Overall US Hospice Quality According to Decedent Caregivers—Natural Language Processing and Sentiment Analysis of 3389 Online Caregiver Reviews

Abstract

Objectives: With an untapped quality resource in online hospice reviews, study aims were exploring hospice caregiver experiences and assessing their expectations of the hospice Medicare benefit. Methods: Topical and sentiment analysis was conducted using natural language processing (NLP) of Google and Yelp caregiver reviews (n = 3393) between 2013-2023 using Google NLP. Stratified sampling weighted by hospice size to approximate the daily census of US hospice enrollees. Results: Overall caregiver sentiment of hospice care was neutral (S = .14). Therapeutic, achievable expectations and misperceptions, unachievable expectations were, respectively, the most and least prevalent domains. Four topics with the highest prevalence, all had moderately positive sentiments: caring staff, staff professionalism and knowledge; emotional, spiritual, bereavement support; and responsive, timely or helpful. Lowest sentiments scores were lack of staffing; promises made, but not kept, pain, symptoms and medications; sped-up death, hasted, or sedated; and money, staff motivations. Significance of Results: Caregivers overall rating of hospice was neutral, largely due to moderate sentiment on achievable expectations in two-thirds of reviews mixed with unachievable expectations in one-sixth of reviews. Hospice caregivers were most likely to recommend hospices with caring staff, providing quality care, responsive to requests, and offering family support. Lack of staff, inadequate pain-symptom management were the two biggest barriers to hospice quality. All eight CAHPS measures were found in the discovered review topics. Close-ended CAHPS scores and open-ended online reviews have complementary insights. Future research should explore associations between CAHPS and review insights.