Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “diagnostic performance”

Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 145 records · Page 8Linked to original sources
PubMed1979

Aids to the evaluation of diagnostic performance, as applied to immunological creatine kinase-MB assay in myocardial infarction.

Immunoassay of creatine kinase-MB provides numerical information, which makes it possible to estimate quantitatively the diagnostic performance following myocardial infarction. Two graphical methods for such an evaluation are presented. The relationship between technical sensitivity and specificity was analyzed using a continuous function, the receiver-operator characteristic curve. Using this function, a diagnostic threshold ("upper normal limit") was chosen, which balances both technical specificity and sensitivity on the first day following infarction. On the second and third days this threshold caused a progressive loss of technical sensitivity. The relationship between effectiveness and the prevalence of myocardial infarction in the tested population was evaluated with a nomogram correlating these two quantities. With the chosen diagnostic threshold, effectiveness is independent of prevalence on the first day, but the loss in technical sensitivity on subsequent days causes effectiveness to decay when the prevalence is high.

Creatine Kinase↗
PubMedDate not supplied

An overview of the College of American Pathologists' programs in surgical pathology and cytopathology. Data summary of diagnostic performance in cervical cytopathology.

The survey programs in surgical pathology and cytopathology of the College of American Pathologists are described, along with a summary of the data produced on diagnostic performance in cytopathology. The levels of agreement between individual diagnoses and consensus diagnoses were highest for slides in the "negative" and "positive" categories and somewhat less for slides in the "suspicious" categories. Submission of slides with 70% or less agreement and slides on which a consensus had not been reached to a new group of observers yielded similar levels of agreement, indicating that there is a fairly small percentage of cervical smears that are diagnostic problems. Agreement levels, especially in relation to false negatives and false positives, represent a commendable achievement, especially considering that cervical cytology is a screening procedure and that "suspicious" smears result in further investigations. It is also important to recognize that additional clinical data and studies are often taken into consideration before definitive treatment is begun or additional follow-up smears are taken.

Cytodiagnosis↗
PubMed2005

Diagnostic performance of serial haematocrit measurements in identifying major injury in adult trauma patients.

UNLABELLED: Determination of occult haemorrhage is an essential part of trauma assessment. We evaluated the diagnostic utility of decreasing haematocrit (DeltaHct) in detecting major injury. Additionally, we tested the correlation between the volume of infused intravenous fluid (IVF) and DeltaHct. METHODS: Prospective observational study at a level one trauma centre. INCLUSION CRITERIA: Patients with suspected major injury. exclusion criteria: Patients who received blood transfusion in the first 4 h, and those who deceased or were transferred to other units before the completion of the observation period (4 h). We measured IVF and DeltaHct at 4 h after triage. We classified patients as having minor or major injury on the basis of injury severity score > or =15. Receiver Operating Characteristic (ROC) curve was used to test the diagnostic performance of DeltaHct in identifying major injury. We tested the operating characteristics of DeltaHct cut-off values of 5 and 10 in detecting major injury. We also measured the correlation of IVF and DeltaHct in a subgroup of patients with low potential for blood loss (ISS<3) to account for possibility of haemodilution. RESULTS: Four hundred and ninety-four patients (convenience sample) were enrolled (age 36+/-17 years, 82% male, 57% blunt trauma). Sixty-three patients (13%) had major injury. The area under the ROC curve for DeltaHct was not significantly different from the unity line (p=0.20). DeltaHct-4 h>5 points had a sensitivity of 40% (95% CI, 29-52%), specificity of 94% (95% CI, 92-96%), likelihood ratio for a positive test (LR+) of 7.1 (95% CI, 4.4-11.7), and likelihood ratio for a negative test (-LR) of 0.64 (95% CI, 0.52-0.78) in identifying major trauma. DeltaHct-4 h>10 points had sensitivity of 16% (95% CI, 9-27%), specificity of 95% (95% CI, 92-0.96%), +LR of 3.0 (95% CI, 1.5-5.9), and -LR of 0.89 (95% CI, 0.80-0.99). In our subgroup analysis, we detected no significant correlation (p=0.09) between the IVF and DeltaHct-4 h. CONCLUSIONS: DeltaHct-4 h>5 or 10 points is suggestive of major injury (high specificity and +LR). However, the failure to drop the Hct cannot be used to rule out major injury (low sensitivity and -LR).

