Pith. sign in

REVIEW 3 major objections 5 minor 43 references

Unsupervised Latent Pattern Analysis for Estimating Type 2 Diabetes Risk in Undiagnosed Populations

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read An unsupervised score built from diagnosed patients' comorbidity and medication patterns estimates T2DM risk in undiagnosed people without needing confirmed negatives.

desk verdict A cleanly-written unsupervised risk-scoring method for T2DM in claims data, but the validation is circular and the score's direction has an unaddressed sign error; worth a referee to push for external validation. read the letter →

arxiv 2505.21824 v2 pith:3ASXU4Z2 submitted 2025-05-27 cs.LG stat.AP

Praveen Kumar , Vincent T. Metzger , Scott A. Malec This is my paper
classification cs.LGstat.AP
keywords type2diabetesmellitusundiagnosednon-negativematrixfactorizationKullback-Leiblerdivergenceriskscoreclaimsdatamultimorbidityunsupervisedlearning
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Type 2 diabetes is frequently undiagnosed, and supervised machine-learning models are hampered because claims data provide confirmed positives but no confirmed negatives. This paper proposes an unsupervised pipeline that learns latent comorbidity and medication patterns from diagnosed T2DM patients alone, then scores every undiagnosed patient by how closely their record matches those patterns. The score combines NMF-derived rank-weighted coefficients with KL-divergence prevalence weights, and patients above the 90th percentile of the diagnosed-score distribution are labeled high risk. Applied to 17.1 million US claims records, 3.18% of undiagnosed individuals fall into the high-risk group, in line with national estimates of undiagnosed diabetes. The authors argue this yields an interpretable, scalable tool for prioritizing undiagnosed patients for screening.

What carries the argument

The central object is the risk score $r_i$, built from two learned weights per covariate. $w_j$, the rank-weighted coefficient, is computed from the NMF coefficient matrix $H$ of diagnosed patients: $w_j = \sum_i H_{ij}/r_{ij} / \sum_i 1/r_{ij}$ over latent components, emphasizing covariates that rank high in many components. $d_j$ is the KL divergence between the covariate's prevalence in diagnosed and undiagnosed groups, with smoothing $\epsilon=10^{-8}$. The product $w_j d_j$ assigns each covariate a single importance weight, and the score sums those weights over the covariates present in a patient's record. NMF supplies the latent comorbidity-and-polypharmacy patterns; KL divergence supplies the discrimination signal; the indicator sum makes the result interpretable as a weighted count of T2DM-associated findings.

What would settle it

Give a cohort of people scored high-risk by this method a hemoglobin A1C test (or track their future T2DM diagnoses) and compare with people scored low-risk; if incidences are equal, the score is measuring general multimorbidity rather than T2DM risk. The paper's own limitation section identifies the absence of HbA1C data as the reason this test was not performed.

Watch

Extended reading notes

Core claim

The paper's central claim is that undiagnosed patients can be ranked for T2DM risk by measuring how similar their pre-diagnosis clinical fingerprint is to that of people already diagnosed with T2DM. The fingerprint is a sparse vector of ICD-10 diagnoses and RxNorm medications; NMF factorizes the diagnosed-only matrix into latent components, and each covariate receives a rank-weighted coefficient $w_j$ that rewards consistent high contributions across components. A KL-divergence $d_j$ then measures how much more prevalent that covariate is among diagnosed than undiagnosed patients, and the raw risk score is $r_i = \sum_{j\in S} w_j d_j I(x_{ij}>0)$. Scores are arctangent-scaled to $[0,1)$ and converted to percentiles against the diagnosed distribution; undiagnosed patients above the 90th percentile are called high risk and are suspected to be undiagnosed T2DM cases. The paper supports this with four validations: temporal separation, expert and AI agreement, Jaccard similarity of covariate profiles, and improved classifier performance when low-risk undiagnosed patients serve as negatives.

