Fax: (213) 740-8905
Cancer treatment adherence among low-income women with breast or gynecologic cancer
A randomized controlled trial of patient navigation
Article first published online: 23 JUN 2009
Copyright © 2009 American Cancer Society
Volume 115, Issue 19, pages 4606–4615, 1 October 2009
How to Cite
Ell, K., Vourlekis, B., Xie, B., Nedjat-Haiem, F. R., Lee, P.-J., Muderspach, L., Russell, C. and Palinkas, L. A. (2009), Cancer treatment adherence among low-income women with breast or gynecologic cancer. Cancer, 115: 4606–4615. doi: 10.1002/cncr.24500
- Issue published online: 17 SEP 2009
- Article first published online: 23 JUN 2009
- Manuscript Accepted: 16 FEB 2009
- Manuscript Revised: 12 FEB 2009
- Manuscript Received: 14 NOV 2008
- National Cancer Institute. Grant Number: RO1 CA94827
- adjuvant cancer treatment;
- patient navigation;
The authors implemented a controlled, randomized trial that compared 2 interventions: the provision of written resource navigation information (enhanced usual care [EUC]) versus written information plus patient navigation (TPN) aimed at improving adjuvant treatment adherence and follow-up among 487 low-income, predominantly Hispanic women with breast cancer or gynecologic cancer.
Women were randomized to receive either TPN or EUC; and chemotherapy, radiation therapy, hormone therapy, and follow-up were assessed over 12 months. Patients with breast cancer were analyzed separately from patients with gynecologic cancer.
Overall adherence rates ranged from 87% to 94%, and there were no significant differences between the TPN group and the EUC group. Among women with breast cancer, 90% of the EUC group and 88% of the TPN group completed chemotherapy (14% of the EUC group and 26% of the TPN group delayed the completion of chemotherapy), 2% of the EUC group and 4% of the TPN group failed to complete chemotherapy, and 8% of the EUC group and 7% of the TPN group refused chemotherapy. Radiation treatment adherence was similar between the groups: Ninety percent of patients completed radiation (40% of the EUC group and 42% of the TPN group delayed the completion of radiation); in both groups, 2% failed to complete radiation, and 8% refused radiation. Among gynecologic patients, 87% of the EUC group and 94% of the TPN group completed chemotherapy (41% of the EUC group and 31% of the TPN group completed it with delays), 7% of the EUC group and 6% of the TPN group failed to complete chemotherapy, 6% of the EUC refused chemotherapy, 87% of the EUC group and 84% of the TPN group completed radiation (51% of the EUC group and 42% of the TPN with delays), 5% of the EUC group and 8% of the TPN group failed to complete radiation, and 8% of the EUC group and 5% of the TPN group refused radiation.
Treatment adherence across randomized groups was notably higher than reported in previous studies, suggesting that active telephone patient navigation or written resource informational materials may facilitate adherence among low-income, predominantly Hispanic women. Adherence also may have be facilitated by federal-state breast and cervical cancer treatment funding. Cancer 2009. © 2009 American Cancer Society.
Despite declines in overall cancer-related mortality,1 disparity in cancer survival among racial/ethnic minorities remains significant.2-6 This disparity is attributable in part to structural constraints and contextual factors that restrict access to healthcare.2, 7 Low-income and minority women are less likely to receive adjuvant treatment for breast and gynecologic cancers8-14 and are more likely to terminate their treatment prematurely and to have higher mortality rates.15-18 However, in contrast to research on interventions to improve screening and abnormal follow-up,19-24 few studies have tested interventions to improve adjuvant treatment adherence among lower socioeconomic populations.25-29
Previous studies have indicated that low-income minorities experienced poorer radiation treatment adherence for cervical cancer compared with national rates (16% vs 63%), had a high rate of treatment interruptions (64%) and discontinuation of treatment for nonmedical reasons (20%),30 were noncompliant with hematologic oral self-administered medication >70% of the time (measured by drug serum levels), and missed appointments >30% of the time.31, 32 An intervention trial on low-income patients with hematologic malignancies demonstrated that control patients were noncompliant 73.1% of the time and fully compliant 21% of the time. After controlling for all other variables, disease severity, compliance with allopurinol, and an educational intervention were associated with significantly better adherence and survival rates.33 Predominantly indigent, minority patients with early breast cancer were less compliant with a standard breast-conservation and radiation therapy program than patients who were reported in clinical trials.34 Suggesting a positive effect on adherence from access to treatment insurance, a Surveillance, Epidemiology, and End Results-Medicare database of 24,510 patients documented high rates of completion (87%) of ≥25 radiation therapy sessions (although with greater noncompletion rates among black women who underwent mastectomy),35 whereas an analysis of 3193 Medicare patients with colon cancer indicated that 78.2% completed the prescribed course of chemotherapy.36
Our controlled, randomized trial compared patient adherence to adjuvant cancer treatment and post-treatment follow-up among 487 low-income, predominantly Latino women with breast or gynecologic cancers who were randomized to receive either a structured patient navigation (TPN) or modestly enhanced usual care (EUC). To better understand the hypothesized role of socioeconomic and culturally mediated attitudes and socioemotional support in adherence, we conducted semistructured qualitative interviews with 29 women from both study groups.
