Dependent Variables
For each quality indicator, cases eligible for the indicator were categorized as having received the specified care or not. Patient refusals of specified care or documentation of a contraindication were also considered to have met criteria for receipt of appropriate care. For bivariate and multivariate analyses, indicators were grouped into 4 domains: treatment toxicity (1 indicator), pain screening and management (6), palliative cancer treatment (4), and hospice (1). For these domain analyses, cases were considered to have met criteria for receipt of recommended supportive care if they received the care specified
for all of the indicators for which they were eligible in the domain.
Independent Variables
Patient characteristics included stage of disease, age, gender, race (white, black, other/unknown), marital status (married/living with partner, other), urban/rural residence, and performance status (poor/not poor), as shown in Table 1. Performance status was abstracted from the medical record as ECOG (Eastern Cooperative Oncology Group) or Karnofsky when present, but most of the time it was recorded in patients’ charts qualitatively (eg, “poor performance status”). Cases were attributed to facilities based on the VACCR attribution, which is usually the VHA facility where treatment is initiated. Facility characteristics included geographic region; number of unique patients; facility complexity level; chemotherapy availability on-site; radiation therapy availability on-site; and types of palliative care services, psychosocial support, and patient tracking and case management available on-site (Table 2).
Statistical Analyses
Bivariate analyses were performed to determine which patient and facility characteristics were associated with quality of care for treatment toxicity, pain, palliative treatment, and hospice. Multilevel logistic regression models were developed for each of these 4 domains to examine the relationship between indicator care receipt and patient and facility characteristics. Collinearity was assessed by examining Pearson correlations between independent variables, and highly collinear variables were not included in multivariate analyses.
Independent variables significantly associated with dependent variables in bivariate analyses or considered necessary for adjustment from conceptual standpoints were included in the multivariate models (cancer type [NSCLC or SCLC], age, sex, race, marital status, urban/rural classification, poor performance status, Veterans Integrated Service Network location, facility complexity level, facility region, as well as the facility having the following cancer-related services: tumor board, palliative care unit, palliative care services, chaplain, method for tracking patients through treatment and posttreatment care (to ensure receipt of timely and appropriate care), and presence of a case manager]. Testing yielded no significant interactions between any of the variables; thus interaction terms were not included in the final regressions. All analyses were performed using SAS statistical software, version 9.1 (Cary, NC).
Results
Patient Characteristics
Table 1 summarizes patient characteristics of this elderly (mean age, 67 years; SD, 9), mostly white, and mostly male veteran cohort. In all, 15% of the cohort was black, nearly one-third lived in a rural area, and 46% were married or living with a partner. Eighteen percent had poor performance status, 38% were referred for palliative care, and 81% died during the study period.
Organizational Characteristics
Patients in the cohort were cared for at one or more of 111 VAMC facilities (Table 2), with the largest proportion (45%) seen at facilities located in the South. Most patients (76%) were seen in low-complexity facilities (54% of facilities) and almost all (95%) used a facility reporting having at least 1 tumor board (76% of facilities) in the 2009 VHA Oncology Services Survey. Most facilities (90%) reported having chemotherapy on-site, with 99% of patients in our cohort receiving care at one of these facilities. Radiation therapy at most sites was provided by referral to a non-VA facility, although 31% of facilities had radiation therapy available on-site and these VAMCs provided care to 40% of the patients. A substantial proportion of facilities offered inpatient (91%) and outpatient (76%) palliative care consultation, serving the majority of patients, 95% and 78%, respectively. Various forms of psychosocial support were reported at facilities serving over half of the study patients. Forty-five percent of veterans were seen at facilities that provided care tracking for lung cancer patients (41% of facilities) and at 14% of facilities (12% of patients) the tracking method was monitored by a nurse practitioner or physician assistant. Sixty-eight percent of patients were cared for at facilities reporting the presence of a case manager (69%). (The fact that a site has chemotherapy or any of the other aforementioned services in-house does not imply that patients receiving care at this site were eligible for or received these particular services, but rather that they were cared for at a facility reporting presence of the service.)