Load-bearing premise

The load-bearing premise is that an undiagnosed person whose comorbidity and medication record resembles the pre-diagnosis record of diagnosed T2DM patients is genuinely at elevated risk for T2DM, a transfer the study never checks with laboratory values or follow-up diagnoses.

Editorial extensions

If this is right

  • A health system can generate a screening-priority ranking for undiagnosed patients using only confirmed-positive cases, removing the need for hand-labeled negative examples.
  • Low-risk undiagnosed patients can stand in as a cleaner negative set, which the paper shows improves a downstream classifier's accuracy, Matthews correlation, AUC-ROC, and Brier score.
  • The same pipeline, NMF on positives, rank-weighted coefficient selection, KL-divergence weighting, and percentile thresholding, should transfer to other underdiagnosed phenotypes with reliable confirmed-case codes.
  • The roughly 3.18% of undiagnosed people assigned high risk become a concrete target for follow-up screening and early intervention.
  • Because the score is a weighted count of indicators $I(x_{ij}>0)$ rather than raw encounter counts, it is insensitive to how often a patient happens to visit the clinic.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The score is best interpreted as a similarity-to-diagnosed-T2DM index, not a calibrated probability; the paper says the arctangent transform should not be read as a probability, so percentile ranks carry the operational meaning.
  • The closeness of the 3.18% high-risk share to national undiagnosed-prevalence statistics is suggestive but not a direct validation, because the category thresholds were set by expert review against the diagnosed-score distribution rather than by any biomarker.
  • A missing experiment that would settle the transfer assumption is to follow high-risk undiagnosed patients with hemoglobin A1C tests or future diagnosis codes; the paper itself notes the claims data contain no HbA1C measurements.
  • The ranking could be coupled with positive-unlabeled class-proportion estimation to convert relative scores into an estimated count of undiagnosed cases in a particular region or demographic.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes an unsupervised framework for estimating T2DM risk in undiagnosed individuals from claims data. The method applies NMF to pre-diagnosis covariates of diagnosed T2DM patients, derives rank-weighted coefficients (RWC) for features, computes KL divergence between diagnosed and undiagnosed feature prevalences, and forms a risk score as a weighted sum over selected features. Risk categories are defined by percentiles of the diagnosed score distribution. Validation comprises distribution comparisons, agreement with an SME and three AI models, Jaccard similarity between high-risk groups, and an XGBoost classifier trained with negative labels selected by risk-score thresholds. The paper reports that 3.18% of undiagnosed individuals are high-risk, which it claims aligns with CDC undiagnosed diabetes prevalence.

Significance. If the central claim were established, the approach would offer a label-efficient and interpretable alternative to supervised T2DM risk models in settings where confirmed negatives are unavailable, with potential applicability to other phenotypes. The authors include several strengths: a large real-world claims cohort, a clearly specified algorithmic pipeline, and reproducible descriptions of the NMF and score computations. However, the evidence for the core claim that the score estimates T2DM risk in undiagnosed individuals is self-referential, and the KL divergence formulation contains a directional error. The significance of the work therefore rests on unvalidated assumptions rather than demonstrated results.