MATERIALS AND METHODS
The Improving Patient Access and Adherence to Cancer Treatment (IMPAACT) trial and the qualitative study were approved by University of Southern California Health Sciences Institutional Review Board. We recruited women with breast or gynecologic cancer to test a culturally tailored patient navigation model that previously was deemed effective in improving abnormal screen follow-up.24 The TPN model combined interactive health education (decisional support), counseling (emotional support), written care site and community service navigation information, and navigator active assistance to facilitate access and adherence to adjuvant treatment. Patients aged ≥18 years were recruited in oncology clinics at an urban public safety-net medical center if they had a primary diagnosis of breast (stage 0-III) or gynecologic International Federation of Gynecology and Obstetrics stage 0 through stage IVB cancer. We excluded palliative care patients.
Of 596 eligible patients, we enrolled 487 individuals (82%) from June 2002 to July 2004, and there were no statistically significant differences between enrolled patients and nonenrolled patients by age, ethnicity, or cancer stage. We randomly assigned patients to an IMPAACT intervention TPN (n = 248) or to EUC (n = 239).
IMPAACT is integrated with oncology care provided under “real-world” service delivery conditions and is designed to enhance standard care through quality improvement enhancements provided by a bilingual, bicultural patient navigator and an MSW team or through the distribution of written materials. There was no “only usual care” group based on previous trials, because studies have indicated that patient navigation improves screening and abnormal screen adherence.19-24
EUC patients received site-standard oncology care, financial department facilitation of the receipt of Medi-Cal treatment funding, and supportive services available to all cancer patients plus an English or Spanish listing of community resources (eg, mental health services, cancer support groups) and of medical center social work, financial, transportation, and childcare resource services; and a patient and family member educational pamphlet on depression and cancer at baseline. These written materials were not provided routinely as part of usual care in this public sector safety-net care system.
TPN is tailored to improve treatment and follow-up access and adherence by influencing predisposing (knowledge/attitudes), reinforcing (social support/cues to action), and enabling (barrier-reduction skill), consistent with the Health Belief Model37 and Socio-Cultural Explanatory Theory.38, 39 TPN provides an initial, structured telephone interview assessing adherence barriers; provides health education, problem-solving, and self-management support; and applies a structured adherence risk algorithm to assign service intensity as follows: Level 1 service (6 month and 12 month follow-up calls), Level 2 (telephone or in-person navigation services), and Level 3 (MSW brief depression or anxiety counseling and/or referral).
Measures: Demographic and Clinical Characteristics
Patient demographics, cancer site, stage, and treatment phase at study enrollment came from medical records or baseline interviews. Among women who reported pain, further assessment was made using the Brief Pain Inventory.40, 41 Patients self-reported functional status using the Karnofsky performance status scale (KPSS), an 11-point rating scale that ranges from normal functioning (10) to death (0).42 The Functional Assessment of Cancer Therapy Scale-General (FACT-G),43, 44 a valid and reliable, 27-item, health-related quality-of-life instrument, was used to assess physical, functional, social/family, and emotional well being. Data on comorbid health conditions were derived from patient self-report. Patient Health Questionnaire 9 was used to assess major depressive disorder,45, 46 and the Brief Symptom Inventory was used to assess anxiety.47 Experienced abstractors collected adherence data from all available charts (n = 444 respondents; 91.2% of study enrollees). Adherence to external beam radiation or intravenous chemotherapy was defined as completed as scheduled (CAsSched), completed but delayed (C-Delay) because of missed treatment appointments, did not complete (DNComplete), or declined, unless the interruption was physician prescribed or resulted from machine breakdown. Information on the receipt of hormones and antidepressant medication was obtained from site pharmacy records. We randomly recruited qualitative study patients from 164 women after the completion of the 6-month interview; 29 women provided written consent (15 women in the TPN group and 14 patients in the EUC group; n = 24 Hispanic women). We tape recorded individual in-person or telephone-guided interviews (11 in English, 18 in Spanish).
