RPD102 - Prevalence of false positives when the Z51.5 code is used at hospitals without palliative care programs

Key Message : In nationally representative data, nearly 80% of hospitals that lack PC programs use the "palliative care" (Z51.5) code; 15% of supposed SPC encounters were actually at non-PC hospitals. Combined with prior studies showing low sensitivity, this raises grave concerns about many studies of “palliative care” relying on that code.

Abstract:

Background: Many studies have used datasets that rely solely on the “encounter for palliative care” Z51.5 ICD10 code to determine specialty palliative care (SPC) involvement. Prior validation studies at hospitals with high SPC involvement showed low sensitivity and moderate positive predictive value (1, 2).

Purpose: Our objective was to determine whether and how often the Z51.5 code was being used at hospitals that do not actually offer SPC, and what portion of the total cases this represented.

Methods: Data from the AHRQ “National Inpatient Sample” of hospital discharges for 15 representative states in 2019 was linked that to American Hospital Association survey data regarding palliative care service availability. We analyzed the use of the Z51.5 code at PC and non-PC hospitals in hospital billing data.

Results: At the 357 hospitals did not offer SPC, 279 (78.2%) used the Z51.5 code at least once. Those non-PC hospitals had 1.76 million discharges, 41,607 (2.4%) of which had the Z515 code. At the 413 hospitals that offered SPC, 234,179 (3.2%) of the 7,361,769 discharges had the Z51.5 code. Based on prior studies where the false discovery rate was 27% to 46%, between 62,228 and 107,772 of those Z51.5 cases at SPC hospitals may also be false (3, 4).
Implications: Over 130 published studies of “palliative care” relied solely on the Z51.5 code, despite validation studies at high-SPC hospitals indicating such data may miss 43% to 50% of actual SPC encounters. This new study estimates that 37.5% to 54.2% of supposed SPC encounters in such data may be false positives. A dataset in which half of SPC encounters are missed and half of supposed SPC encounters are false should not be used for PC research, and the previously published studies should be viewed with this in mind.

References: 1. Cassel JB. What Does “Palliative Care” Represent in Research Using Secondary Data? J Pain Symptom Manage. 2024;68(4):421-428. doi:10.1016/j.jpainsymman.2024.07.011
2. Heitner R, Chambers B, Silvers A, Bowman B, Johnson KS. “Palliative Care” as an Outcome Measure and Its Impact on Our Interpretation of Racial Disparities. J Pain Symptom Manage. 2024;67(1):e114-e116. doi:10.1016/j.jpainsymman.2023.10.003
3. Hua M, Li G, Clancy C, Morrison RS, Wunsch H. Validation of the V66.7 Code for Palliative Care Consultation in a Single Academic Medical Center. J Palliat Med. 2017;20(4):372-377. doi:10.1089/jpm.2016.0363
4. O’Keefe S, Czaja AS. Validation of Administrative Codes for Palliative Care Consultation Among Critically Ill Children. Hosp Pediatr. 2021;11(2):179-182. doi:10.1542/hpeds.2020-001263