major comments (3)
  1. [Sec. 2.7, Eq. (6) and Eq. (7)] The KL divergence in Eq. (6) is always nonnegative, so it cannot indicate whether a feature is more prevalent in diagnosed or undiagnosed individuals. The statement in Sec. 2.7 that 'A large KL divergence suggests that feature j is more prevalent among diagnosed individuals' is incorrect. For example, cervicalgia has prevalence 6.36% in diagnosed and 11.19% in undiagnosed (Table 1), yet its d_j is still positive. Because Eq. (7) sums w_j * d_j for every present feature, the risk score rewards features that are more common in undiagnosed individuals just as much as features enriched in diagnosed patients. This undermines the directional interpretation of the risk score and the central claim that it estimates T2DM risk.
  2. [Sec. 3.5, Method 4 (Table 5)] The negative labels in Method 4 are selected from the model's own risk-score thresholds, and then the same covariate space is used to train XGBoost. The observed decline in classification performance as the threshold increases is an expected consequence of labeling undiagnosed individuals by the model's own ranking: higher thresholds include individuals the model itself considers more T2DM-like, which mechanically makes the binary separation harder. This does not provide independent evidence about whether low-risk undiagnosed individuals are true negatives. The conclusion that the model 'can also reliably identify negative cases using a low-risk threshold' is therefore circular.
  3. [Sec. 2.2 and Sec. 4.1] The central transfer assumption, stated in Sec. 2.2 as 'assuming that undiagnosed individuals with similar covariate profiles may also be at risk,' is never tested against actual T2DM outcomes. The paper acknowledges in Sec. 4.1 the lack of HbA1C measurements and reports no follow-up diagnosis data. The only external anchor, the rough match between the 3.18% high-risk proportion and the CDC undiagnosed prevalence in the Discussion, is post-hoc: the risk thresholds were set in consultation with an SME (Sec. 2.8) before the CDC comparison. As a result, the Abstract and Sec. 3.5 claim that high-risk categories 'may in fact be undiagnosed T2DM cases' is unsupported by the presented evidence.
minor comments (5)
  1. [Sec. 2.2] The text contains a typo: 'undiagnose individuals' should be 'undiagnosed individuals.'
  2. [Sec. 2.7, Eq. (6)] The typesetting of Eq. (6) is broken, with stray newline characters and unclear fraction formatting; the equation should be presented as a standard KL divergence expression.
  3. [Algorithm 3] Lines 5-6 of Algorithm 3 split the two terms of the KL formula with a stray '+' and the log arguments are awkwardly nested; the pseudocode should be cleaned up for readability.
  4. [Table 1 caption] The caption contains a typo: 'T able' should be 'Table'; also, the phrase 'Comorbidities and medications are top covariates with high R WC values' could be clarified to specify that the listed items are examples drawn from the top RWC features.
  5. [Sec. 2.8] The justification for choosing arctangent over sigmoid relies on informal claims about 'overconfident estimates' and 'rapid asymptotic behavior' without quantification; a brief reference to the score distribution or a comparison of transformations would be more rigorous.

Circularity Check

3 steps flagged · score 7.0 of 10

Validation methods 1, 3, and 4 restate the construction of the risk score rather than testing it; the T2DM-risk claim rests on an unverified transfer assumption.

  1. self definitional [Section 2.10.1, Validation Method 1 (with Equations 6 and 7)]
    "Since both the training and validation sets comprised diagnosed individuals, we expected their risk score distributions to be similar and substantially higher than those of the undiagnosed cohort. This comparison serves as a form of temporal validation, evaluating the model’s ability to identify individuals at elevated risk who subsequently received a clinical diagnosis of T2DM."

    Equation 7 defines the score as r_i = sum_{j in S} w_j * d_j * I(x_ij > 0), where d_j is the KL divergence between diagnosed and undiagnosed prevalence (Equation 6). Any covariate more prevalent in diagnosed patients contributes positively, so diagnosed patients have higher scores by construction. Moreover, the KL step (Algorithm 1) computes p_d^j from the full diagnosed matrix, which includes the 8,100 validation patients, so the 'validation' separation is partly fitted on the very patients being evaluated. No follow-up diagnosis or HbA1C is used; the comparison is a restatement of the fitting target, not an independent test of T2DM risk.

  2. self definitional [Section 2.10.3, Validation Method 3; results in Section 3.4]
    "To evaluate the robustness of the model’s risk estimation, we hypothesized that high-risk diagnosed individuals and high-risk undiagnosed individuals would share similar covariate patterns, while high-risk diagnosed individuals and low-risk undiagnosed individuals would show dissimilar profiles."

