In this prospective study in diverse ICUs at three hospitals in different parts of the country, we found that performance of evidence-based care processes that providers, patients, and their families endorse as essential components of high-quality ICU palliative care was inconsistent and infrequent. Interdisciplinary family meetings to discuss goals of care in relation to the patient’s condition, prognosis, and preferences were documented for <20% of patients as late as 5 days after admission to the ICU for critical care treatment. Information in printed form was rarely distributed to families, despite evidence from two randomized, controlled, multicentered trials demonstrating the effectiveness of this low-cost approach (14
). Assessment and management of pain were more consistently and frequently performed for patients in this study.
Our cohort included >500 patients receiving intensive care for medical or surgical illness for at least 5 days and >300 of their family members. The study included large urban academic hospitals and smaller community hospitals and ICUs of diverse size, case mix, and care models. At each of these sites, we identified important opportunities for improvement in the quality of palliative care provided to critically ill patients and their families, as measured by the Care and Communication Bundle of core ICU processes (9
). We also found that selected characteristics of patients and of families, and the site where they received care, were associated with variations in care process performance. These findings may help in development of targeted interventions to improve ICU palliative care.
To our knowledge, this is the first prospective multisite study to evaluate performance directly, formally, and systematically on validated measures addressing multiple domains of ICU palliative care quality. We previously described development of this set of measures including pilot and feasibility testing through informal review of medical records by clinical nurses for a small purposive sample of patients from each of a different group of ICUs across the country that had also participated in other ways in developing the measures (9
). Documentation of some of these processes was part of the data collection for a multicentered trial testing an intervention to improve care for patients who died in ICUs in ten hospitals in the state of Washington; those processes were abstracted from medical records along with a long list of other variables being evaluated as potential predictors of the Quality of Dying and Death (16
) rated by bereaved families (17
). In the present study, we focused specifically on performance of all care processes in the Care and Communication Bundle for a large consecutive cohort of all patients in the study ICUs for at least 5 days, regardless of the patient’s vital status at discharge. We used detailed specifications posted by the National Quality Measures Clearinghouse. In addition, we conducted multivariable regression analyses to identify predictors of performance of five of these care processes. Ho et al (20
) compared overall ratings by physicians and nurses of the quality of palliative care in ICUs in a convenience sample of 13 hospitals; these ratings assessed care in the ICU generally rather than care of any specific patient.
In our analyses, the risk of death (as measured by Charlson comorbidity index) emerged as the most consistent predictor of ICU palliative care process performance. This may reflect appreciation on the part of critical care clinicians at these sites of the special importance of providing high-quality palliative care for acutely ill patients with the heaviest burden of chronic illness. It may also be a reflection of clearer and stronger expression of need for this care by the patients or families themselves. Whereas palliative care is certainly appropriate when acute illness is superimposed on multiple chronic comorbid conditions, there is a broader mandate to integrate palliative care for all ICU patients and their families, as expressed by the Institute of Medicine (1
), by all societies representing critical care professionals (2
) and by large national hospital organizations. Our data demonstrating low overall levels of performance of key palliative care processes for patients treated for at least 5 days in ICUs suggest that important opportunities to improve palliative care exist for patients at high risk for hospital death or other unfavorable outcomes.
Our analyses also suggest an association between family race/ethnicity and performance of several ICU palliative care processes, including identification of a medical decision-maker and investigation for an advance directive. These findings are consistent with previous research indicating that Hispanic patients may be less likely to prepare an advance directive (20
). This may have discouraged ICU clinicians from inquiring whether such patients had previously expressed their preferences for intensive care treatments. In addition, clinicians may have been deterred by the challenges of cross-cultural discussions. Because some patients, including patients from Hispanic families, do express their preferences in advance, and because the majority of patients cannot participate personally in communication or decision-making about their care during severe critical illness, it remains important for clinicians to make every effort to determine without delay whether an advance directive exists for each patient and, if so, what it provides. Investigation for an advance directive helps focus attention on the values, goals, and preferences of the patient and can serve to open or continue an ongoing dialogue with the family about appropriate goals of care (21
). Emotional and practical stresses are a universal problem for ICU families across diverse backgrounds (7
). Thus, an offer of social work support, as defined in the Care and Communication Bundle, is a routine component of high-quality ICU palliative care (7
). Open visitation for ICU families has gained increasing support within the critical care community (19
) and is strongly advocated not only by families but also by patients (7
). The association we found between daily family visits and performance of core palliative care processes provides support for liberal approaches to visiting.
