Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “Prediction Algorithms”

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 361 records · Page 20Linked to original sources
PubMed2001

[Coronary risk appraisal for primary prevention of coronary heart disease in a community].

PURPOSE: In discussion on application of "National Health Promotion Toward 21st Century in Japan" in Kanagawa prefecture, it was noted that the age-adjusted mortality rate of death from ischemic heart disease in this part of Japan was higher than that for the whole nation in 1996. To facilitate development of a strategy for primary prevention of coronary heart disease (CHD), the present study was conducted to predict 2-yr incidence of CHD and decrease with simulations assuming improvement in CHD risk factors. METHODS: Using CHD risk prediction algorithm; the Weibull accelerated failure regression model based on the Framingham Heart Study, a 2-yr incidence of CHD was predicted for 1652 residents (515 male, 1137 female) on the basis of results of a health check up in 1998. We then estimated the probable decrease in CHD recalculated assuming decrease in total cholesterol (TC), increase in HDL-cholesterol (HDL-C), decrease in systolic blood pressure (SBP), or quitting the smoking habit. RESULTS: 1. The 2-yr probability of developing CHD for men free of heart disease was 2.79 +/- 2.17%, and that for men who had heart disease was 10.25 +/- 2.17%. The 2-yr probability for women free of heart disease was 16.80 +/- 14.40%, and that for women who had heart disease was 3.66 +/- 1.09%. As the reported probability of developing CHD in the U.S.A. is remarkably higher than in Japan, the fact that the present model was based on American data explains why these predicted probabilities are higher than values reported from Japanese cohort studies. 2. For men free of heart disease, a strategy for high risk case such as a decrease in TC and an increase in HDL-C, or quitting the smoking habit, was more effective than a population-based strategy. For women free of heart disease, the population-based strategy was more effective. 3. Women more than 60-yrs old who had a high 2-yr probability of developing CHD were divided into three groups; high, middle, and low risk. The mean body weight, mean body mass index, mean diastolic blood pressure, and mean blood glucose in the high risk group were significantly higher than the values in the other groups. Decrease in systolic blood pressure was a more effective strategy for decrease in CHD incidence in the high risk group than in the other groups. CONCLUSIONS: CHD risk prediction of this type may be considered useful for setting target CHD risk factors and for focusing interventions to prevent CHD effectively.

Adult↗
PubMedDate not supplied

Predicting antigenic determinants in proteins: looking for unidimensional solutions to a three-dimensional problem?

In a recent review, Hopp (Peptide Research 6:183-190, 1993) claimed that the Hopp and Woods hydrophilicity method for locating antigenic determinants is superior to all other existing methods for predicting the B cell epitopes of proteins but that it is not useful to aid the investigator in producing peptide-protein cross-reactive antisera. In this article, we challenge both these assertions. Most investigators utilize antigenicity prediction algorithms because they wish to produce anti-peptide antibodies capable of cross-reacting with the intact protein. All prediction methods are based on propensity scales for the 20 amino acids, which describe the tendency of each residue to be associated with properties such as hydrophilicity, surface accessibility or segmental mobility. When we compared the prediction efficacy of 22 different scales, taking into account both correct and incorrect predictions, we found that none of the scales gave a level of correct prediction higher than about 50%-60%. If no antigenicity was found in a particular region of the protein, we took the view that hydrophilicity peaks located in that region amounted to wrong predictions. The much higher success rate reported by Hopp for this method stems from the way he assesses prediction efficacy, i.e., by counting the number of known epitopes located inside and outside hydrophilicity peaks. Reasons for the low success rate of antigenicity prediction are discussed. In most cases, it is unrealistic to try to reduce the complexity of discontinuous, conformational epitopes to simple, linear peptide models.

Chemical Phenomena↗
PubMed1999

Evaluation and improvement of multiple sequence methods for protein secondary structure prediction.