Means, standard errors, and percentages were used to describe the general characteristics and distributions of predicting and outcome variables. Logistic regression and polytomous logistic regression models were used to test intervention effects on treatment adherence. We assessed breast and gynecologic patients separately using SAS statistical software (version 9.1; SAS Institute, Cary, NC). Qualitative transcript data were coded and analyzed using standard methodology of “coding consensus, co-occurrence, and comparison.”48-51
Study Enrollment and Attrition
The Consolidated Standards of Reporting Trials (Fig. 1) details study enrollment and attrition over 12 months. Attrition rates did not vary significantly between groups, and the reasons for attrition were similar between groups. Baseline characteristics were compared between patients who had completed outcome measures at 2 follow-up waves (61%) and noncompleters at any follow-up wave (39%). Patients who were lost to follow-up were more likely to be 8.9% US born versus 63% foreign born (P = .019), to have lived <10 years in the United States (68.7% vs 80.6%; P = .003), and to have gynecologic cancer (65.6% vs 42%; P < .001).
Patients were predominantly Latina, Spanish-speaking, foreign born, and unemployed and had Medi-Cal cancer treatment funding or local government medical insurance at enrollment (Table 1). There were no significant differences between the study groups at baseline on key characteristics, including cancer treatment phase (before treatment, active cancer treatment, or follow-up care), age, ethnicity, cancer stage, self-reported KPSS, major depression, or severe pain.
|Demographics/Characteristics||Breast Cancer||Gynecologic Cancer|
|No. (%)||P||No. (%)||P|
|EUC, N=114||TPN, N=123||EUC, N=125||TPN, N=125|
|Age ≥50 y||63 (55)||69 (56)||.90||65 (52)||63 (50)||.80|
|Hispanic||80 (70)||89 (72)||.71||98 (78)||103 (82)||.43|
|English speaking||30 (28)||32 (28)||.94||34 (28)||28 (23)||.40|
|Foreign born||96 (84)||109 (89)||.32||108 (86)||109 (87)||.85|
|Born/in US ≥10 y||67 (72)||84 (82)||.09||73 (69)||70 (65)||.59|
|<High school education||58 (51)||64 (53)||.71||78 (63)||66 (53)||.11|
|Unemployed||95 (83)||89 (72)||.04||101 (81)||111 (89)||.08|
|Unmarried||60 (53)||68 (55)||.68||78 (62)||83 (66)||.51|
|Medi-Cal/Medicare||39 (34)||40 (33)||.95||57 (46)||58 (46)||.89|
|Local government||41 (36)||47 (38)||35 (28)||32 (26)|
|Other||7 (6)||9 (7)||2 (2)||1 (1)|
|None||27 (24)||27 (22)||31 (25)||34 (27)|
|Stage 0, I, II, unstaged||93 (82)||103 (84)||.66||73 (58)||74 (59)||.90|
|Stage III, IV, recurrent||21 (18)||20 (16)||52 (42)||51 (41)|
|Cancer treatment phase|
|Before treatment||50 (44)||68 (55)||.19||27 (22)||23 (18)||.80|
|Active treatment||59 (52)||52 (42)||72 (58)||74 (59)|
|Follow-up||5 (4)||3 (2)||26 (21)||28 (22)|
|Comorbid medical condition||65 (57)||66 (54)||.60||69 (55)||76 (61)||.37|
|Major depression and Dysthymia||21 (18)||21 (17)||.99||8 (6)||8 (6)||.92|
|Major depression, PHQ−9≥10||15 (13)||15 (12)||14 (11)||11 (9)|
|Dysthymia||18 (16)||20 (16)||15 (12)||17 (14)|
|Anxiety, BSI score ≥14||7 (20)||2 (6)||.08||7 (11)||9 (14)||.59|
|Reporting pain||48 (42)||47 (39)||.58||69 (55)||69 (55)||1.00|
|BPI score ≥7||7 (20)||2 (6)||.08||7 (11)||9 (14)||.59|
|KPSS score, mean ± SE||6.8 ± 0.2||7.1 ± 0.2||.27||6.3 ± 0.2||6.7 ± 0.2||.09|