The “bundle” strategy used in this study has gained wide acceptance as a valuable method for measuring and improving quality in a variety of ICU practice areas (27
). With this strategy, measures of processes that individually improve care in a particular clinical area are applied together for greater impact (27
). Bundles are used to implement best practices for prophylaxis against catheter-related bloodstream infection (30. 31) and for management of ICU patients with sepsis. Although the Care and Communication Bundle, like other ICU bundles, is composed of care process rather than outcome measures, these care processes address established domains of ICU palliative care quality (6
) and the processes themselves have been specifically endorsed as important by ICU clinicians, patients, and families (7
). By focusing on performance of measures in this Care and Communication Bundle, we sought an approach that would be well-aligned with current efforts to improve the quality of other aspects of ICU care while evaluating the timeliness, reliability, and consistency of ICU palliative care process performance.
In this study, we evaluated performance of care processes but did not implement an intervention to improve performance. Measurement is essential for quality improvement, identifying opportunities, recognizing accountability, and allowing comparison of results of an improvement effort with baseline (or benchmark) performance. However, although measurement may also help to trigger clinical activities that are targets of such an effort, it does not ensure that performance will improve. To improve quality in any area, including palliative care, an intervention including attention to work systems is needed (32
). Improvement also depends on engagement of various disciplines collaboratively to anticipate challenges, develop strategies, strengthen teamwork, and thereby achieve goals of high-quality care. In addition, results of ongoing assessment should be provided to clinicians as performance feedback, together with structured discussion of steps to improve care. The dramatic success of the Michigan Keystone Project (31
) and others to prevent catheter-related bloodstream infections has encouraged efforts by groups of ICUs to improve palliative care using similar methods, including several initiatives based in whole or in part on measures in the Care and Communication Bundle (33
). Preliminary results indicate that these efforts have been effective in improving performance of palliative care processes in the bundle (33
). The IPAL-ICU Project (Improving Palliative Care in the ICU) Project, a Web-based resource sponsored by the National Institute on Aging and the Center to Advance Palliative Care (35
), provides expert guidance, tools, and other materials to promote the success of such interventions.
This study has several limitations. First, practice in the ICUs we sampled in three hospitals may not be representative of that in other ICUs and hospitals across the United States. These ICUs, however, were selected for their diversity in geographic location, hospital size, ICU care model, and case mix, and our patient and family cohorts were large and heterogeneous. Second, the source of data on performance of care processes was medical record (paper charts and electronic records) review rather than direct observation of practice. Although it is possible that processes of care were performed but not documented, medical record documentation provides the most practical and feasible source of data for routine monitoring, feedback, and improvement of ICU care. The Care and Communication Bundle focuses on aspects of practice such as determination of resuscitation status, investigation for an advance directive, and clinicians’ discussions with the family about goals of care that could be routinely included in chart documentation. In addition, the medical record is a key mechanism for communication among clinicians, helping to coordinate activities of multiple caregivers and ensuring that each is aware of key processes that have and have not been performed by various clinicians. Third, we did not measure some specific characteristics of the sites. The data suggest that unmeasured site characteristics do influence performance of care processes and deserve further study. Finally, our purpose here was to describe the prevalence and predictors of ICU performance of certain palliative care processes, not to evaluate associations between these processes and outcomes of care, such as family satisfaction or well-being. Measuring the relationship between care processes and such outcomes is an important additional step.