Concordance of Patient and Caregiver Proxy Report of Quality of Colorectal Cancer Care (S746)

Abstract

•Identify the concordance of proxy and patient report of quality of colorectal cancer care.•Describe the limitations of using the kappa statistic for describing agreement in highly skewed observations such as those seen in patient satisfaction results. Patient-reported evaluations of interpersonal quality of care are essential elements of quality of care assessment. During times of serious illness, proxy report of care experiences may be necessary. The validity of this proxy approach, however, depends on the degree to which caregiver and patient assessments are concordant. To examine concordance of patient and caregiver proxy reports of quality of cancer care and to identify any patient or caregiver characteristics that may influence the level of concordance. A VA cohort of colorectal cancer (CRC) patients and their caregivers were surveyed on the quality of VA cancer care in 3 specific domains: surgery, chemotherapy overall and chemotherapy nursing care. Because the kappa statistic penalizes skewed distributions, like commonly seen in patient satisfaction reports, agreement between patients and caregivers on quality of care were measured using Gwet’s AC2 statistic. Stratified analyses on caregiver burden, race, education, and age as well as patient’s stage of disease were used to assess variation in agreement. 417 caregiver-patient dyads completed the survey (70% response rate), of whom 362 (86.8%) had surgery and 195 (46.8%) had chemotherapy. Overall agreement was very high for the 3 quality of care domains: AC2 (95% CI): 0.870 (0.838, 0.903) for surgery, 0.835 (0.786, 0.884) for chemotherapy overall and 0.906 (0.868, 0.944) for chemotherapy nurses. Stratified analyses of agreement did not reveal a consistent pattern of patient or caregiver characteristics resulting in higher or lower levels of agreement. Family caregiver reports on the quality of CRC care were highly concordant with patient reports and not consistently affected by patient or caregiver characteristics.