A new dataset of 396 protein domains is developed and used to evaluate the performance of the protein secondary structure prediction algorithms DSC, PHD, NNSSP, and PREDATOR. The maximum theoretical Q3 accuracy for combination of these methods is shown to be 78%. A simple consensus prediction on the 396 domains, with automatically generated multiple sequence alignments gives an average Q3 prediction accuracy of 72.9%. This is a 1% improvement over PHD, which was the best single method evaluated. Segment Overlap Accuracy (SOV) is 75.4% for the consensus method on the 396-protein set. The secondary structure definition method DSSP defines 8 states, but these are reduced by most authors to 3 for prediction. Application of the different published 8- to 3-state reduction methods shows variation of over 3% on apparent prediction accuracy. This suggests that care should be taken to compare methods by the same reduction method. Two new sequence datasets (CB513 and CB251) are derived which are suitable for cross-validation of secondary structure prediction methods without artifacts due to internal homology. A fully automatic World Wide Web service that predicts protein secondary structure by a combination of methods is available via http://barton.ebi.ac.uk/.

Algorithms↗
PubMed2003

Stochastic modeling of RNA pseudoknotted structures: a grammatical approach.

MOTIVATION: Modeling RNA pseudoknotted structures remains challenging. Methods have previously been developed to model RNA stem-loops successfully using stochastic context-free grammars (SCFG) adapted from computational linguistics; however, the additional complexity of pseudoknots has made modeling them more difficult. Formally a context-sensitive grammar is required, which would impose a large increase in complexity. RESULTS: We introduce a new grammar modeling approach for RNA pseudoknotted structures based on parallel communicating grammar systems (PCGS). Our new approach can specify pseudoknotted structures, while avoiding context-sensitive rules, using a single CFG synchronized with a number of regular grammars. Technically, the stochastic version of the grammar model can be as simple as an SCFG. As with SCFG, the new approach permits automatic generation of a single-RNA structure prediction algorithm for each specified pseudoknotted structure model. This approach also makes it possible to develop full probabilistic models of pseudoknotted structures to allow the prediction of consensus structures by comparative analysis and structural homology recognition in database searches.

Algorithms↗
PubMed1976

An assessment of protein secondary structure prediction methods based on amino acid sequence.

Five of the several secondary structure prediction methods based on protein amino acid sequence has been computerized, allowing the calculation of joint prediction histograms which have been shown to be superior to any individual prediction. The known structures of about 40 proteins experimentally determined by X-ray crystallography are compared with the predictions resulting from calculated histograms. The accuracy of the predictions for helices is generally much better than for both beta-sheet regions and for turns. The overall agreement between prediction and observation within the amino terminal half of the protein molecules is clearly superior to that for the carboxyl half, suggesting an amino nucleating core. Predictions for smaller proteins and thermally stable proteins are generally good, indicating the sensitivity of the methods to short-range but not long-range interactions. In less than half the cases tested were the predictions useful; there was no way of knowing ahead of time if a favorable prediction would result. Given the lack of dramatic improvement with an increase in data base for the schemes and the generally poor agreement factors, it appears that a perfect predictive algorithm must include a consideration of energy minimization, thermalization, and long-range interactions. Extreme caution is suggested in applying present prediction routines to unknown protein structures.

Amino Acid Sequence↗
PubMed2006

A consensus epitope prediction approach identifies the breadth of murine T(CD8+)-cell responses to vaccinia virus.

The value of predictive algorithms for identifying CD8+ T (T(CD8+))-cell epitopes has not been adequately tested experimentally. Here we demonstrate that conventional bioinformatic methods predict the vast majority of T(CD8+)-cell epitopes derived from vaccinia virus WR strain (VACV-WR) in the H-2(b) mouse model. This approach reveals the breadth of T-cell responses to vaccinia, a widely studied murine viral infection model, and may provide a tool for developing comprehensive antigenic maps of any complex pathogen.

Algorithms↗
PubMed1999

Implementation of probabilistic decision rules improves the predictive values of algorithms in the diagnostic management of ectopic pregnancy.

