A validated risk model for early neutropenic events in older cancer patients receiving systemic chemotherapy

2007 ◽  
Vol 25 (18_suppl) ◽  
pp. 9036-9036 ◽  
Author(s):  
M. Shayne ◽  
E. Culakova ◽  
D. C. Dale ◽  
M. S. Poniewierski ◽  
D. A. Wolff ◽  
...  
Keyword(s):  
Cancer Patients ◽  
Predictive Value ◽  
Risk Model ◽  
Low Risk ◽  
Severe Neutropenia ◽  
Validation Dataset ◽  
Cancer Type ◽  
C Statistic ◽  
Risk Patients ◽  
Older Cancer Patients

9036 Background: A prospective, nationwide study was undertaken to develop and validate a risk model for early neutropenic events (NE) in older cancer patients undergoing chemotherapy. Methods: 1,386 patients =65 years of age with lung, breast, colorectal, ovarian cancer or lymphoma were prospectively registered at 117 randomly selected sites. Data on up to 4 cycles were collected upon initiation of chemotherapy. A logistic regression model for cycle 1 NE consisting of febrile neutropenia (FN; fever/infection and absolute neutrophil count nadir <1x109/L) or severe neutropenia (SN; neutrophils <.5x109/L) was derived on 1,378 patients with available data. Validation was performed using a split sample random selection process. Results: No significant differences in distribution of NE or predictive factors were observed between derivation dataset (n=922) and validation dataset (n=464). Major independent baseline clinical risk factors for cycle 1 NE in the derivation model (DM) included: anthracycline based regimens (p<.001), non-chemotherapy immune-modulatory agents (p=.003), elevated bilirubin (p=.016), reduced glomerular filtration rate (p<.001), cancer type (p=.02), planned relative dose intensity =85% (p=.027), and regimens containing cyclophosphamide (p<.001), etoposide (p=.002) or ifosfamide (p=.032). Reduced risk of cycle 1 NE was associated with myeloid growth factor (MGF) prophylaxis (p<.001). DM R2 was 0.478 and c-statistic 0.88 [95% CI 0.86–0.91; p<.001]. At median predicted risk of cycle 1 NE of 7%, model test performance (MTP) showed: sensitivity 90%; specificity 59%; and predictive value positive and negative of 32% and 97%, respectively. Cycle 1–4 FN risk in the DM was 16.6% and 3.3% among high and low risk patients, respectively. The validation model (VM) R2 was 0.508 and c-statistic 0.89 [95% CI: 0.86–0.93; p<.001]. MTP in the VM demonstrated: sensitivity 90%; specificity 65%; predictive value positive and negative of 36% and 97%, respectively. Cycle 1–4 FN risk in the VM was 16.8% and 1.6% in high and low risk patients, respectively. Conclusions: This validated risk model demonstrated good discrimination between older cancer patients at decreased risk for NE, and those at increased risk who may benefit from targeted prophylaxis with MGF. No significant financial relationships to disclose.

Prospective validation of a risk model for first cycle neutropenic complications in patients receiving cancer chemotherapy

2006 ◽  
Vol 24 (18_suppl) ◽  
pp. 8561-8561 ◽  
Author(s):  
G. H. Lyman ◽  
N. M. Kuderer ◽  
J. Crawford ◽  
D. A. Wolff ◽  
E. Culakova ◽  
...  
Keyword(s):  
High Risk ◽  
Test Performance ◽  
Risk Model ◽  
Selection Process ◽  
Low Risk ◽  
Validation Dataset ◽  
Cancer Type ◽  
General Applicability ◽  
C Statistic ◽  
Increased Risk

