The exact text we saved

This is the version exactly as it was saved, character for character. The SHA-256 hash below is a 64-character code worked out from that text. Run the same text through SHA-256 and you get the same code; change a single letter and the code comes out completely different. Each version also stores the previous version’s hash, which chains them together so no version can be swapped out quietly.

TRUVACE RECORD VERSION
record: TRV-2026-2499
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-11-12T07:43:37.559911Z
status: published
lens: trace
sector: health
headline: Detecting spiritual pain in electronic medical records using large language models
dek: PURPOSE: Spiritual pain is one dimension of total pain in terminally ill patients with cancer. Reducing the burden of assessing spiritual pain in electronic medical records is essential for improving the quality of spiritual care. Large language models may offer a solution. This study explored whether large language models can be used to develop an automatic system for accurately identifying spiritual pain in patients with terminal cancer. METHODS: We used unstructured nursing records from patients with cancer w…
gain_title: In a study of 500 nursing records from 109 dying cancer patients in Japan, Gemini and Copilot each identified all 56 records with spiritual pain (sensitivity 1.00), with negative predictive value of 1.00.
problem_title: In the same study, Gemini's positive predictive value was only 0.34 and Copilot's 0.25, and Gemini produced 111 false positives, often mistaking declining physical function or delirium for spiritual pain.
trace_subject: Detecting spiritual pain in palliative cancer nursing records with large language models
gain_reading: In a study of 500 nursing records from 109 dying cancer patients in Japan, Gemini and Copilot each identified all 56 records with spiritual pain (sensitivity 1.00), with negative predictive value of 1.00.
gain_evidence: Sensitivity for Gemini and Copilot was 1.00 | negative predictive value was 1.00 and 1.00
problem_reading: In the same study, Gemini's positive predictive value was only 0.34 and Copilot's 0.25, and Gemini produced 111 false positives, often mistaking declining physical function or delirium for spiritual pain.
problem_evidence: positive predictive value was 0.34 and 0.25 | the models frequently misclassified decreased physical functions and delirium as spiritual pain
quick_read: Researchers at Tohoku University tested whether two AI chatbots, Gemini-2.5-flash and Microsoft Copilot, could find spiritual pain in free-text nursing notes. They used 500 records from 109 cancer patients who died in the hospital's palliative care unit between July and December 2024. Manual review found spiritual pain in 56 records (11.2%). Both models caught all of them, a sensitivity of 1.00.

The cost was a high rate of false alarms. Specificity was 0.75 for Gemini and 0.61 for Copilot, and positive predictive value was 0.34 and 0.25, so most flagged records did not actually show spiritual pain. The authors say this could cause alert fatigue for clinicians and that the tool cannot work as a standalone diagnostic system.

Spiritual care is often under-provided, and the authors note that more non-specialists will need to address it as palliative care demand grows. They suggest a two-stage process, with AI screening followed by clinician verification, but say that needs testing. All results come from one hospital with pre-filtered records.
limitation: The study used records from one Japanese university hospital, only two models, and 500 pre-filtered records from 109 patients, which the authors say may overestimate real-world performance.
tag: Dual reading
key_points: The records were Japanese free-text nursing notes written within 3 weeks before death; 500 were randomly selected from 1,808 records that matched spiritual-pain search terms. | Patients had a mean age of 70.9 years and 57 (52.3%) were male; head and neck cancer (14 patients) and pancreatic cancer (13) were the most common primary sites. | Sensitivity was 1.00 for both models; specificity was 0.75 for Gemini and 0.61 for Copilot; accuracy was 0.78 and 0.66. | Positive predictive value was 0.34 for Gemini and 0.25 for Copilot, while negative predictive value was 1.00 for both. | Of Gemini's 111 false positives, 44 mistook deteriorating physical function and 28 mistook delirium for spiritual pain. | An earlier classical language-processing system reported sensitivity of 0.82, specificity of 0.84 and positive predictive value of 0.28 on the same task. | Two researchers labeled each record independently; agreement (Cohen's kappa) ranged from 0.72 to 0.87 across categories.
rundown: The team prompted both models in Japanese with a definition of spiritual pain from the SpiPas nursing assessment sheet plus example sentences. No model training was needed. Gemini and Copilot were used because they were accessible inside the Tohoku University network, which met personal information protection requirements.

For subdomains of spiritual pain, sensitivity ranged from 0.74 to 0.91 and positive predictive value from 0.11 to 0.57. In the error analysis of Gemini's 111 false positives, other causes included overinterpreting patients' words (15), interpreting physical symptoms as spiritual pain (15), and mistaking perception of prognosis (5) or psychological symptoms (2).

The authors say deteriorating function and delirium can trigger spiritual pain, so even human reviewers may struggle to separate them. They propose clearer prompts, stricter definitions and step-by-step prompting to cut errors, and suggest two-stage screening with clinician verification.

The authors list several limits: a single Japanese hospital, only two models, a small sample with multiple records per patient, and pre-filtering by search terms. They say these factors may have led to overestimated performance.
sources:
- peer_reviewed | Supportive Care in Cancer | https://doi.org/10.1007/s00520-026-11416-0 | 2026-10-29
prev: 0000000000000000000000000000000000000000000000000000000000000000

SHA-256 hash

ff04ebcdc98ad80de147eb40991bf57d34975a8068ee7d6b8e0bd44cfff2bc0e

Previous version’s hash

0000000000000000000000000000000000000000000000000000000000000000