Current algorithms for the diagnosis of ectopic pregnancy do not take into account the heterogeneity in patient profiles. Such heterogeneity can lead to differences in the pre-test probability of ectopic pregnancy. In patients with clinical symptoms, for example, the probability of presence of an ectopic pregnancy is higher than in symptom-free patients. Any additional tests should then be interpreted differently, depending on the pre-test probability. We present a diagnostic algorithm that uses probabilistic decision rules for the evaluation of women with suspected ectopic pregnancy with flexible cut-off levels for test positivity We compare it with a general algorithm that uses fixed cut-off levels. Fictitious cohorts, varying in prevalence of ectopic pregnancy were put together, using data obtained from a cohort of >800 women with suspected ectopic pregnancy. In the inflexible algorithm, ectopic pregnancy was diagnosed whenever it could be visualized at transvaginal sonography, or where serum human chorionic gonadotrophin (HCG) exceeded a rigid cut-off level; ectopic pregnancy was rejected if an intrauterine pregnancy was seen or when serum HCG decreased. In the flexible algorithm, a post-test probability was obtained after each test, using pre-test probabilities and test-based likelihood ratios. Ectopic pregnancy was diagnosed whenever the post-test probability for ectopic pregnancy exceeded 95%, whereas this diagnosis was rejected if the calculated post-test probability fell below 1%. For both algorithms, sensitivity and specificity as well as predictive values were calculated. At each prevalence, the inflexible algorithm was associated with a sensitivity of 93% and a specificity of 97%. In contrast, the sensitivity and specificity of the flexible, individualized algorithm depended on the prevalence of ectopic pregnancy. Consequently, predictive values varied strongly when the inflexible algorithm was used, whereas they were much more stable after using the flexible algorithm. For five possible valuations of false positive and false negative diagnoses, the flexible algorithm reduced the expected disutility, compared with the inflexible algorithm. It is concluded that clinicians should incorporate probabilistic decision rules in algorithms used for the diagnosis of ectopic pregnancy.

Algorithms↗
PubMed2004

DNA microarray analysis of natural killer cell-type lymphoproliferative disease of granular lymphocytes with purified CD3-CD56+ fractions.

Natural killer (NK) cell-type lymphoproliferative disease of granular lymphocytes (LDGL) is characterized by the outgrowth of CD3(-)CD16/56(+) NK cells, and can be further subdivided into two distinct categories: aggressive NK cell leukemia (ANKL) and chronic NK lymphocytosis (CNKL). To gain insights into the pathophysiology of NK cell-type LDGL, we here purified CD3(-)CD56(+) fractions from healthy individuals (n=9) and those with CNKL (n=9) or ANKL (n=1), and compared the expression profiles of >12 000 genes. A total of 15 'LDGL-associated genes' were identified, and a correspondence analysis on such genes could clearly indicate that LDGL samples share a 'molecular signature' distinct from that of normal NK cells. With a newly invented class prediction algorithm, 'weighted distance method', all 19 samples received a clinically matched diagnosis, and, furthermore, a detailed cross-validation trial for the prediction of normal or CNKL status could achieve a high accuracy (77.8%). By applying another statistical approach, we could extract other sets of genes, expression of which was specific to either normal or LDGL NK cells. Together with sophisticated statistical methods, gene expression profiling of a background-matched NK cell fraction thus provides us a wealth of information for the LDGL condition.

Adolescent↗
PubMed2004

Algorithms for sequence analysis via mutagenesis.

MOTIVATION: Despite many successes of conventional DNA sequencing methods, some DNAs remain difficult or impossible to sequence. Unsequenceable regions occur in the genomes of many biologically important organisms, including the human genome. Such regions range in length from tens to millions of bases, and may contain valuable information such as the sequences of important genes. The authors have recently developed a technique that renders a wide range of problematic DNAs amenable to sequencing. The technique is known as sequence analysis via mutagenesis (SAM). This paper presents a number of algorithms for analysing and interpreting data generated by this technique. RESULTS: The essential idea of SAM is to infer the target sequence using the sequences of mutants derived from the target. We describe three algorithms used in this process. The first algorithm predicts the number of mutants that will be required to infer the target sequence with a desired level of accuracy. The second algorithm infers the target sequence itself, using the mutant sequences. The third algorithm assigns quality values to each inferred base. The algorithms are illustrated using mutant sequences generated in the laboratory.

Algorithms↗
PubMed2005

Comparison of inhomogeneity correction algorithms in small photon fields.