    Both 'high-risk' sets are created by thresholding the same risk score, which is a weighted sum over the selected covariates. High-scoring diagnosed and undiagnosed patients necessarily carry the same high-weight covariates, so their most-frequent covariate lists overlap; low-risk undiagnosed patients lack those covariates, so overlap is low. The Jaccard contrast is an artifact of the score's definition and thresholding, not independent evidence that the score identifies T2DM. It confirms only that the score groups people by similarity to the diagnosed profile, which is exactly how it was constructed.

1 more flagged steps
  1. fitted input called prediction [Section 2.10.4, Validation Method 4; results in Section 3.5 and Table 5]
    "All diagnosed individuals were treated as positive cases, and an equal number of undiagnosed individuals were selected as negative examples based on three different risk score thresholds (≤0.1, ≤0.5, and ≤1.0). We hypothesized that classifiers trained on datasets where negative examples were drawn from the low-risk (≤0.1) undiagnosed group would exhibit superior performance, as these individuals are more likely to be true negatives."

    The negative labels are assigned by thresholding the same risk score that is a linear function of the selected covariates. When the classifier is trained on the same claims covariates, it can trivially separate diagnosed positives from low-score negatives. The observed AUC and MCC decline with higher thresholds is the inevitable effect of including higher-scoring, more positive-like individuals as negatives, i.e., injecting label noise from the model's own ranking. The experiment re-describes the threshold rule rather than validating the risk estimate against actual T2DM status; the paper itself confirms there are no HbA1C measurements (Section 4.1).

full rationale

The paper's core construction—NMF on diagnosed patients, Rank-Weighted Coefficients, KL divergence, and Equation 7—is a self-contained unsupervised procedure and is not circular in itself: the risk score does not use undiagnosed outcomes. The circularity lies in the evidence offered for the central claim that the score estimates T2DM risk in undiagnosed individuals. Validation Method 1 is a tautology: d_j is computed from the diagnosed–undiagnosed prevalence gap, so Equation 7 guarantees diagnosed scores are higher; using the validation patients in that prevalence calculation makes the 'temporal validation' even more directly fitted. Method 3 defines high and low risk by the same score and then finds that high-risk groups share covariates, which is forced by the score's definition. Method 4 constructs negative labels by thresholding the same risk score and trains a classifier on the same covariates; the clean separation of low-score negatives is an inevitable consequence of label construction. None of these validations use HbA1C or follow-up diagnosis; Section 4.1 concedes 'Due to the lack of hemoglobin A1C (HbA1C) measurements in claims data, we relied on SME input to set percentile-based risk thresholds.' The CDC prevalence comparison in Section 4 is post hoc because the 90th-percentile threshold was set by SME input before the comparison, so it does not break the circularity. The central transfer assumption in Section 2.2 ('assuming that undiagnosed individuals with similar covariate profiles may also be at risk') is an unverified premise rather than a circular step, but it is the only bridge from the constructed score to the T2DM-risk claim. There is no machine-checked or external outcome-based benchmark. Score 7 reflects that several 'predictions' reduce to the construction, while the score itself is not definitionally the target outcome.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central score rests on a data-dependent feature selection threshold, a hand-picked NMF rank, SME-chosen risk categories, and a transfer assumption from diagnosed to undiagnosed patients. No new physical or biological entities are introduced; the RWC is a methodological construct rather than an invented entity.

