorectal cancer), nursing care (two items; α = 0.86 for lung cancer and 0.84 for colorectal cancer), and coordination and responsiveness of care (six items; α = 0.85 for lung cancer and 0.83 for colorectal cancer). We converted the scores in each domain to 100-point scales, with 0 representing the worst possible care and 100 representing optimal care. Patients or their surrogates also rated the overall quality of their health care since cancer was diagnosed on a 5-point scale ranging from excellent to poor (Table 1).
For each cancer, we compared unadjusted proportions of patients or surrogates who rated cancer care as excellent and unadjusted mean scores for each scale of care experiences by patients' demographic and clinical characteristics. We also calculated Spearman rank correlations of experiences in the three domains of care with overall ratings of care.
In our primary analyses for each cancer, we used linear regression models to assess predictors of patients' experiences with physician communication, nursing care, and coordination and responsiveness of care and logistic regression models to assess predictors of excellent overall cancer care (v less than excellent care). These models adjusted for completion of surveys by surrogates for living patients, study site, sociodemographic characteristics (including Spanish and Chinese survey language), cancer stage, six comorbid conditions, and self-reported health status (a 5-point scale from excellent to poor for living patients and a sixth category for deceased patients). In secondary analyses, we assessed whether receipt of surgery, chemotherapy, or radiation therapy was associated with overall ratings of care or experiences in the three domains of patient-centered care; models predicting either excellent or very good care as a broader specification of high quality; and models that included the domains of interpersonal care as predictors of overall ratings of care. We also tested interactions of surrogate responses with race, ethnicity, language, and health status in predicting adjusted experiences and ratings of care, but only a few of several dozen interactions were statistically significant and with no consistent pattern (data not shown).
Analyses were conducted with CanCORS core data (version 1.10) and patient survey data (version 1.8) using SAS software (version 9.1.3; SAS Institute, Cary, NC). Missing data (< 3% of patients on all covariates) were multiply imputed, and adjusted analyses were performed using these imputed data.22 Two-tailed P values or 95% CIs are reported for all results.
Demographic and clinical characteristics of the cohort are listed by type of cancer in Table 2. A substantially larger proportion of surveys were completed by surrogates of deceased patients for those with lung cancer than with colorectal cancer (31.5% v 8.3%, respectively), reflecting the greater proportion of patients with lung cancer who had stage IV disease at diagnosis (41.1% v 18.5%, respectively). For each cancer, one third of black patients were from Alabama, and approximately 90% of Hispanic and Asian/Pacific Islander patients were from the two areas in California (data not shown).
Mean scores for physician communication, nursing care, and coordination and responsiveness of care ranged from 84 to 90 on these 100-point scales. For lung cancer, the mean scores were 85.4 (standard deviation [SD], 20.1), 89.9 (SD, 19.4), and 84.4 (SD, 19.5), respectively, and for colorectal cancer, they were 88.4 (SD, 17.6), 89.5 (SD, 18.9), and 87.9 (SD, 16.7), respectively. In unadjusted analyses, white (mean scores, 85 and 89 for lung and colorectal cancer, respectively) and black patients (means scores, 85 and 89, respectively) reported the most positive experiences with coordination and responsiveness of care, whereas Hispanic (mean scores, 81 and 84, respectively) and English-speaking Asian/Pacific Islander patients (mean scores, 79 and 87, respectively) reported somewhat worse experiences in this domain, and Chinese-speaking patients (mean scores, 67 and 73 respectively) reported much worse experiences. Similar patterns were evident in experiences with physician communication and nursing care (data not shown). Relative to white and black patients, Chinese-speaking patients reported experiences that were approximately 20 points lower for physician communication and 10 points lower for nursing care.