Adult↗
PubMed1999

Impact of menstrual cycle on the diagnostic performance of LCR, TMA, and PCE for detection of Chlamydia trachomatis in home obtained and mailed vaginal flush and urine samples.

OBJECTIVES: To assess the impact of the menstrual cycle on the diagnostic performance of various assays for detection of Chlamydia trachomatis in home obtained and mailed vaginal flush and urine specimens. METHODS: A ligase chain reaction assay (LCR; Abbott Laboratories), a transcription mediated amplification assay (TMA; Gen-Probe), and an enzyme amplified immunoassay (PCE; Dako Diagnostics) were evaluated for their validity in detecting C trachomatis in vaginal flush, first void urine, and midstream urine specimens obtained by female high school students at home and mailed directly to the diagnostic laboratory. RESULTS: C trachomatis was detected in 45 of 889 females (5.1%). The vaginal flush material was positive by TMA and LCR in 84% and 82% of the chlamydia positive females, respectively. First void urine was positive by TMA in 73% and by LCR in 49% of the cases. Midstream urine was positive by TMA and LCR in 69% and 42% of the females, respectively. On a pool of first void and midstream urine, PCE detected 49% of the chlamydia positive females. The overall prevalence of C trachomatis increased with increasing time after the last menstrual bleeding. In urine samples, but not vaginal flush specimens, obtained 3 weeks after the last menstrual bleeding, the sensitivities of TMA, LCR, and PCE decreased markedly suggesting that inhibitors to the assays are excreted in the urine but not in vaginal secretions at this time. CONCLUSION: Vaginal flush samples are superior to urines for detection of chlamydia infections in females. In screening of young asymptomatic females, samples should be obtained in the latter part of the menstrual cycle.

Adolescent↗
PubMed1997

Initial evaluation of CYFRA 21-1 diagnostic performances as a urinary marker in bladder transitional cell carcinoma.

PURPOSE: CYFRA 21-1, an immunoradiometric assay developed for the detection of a soluble cytokeratin 19 fragment, is evaluated for its diagnostic performance in urine of patients with transitional cell carcinoma. MATERIALS AND METHODS: CYFRA 21-1 was investigated in serum and urine of 128 patients, including 48 with bladder transitional cell carcinoma (group 1), 44 with other urological pathological conditions (group 2) and 36 free of urothelial disease (group 3). Urinary cytopathology was also performed. RESULTS: Mean urinary CYFRA was 123.5 +/- 53, 11.9 +/- 4.8 and 2.3 +/- 0.2 ng./ml. for groups 1 to 3, respectively, and was significantly different. From the receiver operating characteristics curve, the optimal combination of 96% sensitivity and 74% specificity was determined for a threshold value of 4 ng./ml. while overall cytopathology sensitivity was 43%. CONCLUSIONS: Urinary CYFRA 21-1 may be a useful marker for diagnosing transitional cell carcinoma.

Aged↗
PubMed1996

Detection of Chlamydia trachomatis infections in women by Amplicor PCR: comparison of diagnostic performance with urine and cervical specimens.

We used the Roche Amplicor PCR assay to compare urine and cervical swabs as sample material in the detection of Chlamydia trachomatis causing genital infections. The diagnostic performance of Amplicor PCR was compared with that of cell culture and the Gen-Probe PACE 2 assay with cervical specimens. If discrepant from other results, the specimens negative by PCR were diluted and reanalyzed to reveal PCR inhibitors. Of 666 patients, 39 (5.9%) were confirmed to have chlamydial infection. The respective sensitivity and specificity of Amplicor PCR were as follows: urine specimens, 82.0 and 99.7%; cervical specimens, 82.0 and 99.8%. Those for cell culture with cervical specimens were 84.6 and 100%. For the Gen-Probe PACE 2 assay, the sensitivity and specificity with cervical specimens were 79.5 and 100%, respectively. Without the effect of PCR inhibitors, the sensitivity of PCR with urine would have been 97.4%. Provided that the problems currently caused by inhibitors will be solved, the Amplicor PCR assay with urine specimens offers a tempting alternative for the diagnosis of C. trachomatis infection in women.

Bacteriological Techniques↗
PubMed1998

Comparison of the diagnostic performance of two rapid bedside biochemical assays in the early detection of acute myocardial infarction.

