Assessment of Clinical Palliative Care Trigger Status vs Actual Needs Among Critically Ill Patients and Their Family Members

This cohort study assesses whether higher levels of family member–reported palliative care needs are observed among those whose critically ill loved ones meet a clinical palliative care trigger compared with those who do not meet such a trigger.

Abstracted from medical record, including review of admission data, Code Blue form note, and ICU admission clinician notes Advanced cancer Currently active and advanced or metastatic solid or liquid cancers (i.e., ICD-10 codes C18.9, C34.9X, C50.91X, C78.X, C79.X). Solid tumors generally included Stage 4 colon, lung, breast cancers or other cancers described as 'metastatic.' For liquid cancers (e.g., leukemia, lymphoma), documentation of 'refractory' or 'advanced' status in treating oncologist's progress note Abstracted from medical record, including history and physical and progress notes from hematology-oncology service

Very recent ICU admission
≥1 ICU admission in the 3 months before the current admission Abstracted from medical record for admissions within the Duke Health System and history and physical note for admissions from outside the Duke Health System Very recent hospital admissions ≥2 hospital admissions in the 3 months before the current admission Abstracted from medical record for admissions within the three hospitals within the Duke Health System, as well as the history and physical note for admissions from outside the Duke Health System Admission from facility Admission to hospital from skilled nursing facility, inpatient rehabilitation facility, or long-term acute care facility Abstracted from medical record, including review of admission data, the history and physical note, case management notes, and nursing intake form from time of hospital admission Dementia Alzheimer's, vascular, multi-infarct, or other cause (i.e., ICD-10 codes F01.50, F01.51, F03.90, F03.91, G30.0, G30.1, G30.8, G30.9, G31.83) Must have been listed as a key issue in the problem list in medical chart or described as 'moderate,' 'severe,' 'serious,' or 'clinically significant' in admission history and physical note eTable 2. Performance characteristics of the presence of clinical palliative care triggers for identifying serious unmet needs: additional information on analytic approach We determined performance characteristics using a 2 x 2 table set up with the NEST score as the gold standard condition of unmet need and the presence of ≥1 trigger as the test condition. We chose to combine all triggers in this way given that this would presumably optimize sensitivity given the application of clinical triggers as a screening, rather than diagnostic test.
Serious need present (i.e., high NEST score) Serious need absent (i.e., low NEST score) ≥1 Trigger present A B Trigger absent C D Sensitivity is the true positive rate, or the proportion of true positives (i.e., high NEST score) among all positives (i.e., number who meet a trigger): A / A+C.
Specificity is the true negative rate, or the proportion of true negatives (i.e., low NEST score) among all negatives (i.e., number who do not meet a trigger): D / B+D Positive predictive value (PPV) is the proportion of those with a high NEST score (i.e., serious needs) who meet a trigger: A / A+B Negative predictive value (NPV) is the proportion of those with a low NEST score (i.e., less serious needs) who do not meet a trigger: D / C+D Accuracy is the correct classification rate, or the proportion of those who are either true positives (i.e., high NEST score and therefore had serious needs) and who meet a trigger or true negatives (i.e., low NEST score and therefore did not have serious needs) and who do not meet a trigger: A+B / A+B+C+D Positive likelihood ratio (PLR) represents the true positive rate (i.e., sensitivity) divided by the false positive rate (1-specificity), or the ratio of the probability of having serious unmet needs when the patient has a trigger divided by the probability of having low needs when the patient does not meet a trigger.
Negative likelihood ratio (NLR) represents the false negative rate (i.e., 1-sensitivity) divided by the true negative rate (specificity), or the ratio of the probability of having serious unmet needs when the patient does not have a trigger divided by the probability of having low needs when a trigger is not present.
C statistic is equal to the area under the curve of a graph of rates of true positives (sensitivity) and false positives (1-specificity).