Can donor narratives yield insights? A natural language processing proof of concept to facilitate kidney allocation

Abstract

Although expedited placement could ameliorate stagnant kidney utilization, precisely identifying difficult‐to‐place organs is crucial to mitigate potential harms associated with this policy. Existing algorithms have only leveraged structured data from the Organ Procurement and Transplantation Network (OPTN); however, detailed, free text case information about a donor exists. No known research exists about the utility of these data. We developed a model to predict the probability of delay or discard for adult deceased kidney donors between 2010 and 2018, leveraging donor free text data. The resultant model had a c‐statistic of 0.75 compared to 0.80 ( Reduced Probability of Delay or Discard [model], r‐PODD) and 0.77 ( Kidney Donor Profile Index, KDPI) on the test dataset. Analysis of the top predictive words suggest both known and potentially novel clinical factors (ie, a known factor such as hypertension vs a novel factor such as stents), and nuanced social factors (intravenous drug use) could negatively affect kidney utilization. These findings suggest that donor narratives have utility; the natural language processing (NLP) model is only moderately correlated with existing indices and provides directional evidence about additional cardiovascular risk factors that may affect kidney utilization. More research is needed to understand the potential to enhance existing indices of kidney utilization to better enable and mitigate the effects of policy interventions such as expedited placement.

Hardest-to-place kidney transplant outcomes in the United States

“…Our finding that minority of centers utilize hardest-toplace kidneys supports previous recommendations that allocation algorithms to expedite organs could "leverage a sliding scale, a confidence-linked delay to alternative allocation policy based on the number of declines." 38 Our study conclusions are limited by its retrospective nature.…”

“…To predict deceased donor kidneys that will experience placement difficulties, researchers have developed sophisticated models, 21,36,37 but their sensitivity is limited. Our finding that a minority of centers utilize hardest‐to‐place kidneys supports previous recommendations that allocation algorithms to expedite organs could “leverage a sliding scale, a confidence‐linked delay to alternative allocation policy based on the number of declines.” 38 …”

Classifying Comments on Social Media Related to Living Kidney Donation: Machine Learning Training and Validation Study

“…For deceased kidney donation, there are a handful of studies that have utilized modern computer-science methods to analyze motivations and challenges associated with kidney donation. A recent study [ 33 ] discussed the use of natural language processing to glean information about deceased donors and the prospective utility of their kidneys. This information was retrieved from the United Network for Organ Sharing’s DonorNet program, in which organ procurement organizations enter raw text about the donors’ medical and social history, the history of their admissions, and other noteworthy information.…”

The promise of machine learning applications in solid organ transplantation

“… Can donor narratives yield insights? A natural language processing proof-of- concept to facilitate kidney allocation Placona et al 23 74,041 donors Kidney Natural Language Processing model was used to predict the delay or discard of adult deceased donors based on donor-free-text data (C-statistic = 0.75). Performed on par with traditional methods.…”

“…Decision-support tools that identify deceased donor kidneys which may experience placement difficulties and streamline this process may increase donor organ utilization. To better assist in accept/decline decisions for patients needing adult kidney donors, Natural Language Processing (NLP) methods that tap into free-text data beside structured data from donor information were studied 23 . Using this method, both known and new key clinical terms holding significant predictive value were discovered, producing a model with a C-statistic of 0.75 for accept or decline decisions which is comparable with the performance of traditional indices including Reduced Probability of Delay or Discard (r-PODD, C-statistics = 0.80) and Kidney Donor Profile Index (KDPI, C-statistics = 0.77).…”