Skip to main content
Tessa NewsLink
Paste a health news link, or browse
Latest checks

Source study found

Story checked

Machine learning can identify older heart failure patients at risk of losing independence (opens in a new tab)

news-medical.net · 2026-10-20

Short answerEvidenceSource

Short answer

Mixed

Mixed.

The claims we could check match the study, but some claims were not covered by the evidence reviewed.

  • 3 supported
  • 3 not covered

Checked against the study summary. The full text wasn't available, so some details couldn't be settled either way.

Share this check

Follow the evidence trail
1
2

NewsLink checks it

Mixed

Every claim we could check holds up. Three of six claims match the study. This overall rating is based only on the claims we could check. Three claims the study doesn't address.

  • 3 supported
  • 3 not covered
Open claim evidence
3
Then inspect each claim

Evidence layer

Claim by claim

Each claim gets a verdict. Expand it to see the evidence directly below.

6 claims in this story

Showing all 6 claimsChoose a verdict to focus the list.

Claim 1 of 6Not coveredThe model used 77 candidate predictors and incorporated discharge physical-function assessments such as maximum gait speed, the Short Physical Performance Battery, handgrip strength, and the Barthel Index, along with preadmission frailty measured by the Kihon Checklist.View evidenceHide evidence

As stated77 candidate predictors

Why this verdict

The abstract-level profile supports some listed predictors, including maximum gait speed, discharge Barthel Index, preadmission frailty score, and age. However, it does not verify the stated 77 candidate predictors or all named discharge physical-function measures such as SPPB and handgrip strength. This level of detail is not available at abstract depth.

1

Study evidence

A parsimonious Top-10 XGBoost model predicted 1-year functional outcome (maintenance vs decline vs death) with good discrimination and outperformed Kihon Checklist–based scores.AUC 0.75 (95% CI 0.74–0.76)

“We analyzed 6382 older patients (median age, 82 years) from a nationwide prospective cohort (J-Proof HF registry).”
Claim 2 of 6Not coveredAt 1 year, 2,076 patients experienced functional decline and 928 died, while 3,378 maintained functional independence.View evidenceHide evidence

As stated2,076 functional decline; 928 died; 3,378 maintained independence

Why this verdict

The profile reports abstract-level event rates—32.5% functional decline and 14.5% death in a cohort of 6,382—but not the exact counts stated in the story. The counts are broadly consistent with the reported rates and total sample, but exact values are not verifiable from the supplied abstract-level profile.

1

Study evidence

A parsimonious Top-10 XGBoost model predicted 1-year functional outcome (maintenance vs decline vs death) with good discrimination and outperformed Kihon Checklist–based scores.AUC 0.75 (95% CI 0.74–0.76)

“We analyzed 6382 older patients (median age, 82 years) from a nationwide prospective cohort (J-Proof HF registry).”
2

Study evidence

The model outputs were used to identify four distinct risk phenotypes among older patients hospitalized for heart failure.

“The model identified 4 risk phenotypes, including a \"low-mortality/high-decline\" group (n = 1732) with a 57.0% functional decline rate despite an 89.0% survival probability.”
Claim 3 of 6Not coveredThe researchers developed a prototype web application that estimates individual risk at discharge and simulates how changes in physical function, such as faster gait speed, may affect predicted risk, but the article says implementation studies are still needed.View evidenceHide evidence

Why this verdict

The supplied abstract-level paper profile does not mention a prototype web application, individualized risk estimation interface, or simulation of changes in physical function such as gait speed. The implementation caveat is scientifically appropriate, but the web-application claim itself is not verifiable from the supplied abstract-depth evidence.

Claim 4 of 6SupportedA research team led by Kanji Yamada developed an eXtreme Gradient Boosting (XGBoost) model to predict functional maintenance, functional decline, or death 1 year after discharge in older adults hospitalized for heart failure.View evidenceHide evidence

As stated6,382 patients aged 65 years or older

Why this verdict

The abstract-level profile supports that an XGBoost model was developed to predict a 1-year 3-class outcome—functional maintenance, functional decline, or death—in 6,382 patients aged ≥65 hospitalized for heart failure. The profile does not independently verify the named leadership by Kanji Yamada, but the central scientific claim is supported.

1

Study evidence

A parsimonious Top-10 XGBoost model predicted 1-year functional outcome (maintenance vs decline vs death) with good discrimination and outperformed Kihon Checklist–based scores.AUC 0.75 (95% CI 0.74–0.76)

“We analyzed 6382 older patients (median age, 82 years) from a nationwide prospective cohort (J-Proof HF registry).”
Claim 5 of 6SupportedThe Top-10 XGBoost model significantly outperformed the modified Kihon Checklist score and Kihon Checklist total score.View evidenceHide evidence

As statedTop-10 model vs KCL-based scores

Why this verdict

The profile supports that the Top-10 XGBoost model had higher discrimination than Kihon Checklist-based scores, with AUC 0.75 versus KCL-based AUCs of 0.66–0.69 and P < .001. The abstract-level profile does not spell out every named KCL variant, but the comparative outperformance claim is supported.