BACKGROUND: New biochemical markers have been recently proposed as potential aids in the diagnosis of myocardial infarction. Assessment of these markers in real time by near-patient testing would allow more rapid decisions regarding critical therapeutic interventions. HYPOTHESIS: The diagnostic performance of two qualitative, whole blood assays for measurement of myoglobin/creatine kinase (CK)-MB mass (CARDIAC STATus) and cardiac troponin T (TropT) was compared. Their potential in optimizing the rule-in of patients with infarction in order to initiate revascularization therapies was also evaluated. METHODS: A group of 101 patients admitted to the hospital within 12 h from onset of symptoms suggestive of myocardial infarction was studied. A first blood sample was drawn in all patients immediately after admission. In patients not receiving reperfusion therapy (n = 61), three additional samples were also obtained 2, 4, and 6 h later. RESULTS: CARDIAC STATus showed good performance in patients with an admission time > 2 h from onset of pain (diagnostic efficiency, 71-100%). Conversely, TropT was not sensitive enough for acceptable efficiency. In a subset of 39 patients admitted within 6 h who did not meet standard enrollment criteria for thrombolysis, CARDIAC STATus allowed early detection of almost 50% of infarcts, enabling a significant implementation of alternative revascularization therapies. CONCLUSIONS: In patients admitted to the coronary care unit on the basis of case history and clinical examination, CARDIAC STATus assay is more effective than TropT in the early diagnosis of infarction, providing a valuable aid to cardiologists for confirming a suspected infarction rapidly, thus allowing, if indicated, recourse to revascularization therapy in clinically uncertain cases.

Biomarkers↗
PubMed1996

Diagnostic performance of a receptor-binding radiopharmacokinetic model.

UNLABELLED: The aim of this study was to determine which measurement obtained from a radiopharmacokinetic model of a receptor-binding radiotracer provides the highest diagnostic performance for the detection of diffuse hepatocellular disease. METHODS: Twenty-seven healthy subjects and 46 patients with diffuse hepatocellular disease were studied with the receptor-binding radiopharmaceutical, 99mTc-galactosyl-neoglycoalbumin. A radiopharmacokinetic model was used to produce estimates of receptor concentration [R]o, the scaled forward-binding rate constant Kb, hepatic plasma volume, Vh, extrahepatic plasma volume, Ve and hepatic plasma flow, F. Receiver operating characteristic analysis of each model estimate was conducted. RESULTS: Receptor concentration [R]o and the metrics [R]o/tbw and kb[R]o[R]o/tbw provided the best discrimination between healthy and diseased liver. The forward-binding rate constant kb and the metrics F/Ve and Vh/tbw provided no discrimination. CONCLUSION: Based on simplicity and higher measurement precision, [R]o was selected as the most accurate index of hepatic function.

Albumins↗
PubMed1990

Diagnostic performance indices for immunofluorescent tests and enzyme immunoassays of leishmaniasis sera from northern and north-eastern Brazil.

A total of 341 sera were screened for anti-Leishmania IgA, IgG, and IgM antibodies by immunofluorescent (IF) tests and enzyme immunoassay (ELISA). Altogether, 292 of the sera originated from patients with clinically as well as parasitologically diagnosed (positive lesion imprint or the Montenegro skin test) cutaneous leishmaniasis; 49 of the sera were from controls from the same base population. In terms of diagnostic performance, the ELISAs for IgG and IgM yielded indices of diagnostic utility, and the positive predictive value for the IgG-ELISA was 94.6%. A remarkably high specificity (100%) was obtained with the IgA-IF test, but its sensitivity was very low.

Adolescent↗
PubMed2005

Diagnostic performance of 18F-FDG PET and ictal 99mTc-HMPAO SPET in pediatric temporal lobe epilepsy: quantitative analysis by statistical parametric mapping, statistical probabilistic anatomical map, and subtraction ictal SPET.