free parameters (3)
  • NMF component count k = 9
    Chosen from reconstruction error elbow and alignment with nine ICD-10-CM E11 subcategories (Sec 2.5.1); not determined by an external validation criterion.
  • Feature selection threshold = RWC >= mean(RWC) = 0.0384
    The threshold is derived from the NMF output and the observed sharp decline in RWC values (Sec 2.5); arbitrary and data-dependent.
  • Risk category percentile thresholds = 50th and 90th percentile
    Set in consultation with an SME, not from outcome data (Sec 2.8); affects reported high and moderate risk proportions.
assumptions (4)
  • domain assumption Claims-derived diagnosis codes (ICD-10 E11) and prescribed medication codes are reliable indicators of T2DM and its risk factors.
    Used throughout to define the positive cohort and covariates (Sec 2.2); no lab values such as HbA1C are available.
  • domain assumption Undiagnosed individuals with covariate profiles similar to diagnosed T2DM patients before their first diagnosis are at elevated risk of T2DM.
    Core transfer assumption in Sec 2.2; it is the basis for transferring patterns from diagnosed to undiagnosed patients.
  • domain assumption NMF components learned from diagnosed patients transfer to undiagnosed patients and capture clinically meaningful comorbidity patterns.
    The method applies the factorization learned on diagnosed patients to score all patients; no evaluation on separate populations.
  • ad hoc to paper KL divergence weights are directionally associated with T2DM risk.
    Equation 6 treats any large divergence as informative, but KL is not directional: a feature less prevalent in diagnosed patients (e.g., cervicalgia, 6.36% vs 11.19%, Table 1) also yields a positive weight and raises risk scores.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Unsupervised Latent Pattern Analysis for Estimating Type 2 Diabetes Risk in Undiagnosed Populations." pith.science (2026). https://pith.science/paper/3ASXU4Z2

@misc{pith2026250521824,
  author       = {Pith},
  title        = {Pith review of: Unsupervised Latent Pattern Analysis for Estimating Type 2 Diabetes Risk in Undiagnosed Populations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3ASXU4Z2}},
  note         = {Machine review of arXiv:2505.21824}
}
abstract

The global prevalence of diabetes, particularly type 2 diabetes mellitus (T2DM), is rapidly increasing, posing significant health and economic challenges. T2DM not only disrupts blood glucose regulation but also damages vital organs such as the heart, kidneys, eyes, nerves, and blood vessels, leading to substantial morbidity and mortality. In the US alone, the economic burden of diagnosed diabetes exceeded \$400 billion in 2022. Early detection of individuals at risk is critical to mitigating these impacts. While machine learning approaches for T2DM prediction are increasingly adopted, many rely on supervised learning, which is often limited by the lack of confirmed negative cases. To address this limitation, we propose a novel unsupervised framework that integrates Non-negative Matrix Factorization (NMF) with statistical techniques to identify individuals at risk of developing T2DM. Our method identifies latent patterns of multimorbidity and polypharmacy among diagnosed T2DM patients and applies these patterns to estimate the T2DM risk in undiagnosed individuals. By leveraging data-driven insights from comorbidity and medication usage, our approach provides an interpretable and scalable solution that can assist healthcare providers in implementing timely interventions, ultimately improving patient outcomes and potentially reducing the future health and economic burden of T2DM.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

43 extracted references · 42 canonical work pages

  1. [1]

    , Damaskos , C

    barticle Farmaki , P. , Damaskos , C. , Garmpis , N. , Garmpi , A. , Savvanis , S. , Diamantis , E. : Complications of the type 2 diabetes mellitus . Current cardiology reviews 16 ( 4 ), 249 -- 251 ( 2020 ) barticle

    work page 2020

  2. [2]

    : Diabetes overview

    botherref (WHO) , W.H.O. : Diabetes overview. Available at https://www.who.int/news-room/fact-sheets/detail/diabetes. Accessed April 11, 2025 botherref

    work page 2025

  3. [3]

    : Diabetes facts and figures

    botherref (IDF) , T.I.D.F. : Diabetes facts and figures. Available at https://idf.org/about-diabetes/diabetes-facts-figures/. Accessed April 11, 2025 botherref

    work page 2025

  4. [4]

    , (CDC) , P

    botherref Disease Control , C. , (CDC) , P. : About Underlying Cause of Death, 2018-2022. Available at https://wonder.cdc.gov/controller/saved/D158/D389F360. Accessed April 11, 2025 botherref

    work page 2018

  5. [5]

    , Prevention : National Diabetes Statistics Report

    botherref Disease Control , C. , Prevention : National Diabetes Statistics Report. Available at https://www.cdc.gov/diabetes/php/data-research/index.html. Accessed April 11, 2025 botherref

    work page 2025

  6. [6]

