The validity of using health administrative data to identify the involvement of specialized pediatric palliative care teams in children with cancer in Ontario, Canada

Background — Population-based research to identify underserviced populations and the impact of palliative care (PC) is limited as the validity of such data to identify PC services is largely unknown.

Objective —To determine the validity of using such data to identify the involvement of specialized pediatric PC teams among children with cancer.

Design — Retrospective cohort.

Subjects — Ontario children with cancer who died between 2000 and 2012, received care through a pediatric institution with a specialized PC team and a clinical PC database.

Measurements— All patients in the clinical databases were linked to population-based health services administrative databases. Six algorithms were created to indicate the use of formal pediatric PC teams based on the record type (physician billings vs. inpatient records vs. both) and number of eligible codes required (≥1 vs. ≥2). Each was validated against the pediatric PC clinical databases.

Results — The cohort comprised 572 children; 243 were in the clinical databases. Algorithms using only inpatient records had high specificity (80%-95%) but poor sensitivity (21%-56%). Including physician billings increased sensitivity but lowered specificity. The algorithm with overall best performance required ≥2 physician billing or inpatient diagnosis codes indicating PC [sensitivity 0.79 (95% CI 0.73-0.84), specificity 0.58 (95% CI 0.53-0.64)].

Conclusions — Health administrative data identifies involvement of specialized pediatric PC teams with good sensitivity but low specificity. Studies using such data alone to compare patients receiving and not receiving specialized pediatric PC are at significant risk of misclassification and potential bias. Population-based PC databases should be established to conduct rigorous population-based PC research.