Measurement of the structure/process of care is the first step in improving end-of-life care. The primary aim of this study was to psychometrically validate an instrument for directly measuring the bereaved family's perception of the necessity for improvement in structural/procedural aspects of palliative care. Different sets of questionnaires were sent to 800 and 425 families who lost family members at one of 70 certified palliative care units in Japan in the development and validation phases, respectively, and 281 families of the latter group in the follow-up phase. The participants were requested to fill out a newly-developed Care Evaluation Scale (CES), along with outcome measures (the perceived experience and satisfaction levels) and potential covariates (the degree of expectation, the Center for Epidemiologic Studies Depression Scale, and the Social Desirability Scale). We obtained 485, 310, and 202 responses in the development, validation, and follow-up phases (response rates: 64%, 75%, and 72%, respectively). The 28-item CES had an overall Cronbach's coefficient alpha of 0.98; the intra-class correlation coefficient in the test-retest examination was 0.57. A confirmatory factor analysis revealed 10 subscales: physical care (by physicians, by nurses), psycho-existential care, help with decision-making (for patients, for family), environment, family burden, cost, availability, and coordination/consistency. The CES subscales were only moderately correlated with the perceived-experience and satisfaction levels of corresponding areas (r=0.36-0.52 and 0.39-0.60, respectively). The CES score was not significantly associated with the degree of expectation, the changes of depression, or the Social Desirability Scale. The CES is a useful tool to measure the bereaved family's perception of the necessity for improvement in structural/procedural aspects of palliative care. The advantages of the CES are: 1) it specifically evaluates the structure and process of care, 2) it directly identifies needed improvements, 3) it is not affected by the degree of expectation, depression, or social desirability, and 4) it has satisfactory psychometric properties.