Design and Sample
This study was a secondary analysis of data from an American Cancer Society-funded longitudinal study that explored family adaptation to stress of HSCT [10, 21]. Study participants included 178 individuals who underwent either an allogenic or autologous transplant. Eligibility criteria of participants were those who: 1) received a HSCT for the first time; and 2) were diagnosed with hematologic or lymphoid cancer such as leukemia, lymphoma, germ cell tumor, or multiple myeloma. The study was conducted at a National Cancer Institute-designated Comprehensive Cancer Center. All 178 HSCT patients were used for this analysis with data collected at six-time points: baseline, 1-2 week before hospitalization for HSCT (T1), during hospitalization for HSCT (T2), 1 month (T3), 4 months (T4), 8 months (T5), and 12 months (T6) following the initial hospital discharge for HSCT.
Data collection and Measures
A detailed procedure of data collection was reported in the previously published studies [10, 21]. Briefly, first-time recipients for allogeneic or autologous transplant and their families that were interested in the original study to examine family adaptation and stress of HSCT were referred from their healthcare provider to participate. HSCT recipients were provided information regarding the study in-person or over the phone as well as provided all study details in writing. Survey data from HSCT recipients consenting to participate in the study were used in this analysis.
Self-report questionnaires were completed by each participant at six-time points. The instruments included demographic questionnaires, family cohesion, and HSCT symptom checklist including symptom frequency and symptom bother. Demographic questionnaires included recipients’ age, race/ethnicity, gender, education, marital status, employment status, diagnosis, and date and type of HSCT. Family cohesion was measured by the Moos Family environment sub-scale of cohesion. This is a 9-item subscale ranging 0-9, with higher score indicating greater cohesion. Cronbach’s alpha of family cohesion is 0.78 [9]. Symptom frequency and Symptom Bother were measured by the HSCT symptom checklist which includes 16 prevalent cancer symptoms ranging from physical symptoms to social/psychologic symptoms [5]. Physical symptoms include nausea, vomiting, mouth pain, difficulty swallowing, diarrhea, decreased appetite, fever, headache, nosebleed, difficulty sleeping, pain, hair loss; psychosocial symptoms include boredom, tiredness, isolation, loneliness [5]. Participants were asked to indicate on a five-point scale how frequently they experience each of these symptoms (i.e., symptom frequency) and how much those symptoms bother their everyday living (i.e., symptom bother). The sum of items scores for each subscale (symptom frequency and symptom bother) ranges from 16 to 80, with higher scores indicating more frequent symptom and greater bother. Cronbach’s alphas of symptom frequency and symptom bother are 0.81 and 0.80, respectively.
Statistical Analysis
Descriptive statistics were computed for sample characteristics and longitudinal variables (symptom distress-frequency and bother, and family cohesion). Longitudinal parallel-process (LPP) modeling was used for data analysis [22, 23]. The LPP is a type of growth curve modeling that combines multilevel modeling (MLM) and structural equation modeling (SEM). This two-step approach allows examination of the relationships between two (or more) different longitudinal processes at the same time. This provides insight into how family cohesion and HSCT associated symptoms (symptom frequency and symptom bother) change over time; and how these longitudinal changes relate to each other.
In the first step of the LPP modeling, the initial status at T1 (intercept) and the change over time (slope) of each variable (family cohesion, symptom frequency, and symptom bother) were estimated separately by using MLM with SAS Proc Mixed. In this step, we treated HSCT recipients’ age and type of HSCT (allogeneic vs autologous) as covariates because much of the literature agreed on age as well as HSCT type as strong predictors of HSCT-associated symptom distress [4, 7, 24-25]. SEM was then used in the second step to examine the effects of the trajectory (characterized by the intercept and slope) of family cohesion on the trajectories of both symptom frequency and bother using IBM SPSS AMOS. The SEM fit was evaluated using the model-fit indices of Chi-square of the estimated model (χ2), root mean square error of approximation (RMSEA), goodness of fit index (GFI), normed fit index (NFI), incremental fit index (IFI), relative fit index (RFI), and comparative fit index (CFI). A non-significant Chi-square value (p > .05) suggests a good overall model fit to the data. GFI>0.90, NFI > 0.90, IFI>0.90, RFI>0.90, CFI>0.90, and RMSEA<0.08 indicate an adequate fit [26]. The MLM approach in the first step automatically handled missing data issues. Missing values were also found to be missing at random, thus having no negative effect on parameter estimation [27].