We investigated the diagnostic performance of 18F-FDG PET and ictal (99m)Tc-HMPAO SPET in pediatric temporal lobe epilepsy (TLE). Twenty-one pediatric TLE patients were enrolled in this study. Their diagnoses were confirmed by histology and post-surgical outcome (Engel class I or II). The patients' ages were 18 or younger (15+/-3 years). Of the 21 patients, 21 patients underwent 18F-FDG PET scan and 15 underwent ictal (99m)Tc-HMPAO SPET. Preoperative PET and/or ictal SPET images were reviewed by simple visual assessment and by statistical parametric mapping (SPM). Asymmetric indices (AI) were calculated using statistical probabilistic anatomical map (SPAM) on 18F-FDG PET. In nine patients who underwent both ictal and interictal SPET, SISCOM (subtraction ictal SPET coregistered to MR template) was performed. PET correctly localized epileptogenic zones in 20 of 21 (95%) by visual assessment. SPM analysis of PET correctly localized epileptogenic zones in 18 of 21 (86%). Ictal SPET correctly localized epileptogenic zones in 12 of 15 (80%) by visual assessment. SPM analysis of ictal SPET correctly localized epileptogenic zones in 12 of 15 (80%). SISCOM correctly localized 8 of 9 (89%), which was equal to that of visual assessment of ictal SPET. The AIs of the temporal lobes by PET were -15+/-8.4 in the left and 9.9+/-8.9 in the right TLE (normal control: -2.9+/-2.8), and correctly localized epileptogenic zones in all cases. As is found in adult TLE, PET and ictal SPET efficiently localized epileptogenic zones in pediatric TLE. SPM analysis of PET or ictal SPET could be used as an aid to visual assessment. Moreover, SISCOM was equal visual assessment of ictal SPET images in terms of lesion localizations.

Adolescent↗
PubMed1992

Diagnostic peritoneal lavage: integration with clinical information to improve diagnostic performance.

Management of abdominal trauma requires both the detection of injuries sustained and an ability to distinguish patients who require operative repair from those who do not. In this prospective study of 200 patients receiving diagnostic peritoneal lavage (DPL) following blunt trauma, relationships among DPL result, clinical features (information from initial patient assessment), and laparotomy outcome were investigated. The DPL result alone predicted requirement for laparotomy with an accuracy of 93%, a specificity of 96%, a sensitivity of 85%, a positive predictive value (PV-Positive) of 87%, and a negative predictive value (PV-Negative) of 95%. Combining clinical features with the DPL result reduced the number of unnecessary laparotomies (increased PV-Positive and specificity), but increased the number of missed necessary laparotomies (decreased PV-Negative and sensitivity). The best diagnostic performance was found by combining the DPL result with circulatory status, which, in this series of patients, predicted necessary laparotomy with an accuracy of 95%, a specificity of 99%, a sensitivity of 81%, a PV-Positive of 98%, and a PV-Negative of 94%.

Abbreviated Injury Scale↗
PubMed1993

The clinicopathological meeting. A means of auditing diagnostic performance.

The clinicopathological meeting is one of the major areas of contact between clinician and pathologist, and the review of histology reports at such meetings can be regarded as an aspect of quality control for diagnostic pathology. To assess the effects of this review, we have conducted a formal analysis of clinicopathological meetings in the Southampton University Hospitals. Over a 14-week period, 416 cases were reviewed at 58 meetings. Seven categories of meeting were involved; Breast; Dermatology; Ear, Nose, and Throat (ENT) and Oral Surgery; Gastroenterology; General Surgery; Oncology; and Urology. A proforma was completed for each case that was reviewed, evaluating changes in diagnosis, the reason for such changes, and the effect, if any, on patient management. Participating consultants, both clinicians and pathologists, were questioned regarding the aims of their particular meeting, and how successful they felt the meeting to be. Histological review resulted in an altered diagnosis in 9% of cases, a refined diagnosis in 10% of cases, and no diagnostic change in 81%. Most (88.1%) of the diagnostic changes were attributable to the specialist expertise of the reviewing pathologist, and only 4.8% resulted from the clinicians providing extra information. Amended diagnoses led to major management changes in 16 cases (3.8%), minor management changes in 12 cases (2.9%), and no management change in 388 cases (93.3%). The Gastroenterology meeting showed the highest figures for clinically significant diagnostic changes; there was an amended diagnosis in 44.2% of cases reviewed, 83.3% of which altered patient management. These figures were related to the specialist review of endoscopic biopsy specimens. Although the variation in the methods used to select cases for review and the different roles of individual meetings prevent their adoption into a formal systematic audit scheme, analysis of case review at clinicopathological meetings can yield valuable information concerning departmental diagnostic performance.

Dermatology↗
PubMed1995

[Tuberculosis presenting as diffuse pulmonary infiltrates in AIDS patients: diagnostic performance of clinical samples].