    , Lin , J

    barticle Parker , E.D. , Lin , J. , Mahoney , T. , Ume , N. , Yang , G. , Gabbay , R.A. , ElSayed , N.A. , Bannuru , R.R. : Economic costs of diabetes in the us in 2022 . Diabetes care 47 ( 1 ), 26 -- 43 ( 2024 ) barticle

    work page 2022

  7. [7]

    , Qu , K

    barticle Zou , Q. , Qu , K. , Luo , Y. , Yin , D. , Ju , Y. , Tang , H. : Predicting diabetes mellitus with machine learning techniques . Frontiers in genetics 9 , 515 ( 2018 ) barticle

    work page 2018

  8. [8]

    : Random forests

    barticle Breiman , L. : Random forests . Machine learning 45 , 5 -- 32 ( 2001 ) barticle

    work page 2001

Show all 43 references
  1. [9]

    , Hinton , G.E

    barticle Rumelhart , D.E. , Hinton , G.E. , Williams , R.J. : Learning representations by back-propagating errors . nature 323 ( 6088 ), 533 -- 536 ( 1986 ) barticle

    1986

  2. [10]

    : Induction of decision trees

    barticle Quinlan , J.R. : Induction of decision trees . Machine learning 1 , 81 -- 106 ( 1986 ) barticle

    1986

  3. [11]

    , Huang , H

    barticle Lai , H. , Huang , H. , Keshavjee , K. , Guergachi , A. , Gao , X. : Predictive models for diabetes mellitus using machine learning techniques . BMC endocrine disorders 19 , 1 -- 9 ( 2019 ) barticle

    2019

  4. [12]

    : Greedy function approximation: a gradient boosting machine

    botherref Friedman , J.H. : Greedy function approximation: a gradient boosting machine. Annals of statistics, 1189--1232 (2001) botherref

    2001

  5. [13]

    : The regression analysis of binary sequences

    barticle Cox , D.R. : The regression analysis of binary sequences . Journal of the Royal Statistical Society Series B: Statistical Methodology 20 ( 2 ), 215 -- 232 ( 1958 ) barticle

    1958

  6. [14]

    , Kocbek , P

    barticle Kopitar , L. , Kocbek , P. , Cilar , L. , Sheikh , A. , Stiglic , G. : Early detection of type 2 diabetes mellitus using machine learning-based prediction models . Scientific reports 10 ( 1 ), 11981 ( 2020 ) barticle

    2020

  7. [15]

    , Hastie , T

    barticle Friedman , J.H. , Hastie , T. , Tibshirani , R. : Regularization paths for generalized linear models via coordinate descent . Journal of statistical software 33 , 1 -- 22 ( 2010 ) barticle

    2010

  8. [16]

    , Guestrin , C

    bchapter Chen , T. , Guestrin , C. : Xgboost: A scalable tree boosting system . In: Proceedings of the 22nd Acm Sigkdd International Conference on Knowledge Discovery and Data Mining , pp. 785 -- 794 ( 2016 ) bchapter

    2016

  9. [17]

    , Meng , Q

    botherref Ke , G. , Meng , Q. , Finley , T. , Wang , T. , Chen , W. , Ma , W. , Ye , Q. , Liu , T.-Y. : Lightgbm: A highly efficient gradient boosting decision tree. Advances in neural information processing systems 30 (2017) botherref

    2017

  10. [18]

    , Hart , P

    barticle Cover , T. , Hart , P. : Nearest neighbor pattern classification . IEEE transactions on information theory 13 ( 1 ), 21 -- 27 ( 1967 ) barticle

    1967

  11. [19]

    , Friedman , J

    bbook Breiman , L. , Friedman , J. , Olshen , R.A. , Stone , C.J. : Classification and Regression Trees . Routledge , ??? ( 2017 ) bbook

    2017

  12. [20]

    , Vapnik , V

    barticle Cortes , C. , Vapnik , V. : Support-vector networks . Machine learning 20 , 273 -- 297 ( 1995 ) barticle