1

Study evidence

A parsimonious Top-10 XGBoost model predicted 1-year functional outcome (maintenance vs decline vs death) with good discrimination and outperformed Kihon Checklist–based scores.AUC 0.75 (95% CI 0.74–0.76)

“We analyzed 6382 older patients (median age, 82 years) from a nationwide prospective cohort (J-Proof HF registry).”
2

Study evidence

The parsimonious Top-10 XGBoost model discriminated 1-year functional outcomes better than KCL-based frailty scores in the same cohort.AUC 0.75 (95% CI 0.74–0.76) for Top-10 XGBoost vs AUC range 0.66–0.69 for KCL-based scores; P < .001 for difference.

“Performance was evaluated via nested leave-one-site-out validation and benchmarked against the Kihon Checklist (KCL) frailty screening tool.”
Claim 6 of 6SupportedThe model identified four risk phenotypes, including a low-mortality/high-decline group of 1,732 patients with a 57% rate of functional decline despite an 89% probability of survival at 1 year.View evidenceHide evidence

As stated1,732 patients; 57% decline; 89% survival

Why this verdict

The profile directly reports that the model identified 4 risk phenotypes, including a low-mortality/high-decline group of n = 1,732 with a 57.0% functional decline rate and 89.0% 1-year survival probability.

1

Study evidence

The parsimonious Top-10 XGBoost model discriminated 1-year functional outcomes better than KCL-based frailty scores in the same cohort.AUC 0.75 (95% CI 0.74–0.76) for Top-10 XGBoost vs AUC range 0.66–0.69 for KCL-based scores; P < .001 for difference.

“Performance was evaluated via nested leave-one-site-out validation and benchmarked against the Kihon Checklist (KCL) frailty screening tool.”
2

Study evidence

The model outputs were used to identify four distinct risk phenotypes among older patients hospitalized for heart failure.

“The model identified 4 risk phenotypes, including a \"low-mortality/high-decline\" group (n = 1732) with a 57.0% functional decline rate despite an 89.0% survival probability.”
Then look for missing context

Context layer

What the story left out

Important study details the story did not include.

  • Study design and population: secondary analysis of the J-Proof HF nationwide prospective registry, including patients aged ≥65 hospitalized for heart failure, prescribed rehabilitation, and independent before admission.

    The story reflects the older hospitalized heart-failure population but does not convey key eligibility and generalizability constraints, including rehabilitation prescription, preadmission independence, and the Japanese registry context.

    From Prospective registry secondary analysis; ML model development with internal-external (site-held-out) validation; compara

  • Model validation approach: nested leave-one-site-out validation with AUC as the primary reported performance metric.

    The story says the model outperformed KCL-based scores but does not mention the validation strategy or the main performance metric, which are important for interpreting model-development evidence.

    From Prospective registry secondary analysis; ML model development with internal-external (site-held-out) validation; compara

4 things the story did carry across
  • Primary study contribution: development and validation of an XGBoost model for 1-year functional maintenance, functional decline, or death in older hospitalized heart-failure patients.
  • Benchmark finding: the Top-10 XGBoost model outperformed Kihon Checklist-based scores for the same 1-year 3-class outcome.
  • Risk phenotyping: model outputs identified four phenotypes, including a low-mortality/high-decline subgroup with n = 1,732, 57.0% decline, and 89.0% survival.
  • Clinical-use limitation: evidence supports prediction and risk stratification, not proof that model-guided decisions or rehabilitation targeting improves patient outcomes.
Then read the study layer

Study layer

Study at a glance

Scan the study first. Expand only the parts you want to inspect.

Pieces of work

3

Evidence read

study summary

Lead result

secondary data

1Lead resultsecondary dataDevelop and validate a machine-learning model (XGBoost) to predict 1-year functional maintenance vs functional decline vs death in older patients hospitalized for heart failure, accounting for competing mortality risk.Prospective registry secondary analysis; ML model development with internal-external (site-held-out) validationExpand

In plain English

Analysis of 6,382 older (median age 82) hospitalized heart failure patients from the nationwide J-Proof HF prospective registry developed and internally/externally validated an XGBoost model to predict a 1-year, 3-class outcome (functional maintenance vs functional decline vs death). The parsimonious Top-10 model (including maximum gait speed, discharge Barthel Index, preadmission frailty score, and age) achieved good discrimination (AUC 0.75, 95% CI 0.74–0.76) and outperformed Kihon Checklist–based scores. At 1 year, 32.5% experienced functional decline and 14.5% died; clustering of model outputs identified risk phenotypes including a low-mortality/high-decline group with 57.0% decline and 89.0% survival.