Algorithms such as convolution superposition, Batho, and equivalent pathlength which were originally developed and validated for conventional treatments under conditions of electronic equilibrium using relatively large fields greater than 5 x 5 cm2 are routinely employed for inhomogeneity corrections. Modern day treatments using intensity modulated radiation therapy employ small beamlets characterized by the resolution of the multileaf collimator. These beamlets, in general, do not provide electronic equilibrium even in a homogeneous medium, and these effects are exaggerated in media with inhomogenieties. Monte Carlo simulations are becoming a tool of choice in understanding the dosimetry of small photon fields as they encounter low density media. In this study, depth dose data from the Monte Carlo simulations are compared to the results of the convolution superposition, Batho, and equivalent pathlength algorithms. The central axis dose within the low-density inhomogeneity as calculated by Monte Carlo simulation and convolution superposition decreases for small field sizes whereas it increases using the Batho and equivalent pathlength algorithms. The dose perturbation factor (DPF) is defined as the ratio of dose to a point within the inhomogeneity to the same point in a homogeneous phantom. The dose correction factor is defined as the ratio of dose calculated by an algorithm at a point to the Monte Carlo derived dose at the same point, respectively. DPF is noted to be significant for small fields and low density for all algorithms. Comparisons of the algorithms with Monte Carlo simulations is reflected in the DCF, which is close to 1.0 for the convolution-superposition algorithm. The Batho and equivalent pathlength algorithms differ significantly from Monte Carlo simulation for most field sizes and densities. Convolution superposition shows better agreement with Monte Carlo data versus the Batho or equivalent pathlength corrections. As the field size increases the DCF's for all algorithms converge toward 1.0. The largest differences in DCF are at the interface where changes in electron transport are greatest. For a 6 MV photon beam, electronic equilibrium is restored at field sizes above 3 cm diameter and all of the algorithms predict dose in and beyond the inhomogeneous region equally well. For accurate dosimetry of small fields within and near inhomogeneities, however, simple algorithms such as Batho and equivalent pathlength should be avoided.

Algorithms↗
PubMed1989

Adaptive realtime compensatory feedback to open the light-pupil reflex loop.

Integrity of the visual pathway is often clinically determined from response of the pupil to light. The need for more quantitative measurement of pupillary response has led engineers to investigate the light-pupil reflex with systems analysis techniques. The results have made the light-pupil reflex a model of the application of the principles of control theory to physiologic mechanisms. This analysis required that the system be rendered open loop by focusing the stimulus at the center of the pupil. Patients may not be able to restrict their head and eye movement well enough to permit the feedback loop to be rendered open loop optically. Thus, clinical applications of the technique have been limited. A means of opening the feedback loop by using a realtime deterministic autoregressive moving average prediction algorithm integrated into a computer based television pupillometer has been developed. It employs a model reference adaptive scheme to identify the physiologic system, then predicts future values of pupil area to compensate for delays in the measuring apparatus. A computer driven liquid crystal display shutter controls retinal illuminance to render the light-pupil reflex open loop. This technique reduced the mean square error of retinal illuminance by about 75% compared to uncompensated pupil responses.

Algorithms↗
PubMed1999

Scatter and attenuation measurements at distances greater than 10 cm from an 192Ir source.

Most algorithms developed by various investigators for use in brachytherapy treatment planning have typically been designed to calculate the dose within a 10 cm range of a radiation source. These algorithms predict the dose well at distances < 10 cm from the source but were not developed and should not be utilized to predict the dose at distances > 10 cm. On the contrary, treatment planning systems and manual calculations will produce erroneous results when dose points > 10 cm are calculated using these algorithms. The spread in the data generated by the above algorithms is 16% at 15 cm and 42% at 20 cm. Physical measurements were performed at distances between 5 and 50 cm from a high activity 192Ir source in water. The measured data correlated well with the predicted data from 5 to 10 cm, which had a 5% spread. Beyond 10 cm the measured data fell central to the range of the predicted data, with the spread of the predicted data increasing from 5% to 80% with increasing distance from the source. The measured data was fitted with a model incorporating a buildup factor and an attenuation factor. The best fit values are in reasonable agreement with those obtained by two of the investigators, Thomason and Tripathi.

Algorithms↗
PubMed2004

Gene prediction and verification in a compact genome with numerous small introns.

The genomes of clusters of related eukaryotes are now being sequenced at an increasing rate, creating a need for accurate, low-cost annotation of exon-intron structures. In this paper, we demonstrate that reverse transcription-polymerase chain reaction (RT-PCR) and direct sequencing based on predicted gene structures satisfy this need, at least for single-celled eukaryotes. The TWINSCAN gene prediction algorithm was adapted for the fungal pathogen Cryptococcus neoformans by using a precise model of intron lengths in combination with ungapped alignments between the genome sequences of the two closely related Cryptococcus varieties. This approach resulted in approximately 60% of known genes being predicted exactly right at every coding base and splice site. When previously unannotated TWINSCAN predictions were tested by RT-PCR and direct sequencing, 75% of targets spanning two predicted introns were amplified and produced high-quality sequence. When targets spanning the complete predicted open reading frame were tested, 72% of them amplified and produced high-quality sequence. We conclude that sequencing a small number of expressed sequence tags (ESTs) to provide training data, running TWINSCAN on an entire genome, and then performing RT-PCR and direct sequencing on all of its predictions would be a cost-effective method for obtaining an experimentally verified genome annotation.

