Leveraging the electronic health record to identify delivery of goal-concordant care
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
Goal-concordant care (GCC) is recognized as the highest quality of care and most important outcome measure for serious illness research, yet there is no agreed-upon or validated method to measure it.
Objective
Assess feasibility of measuring GCC using clinical documentation in the electronic health record (EHR).
Design
Retrospective chart review study.
Participants
Adults with ≥50% predicted six-month mortality risk admitted to three urban hospitals in a single health system. All participants had goals-of-care (GOC) discussions documented in the EHR 6 months before and 6 months after admission manually classified into one of four categories of goals: (1) comfort-focused, (2) maintain or improve function, (3) life-extension, or (4) unclear.
Main Measures
Pairs of physician-coders independently reviewed EHR notes from 6 months before through 6 months after admission to identify and classify care received between each documented GOC discussion into one of the four goals categories. Epochs between GOC discussions were then coded as goal-concordant if GOC and care received classifications were aligned, goal-discordant if they were misaligned, or uncertain if either classification was unclear or not documented. Coder inter-rater reliability was assessed using kappa statistics.
Key Results
Inter-rater reliability for classifying care received was almost perfect (95% interrater agreement; Cohen’s kappa=0.92; 95% CI, 0.86-0.99). Of 398 total epochs across 109 unique patients, 198 (50%) were goal-concordant, 112 (28%) were of uncertain concordance, and 88 (22%) were goal-discordant. Eighty (73%) patients received care of uncertain concordance during at least one epoch. Forty-eight (44%) patients received goal-discordant care during at least one epoch.
Conclusions
Clinician chart review was a feasible method for measuring GCC and can inform natural language processing and machine learning methods to improve the clinical and research utility of this method. More work is needed to understand the driving factors underlying the high rate of uncertain concordance and goal-discordant care identified among this seriously ill cohort.