    1995

  13. [21]

    , Chen , Y

    bchapter Mani , S. , Chen , Y. , Elasy , T. , Clayton , W. , Denny , J. : Type 2 diabetes risk forecasting from emr data using machine learning . In: AMIA Annual Symposium Proceedings , vol. 2012 , p. 606 ( 2012 ) bchapter

    2012

  14. [22]

    , Blecker , S

    barticle Razavian , N. , Blecker , S. , Schmidt , A.M. , Smith-McLallen , A. , Nigam , S. , Sontag , D. : Population-level prediction of type 2 diabetes from claims data and analysis of risk factors . Big Data 3 ( 4 ), 277 -- 287 ( 2015 ) barticle

    2015

  15. [23]

    , Harish , V

    barticle Ravaut , M. , Harish , V. , Sadeghi , H. , Leung , K.K. , Volkovs , M. , Kornas , K. , Watson , T. , Poutanen , T. , Rosella , L.C. : Development and validation of a machine learning model using administrative health data to predict onset of type 2 diabetes . JAMA net...

    2021

  16. [24]

    , Grauw , W.J

    barticle Klein Woolthuis , E.P. , Grauw , W.J. , Gerwen , W.H. , Hoogen , H.J. , Lisdonk , E.H. , Metsemakers , J.F. , Weel , C. : Identifying people at risk for undiagnosed type 2 diabetes using the gp's electronic medical record . Family practice 24 ( 3 ), 230 -- 236 ( 2007 ...

    2007

  17. [25]

    , Healy , J

    botherref McInnes , L. , Healy , J. , Melville , J. : Umap: Uniform manifold approximation and projection for dimension reduction. arXiv preprint arXiv:1802.03426 (2018) botherref

    2018 arXiv

  18. [26]

    , Kriegel , H.-P

    bchapter Ester , M. , Kriegel , H.-P. , Sander , J. , Xu , X. , : A density-based algorithm for discovering clusters in large spatial databases with noise . In: Kdd , vol. 96 , pp. 226 -- 231 ( 1996 ) bchapter

    1996

  19. [27]

    , Sarkar , J

    barticle Bej , S. , Sarkar , J. , Biswas , S. , Mitra , P. , Chakrabarti , P. , Wolkenhauer , O. : Identification and epidemiological characterization of type-2 diabetes sub-population using an unsupervised machine learning approach . Nutrition & Diabetes 12 ( 1 ), 27 ( 2022 )...

    2022

  20. [28]

    , Makadia , R

    barticle Voss , E.A. , Makadia , R. , Matcho , A. , Ma , Q. , Knoll , C. , Schuemie , M. , DeFalco , F.J. , Londhe , A. , Zhu , V. , Ryan , P.B. : Feasibility and utility of applications of the common data model to multiple, disparate observational health databases . Journal o...

    2015

  21. [29]

    , Verleysen , M

    barticle Fr \'e nay , B. , Verleysen , M. : Classification in the presence of label noise: a survey . IEEE transactions on neural networks and learning systems 25 ( 5 ), 845 -- 869 ( 2013 ) barticle

    2013

  22. [30]

    , Lambert , C.G

    barticle Kumar , P. , Lambert , C.G. : Positive unlabeled learning selected not at random (pulsnar): class proportion estimation without the selected completely at random assumption . PeerJ Computer Science 10 , 2451 ( 2024 ) barticle

    2024

  23. [31]

    , Elkan , C

    bchapter Jaskie , K. , Elkan , C. , Spanias , A. : A modified logistic regression for positive and unlabeled learning . In: 2019 53rd Asilomar Conference on Signals, Systems, and Computers , pp. 2007 -- 2011 ( 2019 ). IEEE bchapter

    2019

  24. [32]

    , Seung , H.S

    barticle Lee , D.D. , Seung , H.S. : Learning the parts of objects by non-negative matrix factorization . nature 401 ( 6755 ), 788 -- 791 ( 1999 ) barticle