|Functional Well Being, mean ± SE||14.1 ± 0.5||14.3 ± 0.5||.87||14.5 ± 0.5||14.8 ± 0.5||.70|
|Emotional Well Being, mean ± SE||15.4 ± 0.5||14.8 ± 0.5||.44||16.5 ± 0.4||16.7 ± 0.4||.81|
|Physical Well Being, mean ± SE||21.2 ± 0.6||22.7 ± 0.5||.04||19.5 ± 0.5||19.0 ± 0.6||.49|
|Social-Family Well Being, mean ± SE||17.0 ± 0.5||16.9 ± 0.5||.87||19.3 ± 0.6||19.0 ± 0.6||.72|
Survival and Quality-of-Life Outcomes
Logistic regression models with control of baseline scores and cancer stage detected no significant effect on survival or quality-of-life improvement except for a significant improvement in emotional well being in gynecologic patients in the EUC group at the end of 12 months (odds ratio, 2.72; 95% confidence interval, 1.25-5.9 [P = .01]).
Treatment profiles did not vary significantly between the TPN group and the EUC group (Table 2). The majority of patients received chemotherapy. Breast cancer patients received hormone therapy after acute treatment. Overall adjuvant treatment adherence rates were notably high, and there were no significant differences between study groups (Table 3).
|Treatment||No. of Patients (%)|
|Breast Cancer||Gynecologic Cancer|
|Chemotherapy and RT*|
|Received chemotherapy and RT||33 (32)||30 (27)||25 (22)||26 (22)|
|Only chemotherapy||25 (24)||33 (30)||24 (21)||25 (22)|
|Only RT||13 (13)||17 (15)||9 (8)||8 (7)|
|Hormone therapy†||71 (62)||82 (67)||NA||NA|
|Pain medication||89 (78)||84 (68)||103 (82)||108 (86)|
|Antidepressants||25 (22)||28 (23)||15 (12)||11 (9)|
|End of trial|
|Follow-up||58 (56)||61 (55)||55 (48)||58 (50)|
|Continuing chemotherapy/RT||8 (8)||9 (8)||4 (3)||9 (8)|
|Hospice care||1 (1)||1 (1)|
|Deceased||5 (5)||6 (5)||20 (17)||14 (12)|
|Receiving care elsewhere||4 (4)||3 (3)||2 (2)||4 (3)|
|Lost follow-up||28 (27)||31 (28)||33 (29)||30 (26)|
|Treatment Adherence||Breast Cancer||Gynecologic Cancer|
|No. (%)||OR||95% CI||P||No. (%)||OR||95% CI||P|
|NC/Declined||6 (10)||8 (12)||1||7 (13)||3 (6)||1|
|C-Delay||9 (14)||18 (26)||1.50||0.40-5.65||.55||22 (41)||16 (31)||1.70||0.38-7.59||.49|
|CAsSched||48 (76)||42 (62)||0.66||0.21-2.05||.47||25 (46)||33 (63)||3.08||0.72-13.12||.13|
|NC/Declined||5 (10)||5 (10)||1||5 (13)||5 (13)||1|
|C-Delay||20 (40)||21 (42)||1.05||0.26-4.19||.94||20 (51)||16 (42)||0.80||0.2-3.25||.76|
|CAsSched||25 (50)||24 (48)||0.96||0.25-3.74||.95||14 (36)||17 (45)||1.21||0.29-5.06||.79|
|Nonadherent||32 (31)||36 (33)||1||44 (38)||40 (34)||1|
|Adherent||71 (69)||74 (67)||0.93||0.52-1.65||.80||71 (62)||76 (66)||1.18||0.69-2.01||.55|
Among the women with breast cancer, 76% of the EUC group and 62% of the TPN group completed chemotherapy as scheduled (CAsSched), whereas an additional 15% of the EUC group and 26% of the TPN group completed chemotherapy with delays (C-Delay; 10% of the EUC group and 22% of the TPN group had delays attributable to toxicity or other medical reasons [ToxDelay], and 5% of the EUC group and 4% of the TPN group had delays attributable to nonmedical reasons [OthDelay]). Only 1 patient in the EUC group and 3 patients in the TPN group did not complete (DNComplete) chemotherapy, and 5 patients in each study group declined chemotherapy. Adherence to radiation treatment was similar between the breast cancer EUC and TPN groups (50% CAsSched, 40% C-Delay, 2% DNComplete, and 8% declined).