Key findings

  • A parsimonious Top-10 XGBoost model predicted 1-year functional outcome (maintenance vs decline vs death) with good discrimination and outperformed Kihon Checklist–based scores.AUC 0.75 (95% CI 0.74–0.76)
  • Observed 1-year event rates in the analyzed cohort.
“We analyzed 6382 older patients (median age, 82 years) from a nationwide prospective cohort (J-Proof HF registry).”
2secondary dataBenchmark the ML model’s predictive performance against an existing frailty screening tool (Kihon Checklist) for the same outcomes.comparative model performance evaluationExpand

In plain English

Within the J-Proof HF prospective cohort (n=6,382; median age 82), the authors benchmarked a parsimonious Top-10 XGBoost model against Kihon Checklist (KCL) frailty screening scores for predicting a 1-year, 3-class outcome (functional maintenance, functional decline defined as BI <85/housebound or bedridden, or death). Using nested leave-one-site-out validation, the XGBoost model showed higher discrimination (AUC 0.75, 95% CI 0.74–0.76) compared with KCL-based scores (AUC range 0.66–0.69), with the difference reported as statistically significant (P < .001). The ML model also yielded risk phenotypes, including a “low-mortality/high-decline” group (n=1,732) with 57.0% functional decline and 89.0% 1-year survival.

Key findings

  • The parsimonious Top-10 XGBoost model discriminated 1-year functional outcomes better than KCL-based frailty scores in the same cohort.AUC 0.75 (95% CI 0.74–0.76) for Top-10 XGBoost vs AUC range 0.66–0.69 for KCL-based scores; P < .001 for difference.
  • The model-derived risk phenotyping identified a subgroup with high probability of functional decline despite relatively low mortality.‘Low-mortality/high-decline’ phenotype (n = 1,732): 57.0% functional decline rate; 89.0% 1-year survival.
“Performance was evaluated via nested leave-one-site-out validation and benchmarked against the Kihon Checklist (KCL) frailty screening tool.”
What this piece can’t prove
  • Cohort restricted to hospitalized older Japanese HF patients eligible for rehabilitation and independent preadmission (BI ≥85), which may limit generalizability.
  • Unclear whether comparator thresholds or score tuning for KCL were optimized within the same validation framework, potentially affecting fairness of comparison.

2 further details could not be confirmed from the summary.

3secondary dataDerive/identify distinct risk phenotypes (patient subgroups) from the model outputs, including a low-mortality/high-decline phenotype, to support targeted rehabilitation strategies.phenotype derivation from model outputsExpand

In plain English

From a nationwide prospective cohort of 6,382 older patients hospitalized for heart failure (median age 82) who were independent preadmission, the authors report deriving 4 risk phenotypes from the outputs of an XGBoost prediction model for 1-year outcomes. One phenotype is described as a "low‑mortality/high‑decline" group (n = 1,732) with a 57.0% 1‑year functional decline rate and an 89.0% 1‑year survival probability. The abstract states these phenotypes can identify patients at high risk of functional decline despite favorable survival to support targeted rehabilitation, but it does not specify the clustering/stratification method or provide phenotype validation details.

Key findings

  • The model outputs were used to identify four distinct risk phenotypes among older patients hospitalized for heart failure.
  • One reported phenotype — labelled "low‑mortality/high‑decline" — included 1,732 patients and had a 57.0% 1‑year functional decline rate despite an 89.0% 1‑year survival probability.
“The model identified 4 risk phenotypes, including a \"low-mortality/high-decline\" group (n = 1732) with a 57.0% functional decline rate despite an 89.0% survival probability.”
What this piece can’t prove
  • Abstract does not specify the analytic method used to derive phenotypes from model outputs (e.g., clustering algorithm, stratification thresholds, or rationale for selecting four phenotypes).
  • Abstract provides no information on the internal stability or external validation of the identified phenotypes.

1 further detail could not be confirmed from the summary.

Finally, the search trail

Method layer

NewsLink found the paper. Tessa takes you deeper.

NewsLink checks the story. Tessa is where you inspect the paper, authors, evidence, and research context.

Papers considered

The selected paper, plus nearby candidates.

PubMed, Europe PMC, Crossref · 36 candidate papers

Candidate

Utilizing SGLT2i to Mitigate Proteinuria in Cardiac Transplant Patients on mTORi

Journal of Cardiac Failure · 2026 · Crossref

Source (opens in a new tab)

And 30 more candidates considered.