Algorithms↗
PubMed2003

Identification of functional clusters of transcription factor binding motifs in genome sequences: the MSCAN algorithm.

MOTIVATION: The identification of regulatory control regions within genomes is a major challenge. Studies have demonstrated that regulating regions can be described as locally dense clusters or modules of cis-acting transcription factor binding sites (TFBS). For well-described biological contexts, it is possible to train predictive algorithms to discern novel modules in genome sequences. However, utility of module detection methods has been severely limited by insufficient training data. For only a few tissues can one obtain sufficient numbers of literature-derived regulatory modules. RESULTS: We present a novel method, MSCAN, that circumvents the training data problem by measuring the statistical significance of any non-overlapping combination of TFBS in a window. Given a set of transcription factor binding profiles, a significance threshold, and a genomic sequence, MSCAN returns putative regulatory regions. We assess performance on two curated collections of regulatory regions; one each for tissue-specific expression in liver and skeletal muscle cells. The efficiency of MSCAN allows for predictive screens of entire genomes.

Algorithms↗
PubMed2000

Hyperopic photorefractive keratectomy with the erodible mask and axicon system: two year follow-up.

PURPOSE: To evaluate efficacy and long-term stability of hyperopic photorefractive keratectomy (H-PRK) using the erodible mask and Axicon system. SETTING: Department of Ophthalmology, St. Thomas' Hospital, London, England. METHODS: Forty-three patients (43 eyes), with a mean preoperative refraction (spherical equivalent) of +4.54 diopters (D) (range +1.75 to +7.50 D), were treated using an Apex Plus(R) excimer laser (Summit Technology). This uses an erodible mask to ablate a 6. 50 mm diameter hyperopic correction and an Axicon to fashion a 1.50 mm blend zone around the correction. The overall ablation diameter was 9.50 mm. Follow-up was 2 years. RESULTS: At 2 years, the mean manifest refraction was +0.16 D (range +4.125 to -4.000 D), with the induced correction appearing stable after 9 months. Based on the Munnerlyn algorithm, predictability was acceptable for corrections up to +4.50 D, with 68% of eyes within +/-1.00 D of the predicted correction. It was poorer for +6.00 D corrections, with 33% of eyes within +/-1.00 D of that expected. Patient satisfaction was high. Forty eyes (93%) had an improvement in uncorrected near visual acuity and 37 (86%), an improvement in uncorrected distance acuity. A peripheral ring of haze, 6.5 mm in diameter, appeared in all eyes 1 month postoperatively. Its intensity was maximal at 3 to 9 months and then diminished over time. There were no significant differences in measurements of the central corneal transparency at 12 and 24 months and those preoperatively. Measurements of flicker contrast sensitivity, forward light scatter (glare), and scotopic halos showed no significant differences between preoperative values and those measured after 6 months. CONCLUSIONS: Achieved H-PRK corrections with the erodible mask and Axicon system agreed closely with the Munnerlyn algorithm, with refractive stability after 9 months. Predictability was acceptable for corrections up to +4.50 D. Axial corneal transparency was not compromised and visual performance, in terms of best spectacle-corrected visual acuity, forward light scatter, and night halos, was not impaired.

Adult↗
PubMed1994

Prediction of response to erythropoietin treatment in chronic anemia of cancer.