Among gynecologic patients, 46% CAsSched, 28% ToxDelay, 13% OthDelay, 7% DNComplete, and 6% declined chemotherapy in the EUC group and 63% CAsSched, 21% ToxDelay, 10% OthDelay, and 6% DNComplete chemotherapy in the TPN group. Approximately 87% of gynecologic patients in each study group completed radiation treatment, including C-Delay (51% of the EUC group and 42% of the TPN group). Only 2 patients in the EUC group and 3 patients in the TPN group failed to complete radiation treatment, and 3 patients in the EUC group and 2 patients in the TPN group refused radiation treatment.
By controlling for study group using polytomous logistic regression models, gynecologic patients with advanced cancer (stage III, state IV, or recurrent) had significantly higher odds of no adhering to chemotherapy (P < .01). Cancer stage was not associated with radiation adherence for gynecologic patients or with either treatment adherence for breast patients. In addition, there were no significant differences in adherence rates when comparing patients' baseline status (ie, before initiating treatment vs during or after acute treatment) between study groups.
Nearly 33% of the patients with breast cancer and >33% of patients with gynecologic cancer in each study group were not fully adherent to follow-up appointments and either had missed at least 1 scheduled follow-up appointment or had failed to return to the clinic. Site pharmacy records indicated that 44 patients obtained tamoxifen or anastrozole, they had an overall 59% adherence rate, and there was no significant difference in the rate between study groups. Of 27 patients who received antidepressants, adherence rates, defined as obtaining the prescribed number of refills, were significantly better among TPN patients (67%; 10 of 15 patients) compared with EUC patients (25%; 3 of 12 patients; P = .03).
Adherence, Mortality, and Quality-of-Life Outcomes
There were no significant interactions between study group or adherence (CAsSched or not CAsSched) for either mortality or quality-of-life outcomes. Therefore, combined data from TPN and EUC were used to evaluate the associations between timely adherence, the death rate, and quality-of-life FACT-G subscales (Table 4). Gynecologic patients who failed to complete the prescribed chemotherapy regimen as scheduled had a higher death rate at 12 months (29.17% vs 10.34%; P = .01). FACT-G subscale mean scores did not vary between timely and not timely adherence groups in regression models that were controlled by baseline scores and cancer stage. Patient costs-of-care worries were similar between the TPN group and the EUC group. Polytomous logistic regression models were used to compare the odds of adherence to treatment between patients with or without reported cost concerns at either 6 months or 12 months. Adherence was not associated significantly with cost concerns in either breast cancer patients or gynecologic patients.
|Variable||Breast Cancer||Gynecologic Cancer|
|Not CAsSched||CAsSched||P||Not CAsSched||CAsSched||P|
|No. of deaths (%)||4 (9.76)||4 (4.44)||.15||14 (29.17)||6 (10.34)||.01|
|FACT-G scores, mean ± SE|
|Functional||15.56 ± 1.07||17.04 ± 0.64||.24||13.91 ± 1.21||16.58 ± 0.96||.10|
|Emotional||17.85 ± 0.80||18.31 ± 0.49||.63||17.78 ± 1.04||17.58 ± 0.82||.89|
|Physical||21.29 ± 0.88||22.56 ± 0.54||.22||19.76 ± 1.07||22.06 ± 0.84||.11|
|Social-Family||18.96 ± 0.97||18.40 ± 0.59||.62||17.65 ± 1.28||18.70 ± 1.01||.53|
|No. of deaths (%)||2 (3.92)||3 (6.12)||.31||2 (4.35)||4 (12.9)||.13|
|FACT-G scores, mean ± SE|
|Functional||16.12 ± 0.88||17.54 ± 0.86||.26||15.70 ± 1.06||16.38 ± 1.49||.71|
|Emotional||17.96 ± 0.63||18.38 ± 0.61||.64||18.35 ± 0.87||17.11 ± 1.26||.43|
|Physical||22.07 ± 0.78||22.60 ± 0.76||.64||21.08 ± 0.94||22.84 ± 1.32||.29|
|Social-Family||18.45 ± 0.84||18.86 ± 0.82||.73||18.73 ± 1.05||15.59 ± 1.46||.09|
Qualitative Study Results
Information collected from the semistructured interviews provided additional insight into economic, cultural, and systems barriers to adherence. Although all women in this study were receiving care in a public safety-net system, lack of insurance to pay for treatment and related out-of-pocket costs were cited as barriers to care by 21% of the women who were interviewed. In contrast, women also cited strong efforts on their part to be adherent to prescribed treatment and follow-up appointments, and they attributed their desire to adhere to treatment to several factors, including respect for the advice of caring physicians and family members and a strong desire to survive.
