We conducted a cross-sectional survey between April 2018 and September 2019 with patients in two of the largest oncology hospitals in Mexico City, which were selected by convenience sampling—one belonging to the Ministry of Health (MoH) and the other to the Mexican Institute of Social Security (IMSS). Up to 90% of the Mexican population receives health care at either MoH or IMSS facilities. The IMSS health network provides health insurance to formal sector workers and their families, covering 65 million people [24]. The MoH provides health care to 54 million people without social security through local health secretariats located in every Mexican state.
The study population comprised outpatient cancer patients aged ≥ 18 years with one of three HMs: lymphoma, acute leukemia, or multiple myeloma. We included patients with at least one hospitalization during the last year, with five years or less since diagnosis, and without mental impairment. Two fieldwork-trained nurses interviewed patients after their medical consultations if they met the inclusion criteria, agreed to participate, and signed the informed consent forms. Two field coordinators verified the diagnosis and treatment in patients’ health records.
Study variables
We used a patient-centered quality of cancer care (PCQoCC) questionnaire previously validated in Mexico to evaluate experiences with health care [25]. This questionnaire has 30 items in five domains: 1) timely care; 2) clarity of information; 3) information for treatment decision-making; 4) care to address biopsychosocial needs; and 5) respectful and coordinated care. Each item has a 4-point Likert response option (1 = totally agree, 2 = agree, 3 = disagree, 4 = totally disagree with the statement about the experience of receiving specific health care process). The score for each domain was calculated by reversing the response options, adding all subscale items, and dividing them by the number of items in each subscale/domain for a minimum score of one and a maximum of four per domain [25].
The study outcome variable was perceived HRQL of HM patients and was measured using the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Core 30 (EORTC QLQ-C30).26 The EORTC QLQ-30 consists of 30 items grouped into one global health subscale; five function subscales (physical, role, emotional, cognitive, and social functioning); three symptom subscales (fatigue, pain, nausea/vomiting); and six single items covering individual symptoms/problems (shortness of breath, loss of appetite, insomnia, constipation, diarrhea, and financial difficulties). Each item has a 4-point Likert response option scale and two global health questions have a 7-point response option scale. We transformed each subscale linearly to a score of 0–100 with 100 being the best in overall health, functional status, or major symptoms. We used the QLQ-C30 summary score to assess the association between patients’ experience with care and HRQL. This summary score encompasses all function and symptom domains assessed by the QLQ-C30; it is the mean of 13 QLQ-C30 domain scores, with a higher summary score reflecting a better health status. The QLQ-C30 summary score has been recommended as a meaningful and reliable measure for oncological research and it has shown greater prognostic value for overall survival than the global HRQL, physical functioning, or any other scale within the QLQ-C30 [27]. It was previously validated in HM patients [28, 29].
Other study covariates included patient sociodemographic characteristics (gender, age, educational attainment, and marital status) and clinical history (time since diagnosis, cancer type and stage, anxiety, and depression). We categorized the following variables: patient age (≤ 45; 46 to 64; ≥65 years); educational attainment (completed elementary school or less; secondary school; high school or higher); cancer type (leukemia, lymphoma, or multiple myeloma); cancer stage/or risk (early stage [I–II] or low and standard risk; advanced stage [III–IV] or high and very high risk); time since diagnosis (≤ 6 months; 7 to 12 months; ≥1 to 5 years). We measured anxiety and depression using the Hospital Anxiety and Depression Scale composed of 14 items previously validated in Spanish with cancer patients [30]. Each item has a 4-point Likert scale response that ranges from 0–3. A summary score of ≥ 11 points in each domain indicates anxiety or depression.
Sample size and statistical analysis
We secured a minimum of 10 participants per covariate in the multiple regression analysis [31].
We performed descriptive and exploratory analyses and found that the QLQ-C30 summary score had a normal distribution; however, the domains of perceived experiences with patient-centered cancer care had a skewed distribution. Therefore, to determine the association between independent and dependent variables, we dichotomized the dependent variables as high and low PCQoCC by using the 75th percentile of the total sample as a cut-off value. The distribution of variables supported this decision, including the low frequency of patients at the 85th, 90th, and 95th percentiles. We built five high PCQoCC variables: 1) timely care = 4.0 points; 2) clarity of information = 4.0 points; 3) information for treatment decision-making ≥ 3.6 points; 4) care for biopsychosocial needs ≥ 2.08 points; and 5) respectful and coordinated care ≥ 3.83 points.
We performed the Student t-test for two-group comparisons and one-way analysis of variance for the difference in means among more than two groups to compare HRQL by patients’ sociodemographic characteristics, clinical characteristics, and PCQoCC domains.
As recommended by VanderWeele [32], we modeled a multiple regression analysis with simultaneous inclusion of all conceptually and clinically relevant variables to determine the independent association between PCQoCC domains and HRQL independent from other covariates. We performed a bootstrapped linear regression model with 10,000 bootstrap replications. The bootstrap method has a less restrictive assumption about the sample being representative of the population, making it a large sample method akin to the Central Limit Theorem [33, 34]. In addition, we considered the cluster effect, as the study included patients from IMSS and MoH hospitals; thus, one of the assumptions was that the measurements within each hospital may not be independent because patients treated in the same hospital were more likely to receive similar quality of care than patients from other hospitals. For this purpose, we adjusted the standard errors by computing clustered robust standard errors for the coefficients.
Stata 14.0 (Stata Corp, College Station, TX, USA) was used for the analysis; p < 0.05 was considered to be statistically significant.
Ethics considerations. The study was approved by the IMSS National Research and Ethics Committee (registry number R-2017-785-042).