Cancer doesn't know what day of the week it is: Temporal trends in day of death, 2000-2017.

12033 Background: Improving end-of-life care is an opportunity to reduce trauma for both patients and caregivers. Studies support the existence of a psychosomatic phenomenon known as the “holiday effect,” allowing critically ill patients to enter a bargaining phase and postpone death until a given event. We hypothesized that there may be a temporal trend as patients with cancer attempt to delay death until the arrival of family on weekends or the passage of a holiday. Methods: All deaths due to malignant neoplasms (ICD C00-C97) from 2000-2017 were collected from the National Center for Health Statistics. Outcomes were the days of week and months of year that death occurred. Weekend mortality was defined as death on Friday, Saturday, or Sunday. Holiday mortality was defined as death in December or January. Chi-squared tests were performed to determine differences in deaths by day or month. Logistic regression examined associations between day or month of death and age, education, race/ethnicity, location of death, marital status, race, and sex. Statistical analyses were performed using SAS statistical software, version 9.4 (SAS Institute Inc). Results: 10,305,990 deaths due to cancer were recorded from 2000-2017. Deaths were not uniform across day of week or month of year (p < 0.001 each). Each year, rates of death increased progressively from Monday to Thursday, peaked on Friday and Saturday, and declined on Sunday. There was a relative 3.4% difference in death rate between the peak on Friday and nadir of deaths on Monday. Each year there was a consistent increase in deaths in December and January, with a nadir in February with a relative difference of 10.2%. Multivariate logistic regression assessed associations in weekend and holiday cancer death. Non-hospital death (home, nursing facility, hospice) was more likely to occur on the weekend (aOR from 1.03-1.05, p≤0.01 for all). Hispanic patients (aOR 1.01, p = 0.001) were more likely to die on weekends. Females (aOR 0.99, p = 0.02), those who were divorced (aOR 0.99, p = 0.04) and those with pancreatic cancer (aOR 0.99, p = 0.01) were less likely to die on weekends. Non-hospital death was less likely to occur on Holiday months (aOR from 0.95-0.96, p≤0.01 for all). Those > 75 were more likely to pass away in Holiday months compared to younger age groups (OR 1.02-1.03, p < 0.05 for both). Cancer type was associated with holiday mortality; those with breast or prostate cancer were more likely to pass away (OR 1.01, p < 0.05 for both) during the holidays compared to those with lung cancer. Conclusions: There is a human element to death. The unequal distributions of deaths due to cancer in this study are small but significant; they suggest a non-biological variation that precludes cancer mortality from being strictly date agnostic. Temporal trends in cancer death highlight an opportunity to improve end of life care and demonstrate how care should shift focus to what matters to patients including time with family.