In adjusted analyses of both cancers, the most negative experiences in all domains of interpersonal care were reported for Chinese-speaking patients and patients in fair or poor health or who had died within 4 months of diagnosis (Table 3). Their scores were 4 to 22 points lower than white patients or those in excellent health, respectively. For both cancers, patients in Los Angeles County reported 4- to 7-point lower adjusted scores than patients in the Cancer Research Network, and patients with depression also reported scores that were significantly lower than other patients. In contrast, adjusted experiences with interpersonal care did not differ consistently by age, sex, black race, Hispanic ethnicity, education, income, marital status, or other comorbid conditions. In secondary analyses of patients' experiences with care, patients who received chemotherapy for lung cancer or colorectal cancer reported significantly better adjusted experiences with physician communication, nursing care, and coordination and responsiveness of care (all P < .001) of approximately 4 to 6 points on these 100-point scales.
Excellent overall ratings of cancer care were reported for 44.4% of patients with lung cancer and 53.0% of patients with colorectal cancer (Table 4). Lower proportions of patients with lung and colorectal cancer rated their care as very good (30.7% and 29.6%, respectively), good (15.5% and 12.3%, respectively), fair (5.8% and 3.7%, respectively), or poor (3.6% and 1.6%, respectively). For both cancers, excellent ratings of care were substantially less common among surrogates of living patients, nonwhite patients, Chinese-speaking patients, patients with low incomes or less education, unmarried patients, patients in less than excellent health, patients with stage IV cancers, and patients with a history of depression.
After multivariate adjustment (Table 5), excellent ratings of care remained significantly less common for both cancers among black patients, English- or Chinese-speaking Asian patients, and patients in less than excellent health (all P < .05). For lung cancer, excellent adjusted ratings of care were also significantly less common among patients younger than age 55 years and patients with depression. For colorectal cancer, excellent ratings of care were significantly more common in patients younger than age 65 years, patients in North Carolina, and patients with diabetes, and such ratings were less common among English-speaking Hispanic patients, lower income patients, and patients in Los Angeles County.
In secondary analyses with treatment indicators added to these models, care was rated as excellent more commonly among patients with colorectal cancer who received chemotherapy (adjusted odds ratio [OR], 1.49; 95% CI, 1.20 to 1.85), but no other treatments were significantly associated with ratings of care (data not shown). In secondary adjusted analyses predicting excellent or very good ratings of care, these ratings were less common among black patients (P < .02) and Chinese-speaking Asian patients (P < .001) for both cancers and among patients with annual incomes less than $40,000 for colorectal cancer (P < .05). These ratings were also less common for both cancers for patients in less than very good health (all P < .05).
Patients' experiences in the three domains of physician communication, nursing care, and coordination and responsiveness of care were significantly correlated with ratings of the quality of care for lung cancer (Spearman rank correlation coefficients of 0.41, 0.29, and 0.49, respectively) and for colorectal cancer (Spearman rank correlation coefficients of 0.39, 0.28, and 0.42, respectively; all P < .001). After adjustment for demographic and clinical factors, a 10-point improvement (SD, 0.5 to 0.6) in patients' experiences with coordination and responsiveness of care was associated with a substantially greater likelihood of rating their cancer care as excellent overall for both lung cancer and colorectal cancer (adjusted OR, 1.83 for lung cancer and 1.63 for colorectal cancer; both P < .001). The likelihood of patients rating their cancer care as excellent was significantly but less strongly associated with 10-point improvements in experiences with physician communication (adjusted OR, 1.24 and 1.26 for lung and colorectal cancer, respectively) and nursing care (adjusted OR, 1.12 and 1.18 for lung and colorectal cancer respectively; all P ≤ .001). However, adding these three scales to the adjusted analyses of overall ratings of care accounted for only a small proportion of the significant differences by race, language, and health status (data not shown).
To our knowledge, this study is the largest population-based survey to assess patients' experiences and ratings of cancer care in the United States. Patients' experiences with physician communication, nursing care, and coordination and responsiveness of care were generally good, ranging from 84 to 90 on a 100-point scale. Coordination and responsiveness of care was most strongly associated with patients' adjusted ratings of care.