Chronic anemia of cancer can be corrected in approximately 50% of the cases by treatment with recombinant human erythropoietin (rHuEPO). Early prediction of responsiveness would avoid the emotional and financial burden of ineffective medical intervention. Eighty patients with chronic anemia of cancer undergoing treatment with rHuEPO (150 U/kg, 3 times per week by subcutaneous injection; after 6 weeks without response, 300 U/kg) participated in this study. Response was defined as a gain of at least 2 g/dL hemoglobin (Hb) within 12 weeks. Multivariate discriminant analysis and logistic regression analysis of response were performed on routine blood tests; serum levels of EPO, iron, ferritin, transferrin, and its receptor; World Health Organization (WHO) performance status; various cytokines; neopterin; stem cell factor; C-reactive protein; and alpha 1-antitrypsin. At baseline, none of these factors showed sufficient prognostic power. The following predictive algorithm was developed: (1) If after 2 weeks of therapy both the serum EPO level is > or = 100 mU/mL and Hb concentration has not increased by at least 0.5 g/dL, unresponsiveness of the patient is very likely (predictive power, 93%); otherwise, response may be predicted with an accuracy of 80%. (2) If both the serum level of EPO is less than 100 mU/mL and Hb concentration has increased by > or = 0.5 g/dL, response is highly probable (predictive power, 95%). (3) Alternatively, a serum ferritin level of > or = 400 ng/mL after 2 weeks of rHuEPO therapy strongly indicates unresponsiveness (predictive power, 88%), whereas a level less than 400 ng/mL suggests response in 3 of 4 patients.

Adult↗
PubMed1997

Simian T cell leukemia virus type I from naturally infected feral monkeys from central and west Africa encodes a 91-amino acid p12 (ORF-I) protein as opposed to a 99-amino acid protein encoded by HTLV type I from humans.

A single protein of 12 kDa, p12 is encoded by the HTLV-I genome from both the singly spliced mRNA pX-ORF-I and doubly spliced mRNA pX-rex-ORF-I. While many full-length sequences of HTLV-1 are known, data on the p12 regions of African STLV-I are unavailable. We have undertaken to sequence the p12 gene in STLV-I from Central and West Africa naturally infected primates, and have compared them to known p12 sequences of HTLV-I. Our data on sequence and in vitro transcription-translation analyses indicate that p12 is a 91-amino acid (aa) protein among STLV-I strains from Central and West Africa, in contrast to the 99-aa protein found among HTLV-I strains around the globe. The p12 sequences of STLV-I exhibit a marked genetic variability at the level of both nucleotide and peptide sequences. Hydropathic and helical wheel analyses reveal that 60% of residues in HTLV-I p12 are hydrophobic, in contrast to 55% in STLV-I from Africa. Although HTLV-I and STLV-I show a similar putative antigenic site, a second potential site was located exclusively in STLV-I from Africa. There are differences in the predicted transmembrane domains in p12 between STLV-I and HTLV-I. Furthermore, the secondary structure data according to the Chou and Fasman algorithm predict an alpha-helical domain at the carboxy terminus in HTLV-I, and this domain may be truncated in STLV-I p12. The amino acid sequence of p12 shows two leucine zipper motifs (LZMs) at the amino terminus and in the middle region, respectively. This is the first report describing the size differences in p12 protein between HTLV-I and STLV-I, which may provide insights into pathogenic mechanisms used by HTLV-I and STLV-I.

Africa, Central↗
PubMed2002

Predictors of bacteremia in febrile children with sickle cell disease.

PURPOSE: Bacteremia is an important cause of death and complications in children with sickle cell disease (SCD), yet predictors of bacteremia in these patients have not been well identified. The purpose of this study was to test whether clinical and hematologic variables commonly used to predict bacteremia in normal young children with fever could accurately predict bacteremia in febrile children with SCD. PATIENTS AND METHODS: The authors reviewed the medical records of all patients with SCD younger than 18 years of age over a 10-year period at a single institution for febrile events. They tested the univariate associations of age, height of fever, white blood cell count (WBC), absolute neutrophil count (ANC), and absolute band count (ABC) with bacteremia. Three separate multivariate analyses were performed using the predictor variables age, temperature, and one of three hematologic variables (ANC, WBC, or ABC) with the outcome bacteremia. RESULTS: There were 175 evaluable febrile events, of which 8 (4.6%) were associated with bacteremia. In the multivariate analyses, all hematologic variables, but not age or height of fever, retained significant associations with bacteremia. CONCLUSIONS: In febrile children with SCD, WBC, ANC, and ABC are all independently associated with bacteremia when adjusting for height of fever and age. Hematologic variables may be useful in developing prediction algorithms to identify febrile patients with SCD at higher risk of bacteremia. These data emphasize the need for a national trial to develop a predictive model with defined thresholds.

Age Factors↗