    1999

  25. [33]

    , Leibler , R.A

    barticle Kullback , S. , Leibler , R.A. : On information and sufficiency . The annals of mathematical statistics 22 ( 1 ), 79 -- 86 ( 1951 ) barticle

    1951

  26. [34]

    Available at https://www.icd10data.com/ICD10CM/Codes/E00-E89/E08-E13/E11-

    botherref ICDData : Type 2 diabetes mellitus E11. Available at https://www.icd10data.com/ICD10CM/Codes/E00-E89/E08-E13/E11-. Accessed April 11, 2025 botherref

    2025

  27. [35]

    , Moomtaheen , F

    botherref Kumar , P. , Moomtaheen , F. , Malec , S.A. , Yang , J.J. , Bologa , C.G. , Schneider , K.A. , Zhu , Y. , Tohen , M. , Villarreal , G. , Perkins , D.J. , et al.: Detecting opioid use disorder in health claims data with positive unlabeled learning. IEEE Journal of Bio...

    2024

  28. [36]

    , Nestsiarovich , A

    barticle Kumar , P. , Nestsiarovich , A. , Nelson , S.J. , Kerner , B. , Perkins , D.J. , Lambert , C.G. : Imputation and characterization of uncoded self-harm in major mental illness using machine learning . Journal of the American Medical Informatics Association 27 ( 1 ), 13...

    2020

  29. [37]

    , Kent , D.M

    barticle Lin , P.-J. , Kent , D.M. , Winn , A. , Cohen , J.T. , Neumann , P.J. , : Multiple chronic conditions in type 2 diabetes mellitus: prevalence and consequences . Am J Manag Care 21 ( 1 ), 23 -- 34 ( 2015 ) barticle

    2015

  30. [38]

    , Carrasco , A

    barticle Luque , A. , Carrasco , A. , Mart \' n , A. , Las Heras , A. : The impact of class imbalance in classification performance metrics based on the binary confusion matrix . Pattern Recognition 91 , 216 -- 231 ( 2019 ) barticle

    2019

  31. [39]

    , Kim , H.Y

    barticle Choi , H.S. , Kim , H.Y. , Han , K.-D. , Jung , J.-H. , Kim , C.S. , Bae , E.H. , Ma , S.K. , Kim , S.W. : Obstructive sleep apnea as a risk factor for incident end stage renal disease: a nationwide population-based cohort study from korea . Clinical and experimental ...

    2019

  32. [40]

    , Brenner , B.M

    barticle Keane , W.F. , Brenner , B.M. , De Zeeuw , D. , Grunfeld , J.-P. , McGill , J. , Mitch , W.E. , Ribeiro , A.B. , Shahinfar , S. , Simpson , R.L. , Snapinn , S.M. , : The risk of developing end-stage renal disease in patients with type 2 diabetes and nephropathy: the r...

    2003

  33. [41]

    , Lieber , A

    barticle Smulyan , H. , Lieber , A. , Safar , M.E. : Hypertension, diabetes type ii, and their association: role of arterial stiffness . American journal of hypertension 29 ( 1 ), 5 -- 13 ( 2016 ) barticle

    2016

  34. [42]

    , Brunzell , J.D

    barticle Carr , M.C. , Brunzell , J.D. : Abdominal obesity and dyslipidemia in the metabolic syndrome: importance of type 2 diabetes and familial combined hyperlipidemia in coronary artery disease risk . The journal of clinical endocrinology & metabolism 89 ( 6 ), 2601 -- 2607...

    2004

  35. [43]

    , Ren , H

    botherref Luo , J.-Q. , Ren , H. , Chen , M.-Y. , Zhao , Q. , Yang , N. , Liu , Q. , Gao , Y.-C. , Zhou , H.-H. , Huang , W.-H. , Zhang , W. : Hydrochlorothiazide-induced glucose metabolism disorder is mediated by the gut microbiota via lps-tlr4-related macrophage polarization...

    2023

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.