Distinctive patterns emerged in the associations of patients' sociodemographic characteristics with their ratings and experiences with care. In contrast to our hypothesis, however, experiences with coordination and responsiveness of care did not account for a substantial portion of the differences in ratings of quality of care by race, ethnicity, and language. Relative to white patients, black patients reported similar experiences with physician communication, nursing care, and coordination and responsiveness of care, but black patients were significantly less likely to rate their cancer care as excellent. These seemingly paradoxical findings suggest the need to examine other aspects of cancer care, including interactions with other health care personnel, to understand why black patients rated their care less highly. In adjusted analyses, English-speaking Hispanic patients with colorectal cancer reported somewhat worse experiences with coordination and responsiveness of care, and they were also less likely than white patients to rate their cancer care as excellent. Ratings of quality of care were also significantly worse among lower income patients with colorectal cancer. These findings also merit further exploration to understand why these patients viewed their care less favorably.
Among Asian/Pacific Islander patients, those who spoke English reported moderately worse experiences with interpersonal care for lung cancer and were less likely to rate their care as excellent relative to white patients for both cancers. Those who spoke Chinese reported much worse experiences with all domains of interpersonal care for both cancers, and they were much less likely to rate their care as excellent. These suboptimal experiences and ratings of care among Asian/Pacific Islander patients, particularly Chinese-speaking patients, are consistent with prior surveys assessing Asian patients' views of primary and specialty care.23,24 For patients with limited English proficiency, language-concordant clinicians or well-qualified interpreters promote better interpersonal care and greater satisfaction.25,26 Systems to improve the coordination of care may also reduce racial, ethnic, and socioeconomic disparities in the quality of cancer care.27
We focused on lung cancer and colorectal cancer as the two leading causes of cancer mortality in the United States with different rates of progression. For lung cancer, the inclusion of surrogate respondents was especially important because this cancer causes substantial early morbidity and mortality. Significantly worse ratings of care by surrogates of deceased patients may represent truly worse care or the effect of negative outcomes on the surrogates' perceptions of the quality of care.
For both lung and colorectal cancer, significantly worse adjusted experiences with interpersonal care were reported for patients with worse health status or who had died soon after diagnosis. More focused assessments of care among patients with advanced cancer or poor health status and their families may illuminate ways to improve care for these patients. Significantly higher adjusted scores on all three scales of interpersonal experiences with care were evident among patients who received chemotherapy, suggesting that interpersonal care is better for patients who have more frequent contact with their cancer care providers during the initial months after diagnosis.
Important strengths of our study included the racial and ethnic diversity of our cohort and the use of translated survey instruments and bilingual interviewers to enroll Spanish-speaking and Chinese-speaking patients. In addition, our cohort reflected experiences with care in multiple geographic regions and health care systems, and we interviewed patients or their surrogates within just a few months after diagnosis. Our study also had some limitations. First, we did not directly observe the interpersonal care experienced by patients. Second, because we did not simultaneously interview patients and their surrogates about the patients' experiences, we could not assess whether surrogates' ratings of care were consistent with ratings that patients would have provided. Third, our measures of patients' experiences and ratings of care may have been limited by ceiling effects for approximately one half of our cohort who reported the highest possible responses on these measures.
Our study has important implications for clinical practice and health policy. Patients' experiences with the coordination and responsiveness of cancer care are closely related to their perceptions of the overall quality of care.8,28 Standardized assessments of patients' experiences with their cancer care may promote better patient-centered care as they live with cancer or recover from treatment.29 Stratifying these assessments by patients' race, ethnicity, language, and health status may also help to identify disparities in these domains.30 Future research should assess how the quality of patient-centered care is related to clinical measures such as the completeness of cancer staging and the use of appropriate curative and palliative therapies.
We are grateful to Lin Ding, PhD, for conducting statistical analyses and to Debby Collins for assisting with preparation of the manuscript.