The high rate of adjuvant treatment adherence by women in both the TPN group and the EUC group was striking and was in contrast to the previous adherence intervention trials among low-income minority women described above.30-33 Thus, the current study raises key questions regarding the degree of patient navigation assistance that is needed to improve cancer treatment adherence in safety-net healthcare systems. In this study, informational material on community resources and depression and cancer and patient navigation to facilitate the receipt of financial and supportive services may have facilitated treatment adherence.
Social structural factors (eg, lack of health insurance, out-of-pocket treatment, lost wages and transportation costs) also may affect treatment adherence.52 Thus, relatively high adherence rates across study groups may be attributable in part to the study site routine referral to the California Cancer Treatment Fund (under the federal-state Breast and Cervical Cancer Prevention and Treatment Act's funding of poor women) that was implemented within Medi-Cal in January 2002 (before the initiation of recruitment for the current study). The similarity of adherence rates in our trial with the rates from studies of Medicare-insured patients35, 36 suggests that available treatment funding for this population is likely to facilitate treatment participation.
Study limitations included patient attrition, demographic differences and cancer site differences between adherent and nonadherent patients, lack of medical charts for all patients, and the relatively small numbers in subgroup analyses, which raise questions concerning treatment adherence among the patients who were lost to follow-up. Moreover, in a previous longitudinal analysis of this study trial population, socioeconomic stress was associated significantly with depression and poorer quality of life over time.53 However, a brief intervention for patients who met the criteria for major clinical depression (referral for brief counseling followed by referral to a mental health provider) did not result in significant differences in the receipt of antidepressants for patients who reported major depressive symptoms. In contrast, 2 randomized controlled trials of intensive, collaborative care management, including antidepressant medication and/or psychotherapy plus recurrence prevention and maintenance care, significantly improved major depression over 12 months in a similar low-income minority population54, 55 and in a population in Scotland.56
In conclusion, the high adjuvant treatment adherence results from the current study raise the probability that a treatment funding program and routine efforts to ensure that eligible patients receive the funding plus either telephone-delivered navigational information and reminders or written informational materials (in this case, in Spanish when preferred) may facilitate cancer treatment adherence among underserved, low-income Hispanic women in safety-net programs. Further research,57, 58 including an ongoing, multisite study59 on patient navigation community practice models, may provide additional answers to questions of efficacy for specific populations as well as comparative cost data and data on the degree of effectiveness of diverse navigational programs. Evaluation of national and state data on the effects of publicly financed treatment funding also will add to our understanding.
Conflict of Interest Disclosure
Supported by the National Cancer Institute grant RO1 CA94827 (Kathleen Ell, principal investigator).
- 17The impact of socioeconomic status on survival after cancer in the United States. Cancer. 2008; 5: 241-246., , , et al.
- 19Screening for distress can predict loss of follow-up and treatment in cancer patients: results of development and validation of the Distress Inventory for Cancer Version 2 [serial online]. Psychooncology. 2009; 18: 524-533., , , , .
- 40Pain assessment in cancer. In: OsabaD, ed. Effect of Cancer on Quality of Life. Boca Raton, Fla: CRC Press; 1991: 293-305..
- 49The Discovery of Grounded Theory: Strategies for Qualitative Research. Chicago, Ill: Aldine; 1967., .
- 50Basics of Qualitative Research: Techniques and Procedures for Developing Grounded Theory. Thousand Oaks, Calif: Sage Publications; 1998., .
- 51Qualitative Data Analysis: An Expanded Sourcebook. 2nd ed. Thousand Oaks, Calif: Sage Publications; 1994., .