BACKGROUND: In patients with human immunodeficiency virus (HIV) infection, tuberculosis is frequently presented with diffuse pulmonary infiltrates which are indistinguishable from those caused by other respiratory pathogens. It is therefore useful to know the diagnostic performance of different clinical samples. METHODS: We have retrospectively analyzed the clinical histories of 56 patients seen over a 3-year period. All the patients had HIV infection, Mycobacterium tuberculosis isolated in at least one clinical sample and presented with diffuse bilateral infiltrates in thorax radiography. The results of all the clinical samples submitted to the microbiology laboratory. RESULTS: The highest performance in both stainings and cultures were obtained from the biopsy (or aspirate) of adenopathies (93 and 100%, respectively), sputum (57 and 88%) and urine (31 and 64%). A lower than expected sensitivity was obtained in the fibrobronchoscopy samples (bronchoalveolar lavage and transbronchial biopsy). The staining had low sensitivity for predicting positive cultures in all the samples except in the adenopathies. Visualization of granulomas in transbronchial biopsies and bone marrow was more sensitive for diagnosis than the demonstration of resistant acid-alcohol bacilli in the same samples. Globally, rapid diagnosis was obtained in 43 patients (76%). The remaining 13 (24%) patients were not diagnosed until the culture results had been received despite the adequate use of diagnostic procedures. CONCLUSIONS: These findings support the use of empiric treatment when tuberculosis is suspected despite initial negativity of the microbiologic and pathologic examinations following the discarding of other potential causes.

AIDS-Related Opportunistic Infections↗
PubMed2006

The diagnostic performance of chronologic age in the assessment of skeletal maturity.

OBJECTIVE: The aim of this study was to analyze the relationship between chronologic age the and individual skeletal maturity as assessed by means of the cervical vertebral maturation (CVM) method during the circumpubertal period. MATERIALS AND METHODS: The evaluated sample of 600 subjects consisted of 100 subjects (50 males and 50 females) for each of 6 age groups, from 9 years through 14 years of age. Individual skeletal maturity for all subjects was determined by using the CVM method. The relationship between chronologic age and the most prevalent CVM stage at each age group was evaluated statistically by means of indicators of diagnostic test performance that specify the ability of a diagnostic test to identify a condition. RESULTS: The diagnostic performance of chronologic age for the detection of the onset of the adolescent peak in skeletal maturation was very low both in males and in females. In male subjects, the chronologic age of 9 years +/- 6 months presented with strong diagnostic power for the identification of a pre-pubertal stage in skeletal maturation. In female subjects, the chronologic age of 14 years +/- 6 months corresponded with a strong probability of a postpubertal stage in skeletal maturation. CONCLUSIONS: In males, chronologic age can identify a pre-pubertal stage of skeletal development, and in females a post-pubertal stage. In both males and females, chronologic age cannot recognize the onset of the adolescent peak in skeletal maturation.

Adolescent↗
PubMed2005

Multilevel examination of minor salivary gland biopsy for Sjogren's syndrome significantly improves diagnostic performance of AECG classification criteria.

The recently observed low reproducibility of focus score (FS) assessment at different section depths in a series of single minor salivary gland biopsies highlighted the need for a standardized protocol of extensive histopathological examination of such biopsies in Sjogren's syndrome. For this purpose, a cumulative focus score (cFS) was evaluated on three slides cut at 200-mum intervals from each of a series of 120 salivary biopsies. The cFS was substituted for the baseline FS in the American-European Consensus Group (AECG) criteria set for Sjogren's syndrome classification, and then test specificity and sensitivity were assessed against clinical patient re-evaluation. Test performances of the AECG classification with the original FS and the score obtained after multilevel examination were statistically compared using receiver operating characteristic (ROC) curve analysis. The diagnostic performance of AECG classification significantly improved when the cFS was entered in the AECG classification; the improvement was mostly due to increased specificity in biopsies with a baseline FS >or= 1 but <2. The assessment of a cFS obtained at three different section levels on minor salivary gland biopsies can be useful especially in biopsies with baseline FSs between 1 and 2.

Adult↗
PubMed1991

Improved diagnostic performance of the exercise ECG test by computerized multivariate ST-segment/heart rate analysis.

