Patients 21 years of age or older with a diagnosis of advanced disease for a solid tumor or non-Hodgkin lymphoma were screened at six cancer centers in Michigan, Indiana, and Ohio during a 20-month period. Inclusion criteria for the study were patients who were newly diagnosed with advanced (i.e., stage III or IV) or recurrent cancer and undergoing chemotherapy. Exclusion criteria were patients who were not cognitively intact, could not read and speak English, and did not have regular access to a telephone. Power analysis demonstrated that two covariates in the model and a sample size of 113 would produce an 85% power for a two-sided partial t test to detect an R-square increase of 0.06 as a result of the addition of the group effect to the linear regression model. The researchers assumed an alpha of 0.05 and that the model with the two covariates alone accounted for at least 20% of the variance in the dependent variable (Cohen, 1988
Recruiters employed by the study approached potential participants and explained the study, and, if interested, patients signed consent forms. Once consent was obtained, a baseline interview was completed to collect sociodemographic, treatment, and symptom-related information. Following baseline data collection, a stratified randomization schema was used to randomly assign patients from each recruitment site to the conventional care group or the conventional care plus a five contact, eight-week, nurse-delivered intervention group. Follow-up interviews for patients in both arms of the study were completed by non-nurses 10 and 20 weeks after entry into the study to collect data regarding symptom severity and potential confounding variables, such as depressive symptoms, age, gender, and site of cancer. The first interview was conducted 10 weeks after recruitment into the study because of the likelihood that most patients would have completed one to two cycles of chemotherapy in that time; the second interview was conducted after 20 weeks because of the likelihood that most patients would have completed one course of chemotherapy. Interviewers were not nurses and were not aware of which arm of the study patients were in. Approval from the institutional review board of each participating site as well as the investigators’ institutions was obtained before the study was implemented.
The symptom management intervention delivered to patients in the experimental group consisted of five contacts during an eight-week period with an RN who was experienced in oncology. The intervention was based on cognitive behavioral theory and was designed to help patients understand the nature of symptoms, improve patients’ belief in their ability to control symptoms, and teach patients problem-solving skills. The first and last contacts occurred in person; the second, third, and fourth contacts were conducted by telephone. The purpose of the first in-person contact was to establish a rapport with patients, and the last contact was meant to facilitate closure to the intervention. Telephone contacts were used at other times to minimize patient burden that might result from participating in the study. Contacts with nurse interventionists were scheduled at two-week intervals to allow patients enough time to implement and assess the effectiveness of symptom management strategies. During each contact, nurses assessed patients’ pain, fatigue, nausea, vomiting, insomnia, dyspnea, weakness, anorexia, fever, dry mouth, constipation, mouth sores, and depressive symptoms. Patients rated the severity of each symptom and its impact on four dimensions of their quality of life: appetite and eating, daily activities, emotions and mood, and sleep. Once patients identified which symptoms were severe or affecting their quality of life, nurse interventionists helped patients reframe their attitudes and beliefs with regard to controlling individual symptoms. Nurses proposed cognitive and behavioral self-care strategies and assisted patients with plans to carry them out.
A customized computer documentation program was used to lead nurses through the patient encounter from symptom assessment to selecting intervention strategies that were incorporated into patients’ plan of care. The computer documentation program was developed by the principal investigators during a previous study and has been used with more than 300 patients with cancer. Any symptom that patients gave a severity score of 5 or higher on a 0-10 scale or 3 or higher on the quality-of-life scale, which was rated 0-5, automatically posted to patients’ plan of care. Nurses and patients reviewed symptoms that reached the thresholds. Patients then selected which symptoms they would focus on during the following two weeks. Together with each patient, nurses tailored a list of interventions, which patients agreed to implement, to decrease the severity or impact of the symptom. Interventions were grouped according to the following domains: prescribe, teach-assess-evaluate, communicate, and counsel. For a patient with pain, for example, a nurse might suggest recording pain levels throughout the day, using distraction, or enhancing communication with physicians and family caregivers regarding current and acceptable pain levels (see ). Although nurses suggested strategies, patients ultimately were responsible for choosing and implementing them.
Examples of Interventions Used to Assist Patients With Symptom Management
Four nurse interventionists were employed for the study, and each attended an initial two-day training session during which the study’s goals, procedures, and objectives were discussed. On completion of the initial training, nurse interventionists performed two mock interviews that were recorded and reviewed by the principal investigators and a quality assurance coordinator for protocol compliance and appropriateness of interventions. After the mock interviews were considered acceptable, interventionists recorded one intervention per month for the duration of the study, which was reviewed by the quality assurance coordinator. In addition, the quality assurance coordinator reviewed the computer record of every intervention session that was completed during the study for protocol compliance, appropriateness of interventions, and completeness of data. Finally, all nurses participated in monthly telephone conference calls during which strategies were reviewed by the group to ensure uniformity in the delivery of the intervention.
To test the impact of the intervention, age, gender, and cancer site and stage were identified during the baseline interview, and symptom severity and depressive affect were assessed at baseline, 10 weeks, and 20 weeks. The site of cancer was grouped according to breast, lung, and other (e.g., colorectal, gastrointestinal-pancreatic, genitourinary-gynecologic). The stage of cancer was identified during the medical record audit at recruitment as stage III or IV according to tumor, node, metastasis guidelines (National Cancer Institute, 2004
Symptom severity was measured during each interview observation by asking patients to rate the severity of each symptom on a scale of 0 (not present) to 10 (as severe as it possibly could be) and then summing severity ratings for each symptom, with higher scores indicating higher levels of severity (possible range = 0-120). The symptoms that were included in the severity index included pain, fatigue, nausea, vomiting, insomnia, dyspnea, weakness, anorexia, fever, dry mouth, constipation, and mouth sores.
Patients’ reports of depressive symptoms were evaluated by the Center for Epidemiologic Studies-Depression (CES-D) scale
). The 20-item CES-D scale assesses a respondent’s level of depressive symptoms on a four-point Likert-type scale (i.e., 1 = almost all of the time, 2 = most of the time, 3 = some of the time, 4 = rarely or none of the time). Scoring for the CES-D scale consists of reverse coding negative items and summing individual items so that higher scores indicate higher levels of depressive symptoms. Reliability analysis of the CES-D scale revealed a Cronbach’s alpha of 0.89.
Univariate analyses were conducted for all continuous variables to study their underlying distribution. Comparisons of continuous variables between groups (i.e., experimental versus control, those lost to attrition versus those who completed the study) were made by t tests or regression models and adjusted for potentially confounding variables. For variables that did not satisfy the t test assumption of equal variances, the p value from the t test was based on the Satterthwaite method for correcting the degrees of freedom. Furthermore, any violation of normality was primarily in terms of skewness; therefore, t tests were employed to compare means for all variables.
To compare categorical variables between the levels of the group variable or attrition variable, the contingency-table Pearson chi-square test for general association was used. When 20% or more of the cells of the contingency table had expected counts less than five, the two-sided Fisher’s exact test for the overall cross-classification table was used. Linear regression models were used to answer the research question.