Presented in part at the AcademyHealth 2009 Annual Research Meeting, June 27-July 1, 2009, Chicago, IL.
|Study Site||Geographic Unit or Health Care Organization||Estimated Covered Population (No.)*||No. of Enrolled Patients|
|Lung Cancer||Colorectal Cancer|
|Northern California (8 counties)||Alameda, Contra Costa, Sacramento, San Francisco, San Joaquin, San Mateo, Santa Clara, and Solano||7,741||970||1,007|
|Eastern/central North Carolina† (22 counties)||Alamance, Bertie, Chatham, Durham, Edgecombe, Greene, Guilford, Harnett, Johnston, Lee, Lenoir, Martin, Montgomery, Moore, Nash, Orange, Pamlico, Pitt, Randolph, Wake, Wayne, and Wilson||2,666||—||986|
|Integrated delivery systems in:||1,752||742||688|
|Washington State||Group Health Cooperative|
|Massachusetts||Harvard Pilgrim Health Care|
|Michigan||Henry Ford Health System|
|Hawaii||Kaiser Permanente Hawaii|
|Oregon and Washington State||Kaiser Permanente Northwest|
|Veterans Health Administration (15 hospitals)||Atlanta, GA; Baltimore, MD; Biloxi, MS; Chicago, IL (n = 2); Durham, NC; Houston, TX; Indianapolis, IN; Minneapolis, MN; Nashville, TN; New York, NY (n = 2); Seattle, WA; Temple, TX; and Tucson, AZ||520||525||353|
Abbreviation: CanCORS, Cancer Care Outcomes Research and Surveillance.
The Cancer Care Outcomes Research and Surveillance Consortium was supported by National Cancer Institute Grants No. U01 CA093344 to the Statistical Coordinating Center at the Dana-Farber Cancer Institute; U01 CA093324 to Harvard Medical School and Northern California Cancer Center; U01 CA093332 to the Dana-Farber Cancer Institute and Cancer Research Network; U01 CA093348 to the RAND Corporation and University of California, Los Angeles; U01 CA093329 to the University of Alabama at Birmingham; U01 CA093339 to the University of Iowa; and U01 CA093326 to the University of North Carolina; also supported by Department of Veterans Affairs Grants No. U01 CDA093344 (MOU) and HARQ 03-438MO-03 to the Durham Veterans Affairs Medical Center.
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
The author(s) indicated no potential conflicts of interest.
Conception and design: John Z. Ayanian, Alan M. Zaslavsky, Neeraj K. Arora, Katherine L. Kahn, Jennifer L. Malin, Michelle van Ryn, Julie M. Urmie, Jane C. Weeks, David P. Harrington
Financial support: Neeraj K. Arora
Administrative support: Katherine L. Kahn, David P. Harrington
Provision of study materials or patients: Katherine L. Kahn, Mark C. Hornbrook
Collection and assembly of data: Katherine L. Kahn, Jennifer L. Malin, Mark C. Hornbrook, Julie M. Urmie, Jane C. Weeks, David P. Harrington
Data analysis and interpretation: John Z. Ayanian, Alan M. Zaslavsky, Neeraj K. Arora, Katherine L. Kahn, Jennifer L. Malin, Patricia A. Ganz, Mark C. Hornbrook, Yulei He, Jane C. Weeks
Manuscript writing: John Z. Ayanian, Neeraj K. Arora, Patricia A. Ganz, Michelle van Ryn, Mark C. Hornbrook, Catarina I. Kiefe
Final approval of manuscript: John Z. Ayanian, Alan M. Zaslavsky, Neeraj K. Arora, Katherine L. Kahn, Jennifer L. Malin, Patricia A. Ganz, Michelle van Ryn, Mark C. Hornbrook, Catarina I. Kiefe, Yulei He, Julie M. Urmie, Jane C. Weeks, David P. Harrington