The exercise ECG test is the most widely used noninvasive method of evaluating myocardial ischemia. To determine whether it is possible to enhance its diagnostic power, the authors carried out computerized ECG measurements on 118 nonischemic patients and 38 ischemic patients with or without myocardial infarctions (MIs). Forty-seven (all ischemic and nine normal cases) underwent T1-201 SPECT. The diagnostic variables include ST-segment deviation, modified ST/HR-slope, ST-segment shape, and maximum heart rate. These variables are interrelated by means of compartmental and specific decision rules by computer. The diagnostic method discussed in this article is called the multivariate ST/HR analysis (MUSTA). When compared with MUSTA, the T1-201 SPECT images detected myocardial ischemia with a sensitivity of 100% and a specificity of 89%. Its diagnostic performance was influenced by neither previous MIs nor cardiac medication of the 47 patients tested. MUSTA was also compared to the nonmodified ST/HR analyses (ST60/HR and ST80/HR) and the standard exercise ECG test. It performed significantly better than these methods. The authors conclude that the multivariate ST/HR analysis is comparable to T1-201 SPECT in diagnosing myocardial ischemia among the study subjects. Furthermore, MUSTA is executed during the noninvasive exercise ECG test and is easily applied using an IBM/AT-compatible microcomputer. However, further evaluation of MUSTA with a separate and unselected patient population is needed.

Adult↗
PubMed1997

Diagnostic performance characteristics of architectural features revealed by high spatial-resolution MR imaging of the breast.

OBJECTIVE: Our objective was twofold: to determine which architectural features revealed by high spatial-resolution MR imaging of the breast contribute to diagnostic accuracy and to evaluate the diagnostic performance characteristics of those architectural features. MATERIALS AND METHODS: Eligible patients with suspicious mammographic or palpable findings or both underwent MR imaging. Ninety-three patients whose MR images revealed lesions that corresponded to the mammographically visible or palpable findings were included in the study. Patients were examined with sagittal T1-weighted spin-echo MR imaging, fat-saturated T2-weighted fast spin-echo MR imaging, and dynamically enhanced fat-saturated fast gradient-echo MR imaging. All patients underwent subsequent excisional biopsy or cyst aspiration. Lesions were identified initially by an experienced radiologist who was aware of the patient's clinical or mammographic information. Two radiologists who were unaware of the patients' histories and who had less experience in MR imaging of the breast then independently evaluated each lesion for the architectural-features and predicted each lesion's potential for malignancy. RESULTS: Architectural features that were highly predictive of benign disease included smooth or lobulated borders (97-100%), the absence of mass enhancement (100%), and enhancement that was less than the enhancement of surrounding breast fibroglandular tissue (93-100%). Nonenhancing internal septations were specific for the diagnosis of fibroadenoma. Architectural features that were highly predictive of malignant disease included spiculated borders (76-88%) and peripheral rim enhancement in the presence of central lesion enhancement (79-92%). CONCLUSION: Architectural features revealed by high spatial-resolution MR imaging of the breast can help distinguish benign from malignant disease.

Adult↗
PubMed2026

Diagnostic performance of machine learning models for malignant and non-malignant pleural effusion: Systematic review and meta-analysis.

BACKGROUND: Accurately distinguishing malignant pleural effusion (MPE) from non-malignant pleural effusion is clinically important, but the generalisability and methodological quality of machine-learning (ML) models remain uncertain. METHODS: We searched eight databases to 23 April 2026. Diagnostic performance was pooled using random-effects and Reitsma bivariate models, and study quality was assessed using PROBAST+AI. RESULTS: Forty-two studies were included; 17 contributed to the AUC meta-analysis and 14 to the bivariate analysis. The pooled AUC was 0.90 (95&#xa0;% CI 0.85-0.94; 95&#xa0;% prediction interval 0.62-0.98), with sensitivity of 0.80 (95&#xa0;% CI 0.77-0.83) and specificity of 0.87 (95&#xa0;% CI 0.79-0.92). Only nine studies reported external, temporal or independent validation. Externally validated studies had a lower pooled AUC than studies without external validation (0.83 vs 0.92), with lower specificity observed in the two externally validated studies contributing sensitivity and specificity data. All 42 development assessments had high overall quality concerns, and all 42 model evaluations were judged at high risk of bias. CONCLUSIONS: ML models showed good apparent accuracy for distinguishing MPE from non-MPE, but the evidence was limited by substantial heterogeneity, high risk of bias and scarce external validation. The pooled estimates reflect the average performance of different selected models rather than the expected accuracy of a single clinical test. ML models should be regarded as adjuncts to existing diagnostic pathways until they are confirmed by rigorous multicentre prospective external validation and clinical-impact studies.

Humans↗