8561 Background: A nationwide, prospective cohort study was undertaken to develop and validate a risk model for neutropenic complications (NC) in cancer patients receiving chemotherapy. Methods: 3,596 patients initiating a new chemotherapy regimen with solid tumors or lymphoma were registered at 115 randomly selected sites. Data on at least 1 cycle of chemotherapy were available on 3,468. A logistic regression model for cycle 1 NC was derived and then validated using a split sample random selection process. Results: The risk of cycle 1 NC ranged from 5.5%-30.2%, averaging 18.5% across tumor types. No significant differences in distribution of NC or predictive factors were observed between the derivation dataset (n=2,592) or the validation dataset (n=876). Major independent baseline clinical risk factors for cycle 1 NC in the derivation model include: prior chemotherapy (P=.044), number of myelosuppressive agents (P<.0001), anthracycline-based regimens (P<.0001), planned delivery >85% of standard (P<.0001), cancer type (P<.0001), concurrent antibiotics (P=.023) or phenothiazines (P=.006), abnormal alkaline phosphatase (P=.002), elevated bilirubin (P=.031), low platelets (P=.004), elevated glucose (P=.023) and reduced glomerular filtration rate (P=.013). Reduced risk of cycle 1 NC was associated with primary prophylaxis with a myeloid growth factor (P<.0001). Model R2 was 0.273 and c-statistic 0.80 [95% CI: 0.78–0.82; P<.0001]. At the median predicted risk of cycle 1 NC of 11%, model test performance consisted of: sensitivity 84%; specificity 57% and diagnostic odds ratio (DOR) 7.2 while cycle 1 NC risk was 31% and 6% among high risk and low risk half, respectively. The model performed well in the smaller validation dataset with a model R2 of 0.354 and c-statistic of 0.84 [95% CI: 0.81–0.87, P<.0001]. Test performance of the model in the validation sample included: sensitivity 90%; specificity 62%; DOR 14.1 and risks of 35% and 4% in high risk and low risk patients, respectively. Conclusions: Validation in a randomly selected patient sample suggests that this model has general applicability in identifying patients at increased risk for NC. Further validation in other independent cancer patient populations receiving chemotherapy is planned. [Table: see text]

A Conditional Risk Model for Chemotherapy-Induced Anemia (CIA) in Cancer Patients.

Blood ◽  
2007 ◽  
Vol 110 (11) ◽  
pp. 372-372
Author(s):  
Gary H. Lyman ◽  
Jeffrey Crawford ◽  
Nicole M. Kuderer ◽  
Debra A. Wolff ◽  
Eva Culakova ◽  
...  
Keyword(s):  
Platelet Count ◽  
Cancer Patients ◽  
Predictive Value ◽  
Risk Model ◽  
Low Risk ◽  
Performance Characteristics ◽  
Individual Risk ◽  
Hematological Toxicity ◽  
Cancer Type ◽  
Conditional Risk

Abstract Anemia represents the most common hematological toxicity in cancer patients receiving systemic chemotherapy and is associated with considerable morbidity and cost. Current guidelines for chemotherapy-induced anemia call for intervention at a hemoglobin &lt;10 g/dL with treatment options including transfusion or an erythropoietic-stimulating agent (ESA). A meta-analysis of randomized controlled trials has demonstrated the clinical value of early versus late intervention with an ESA (Lyman Cancer, 2006). While anemia risk models based on pretreatment characteristics have recently been validated, recent safety concerns have limited use of the ESAs to patients with moderate or severe anemia. The gradual onset of anemia and response to ESAs over time provides a rationale for selecting patients for ESA support early in the course of chemotherapy. Methods: 3640 patients with solid tumors or malignant lymphoma initiating a new regimen have been prospectively registered at 110 randomly selected U.S. practice sites. A logistic regression risk model for hemoglobin &lt;10 g/dL based on pretreatment characteristics and hematolgic events during cycle 1 was developed and model predictive performance characteristics estimated. Results: Over a median of 3 cycles of chemotherapy, hemoglobin &lt;10 g/dL was reported one or more times in 1072 (29.5%) patients. Significant independent baseline characteristics associated with subsequent hemoglobin &lt;10 g/dL include: female gender, poor ECOG performance status, history of congestive heart failure, vascular disease or chronic pulmonary disease, cancer type, treatment with an anthracycline-, platinum- or etoposide-based regimen and baseline hemoglobin &lt;12 g/dL or platelet count &lt;150000/mm3. In addition, significant independent predictive hematologic changes during cycle 1 include: decrease in hemoglobin &gt;1 g/dL (OR=4.48; P&lt;.0001), decrease in platelet count &gt;100000/mm3 (OR=1.54;P&lt;.0001) or neutrophil count &lt;500/mm3 (OR=1.94; P&lt;.001) as well as hemoglobin &lt;12 g/dL (OR=2.0;P&lt;.001) at the start of cycle 2. Model R2 = 0.581 and c-statistic = 0.901 [95% CI: .89–.91, P&lt;.0001]. The predicted risk of hemoglobin &lt;10 g/dL ranged from 0 to 100% with mean and median probabilities of 0.16 and 0.30, respectively. Based on a risk cutpoint at the mean, 1290 patients (38%) were classified as high risk. The risks of hemoglobin &lt;10 g/dL in high and low risk subjects were 66% and 9%, respectively. Model test performance characteristics [± 95% CLs] included: sensitivity: 82%[80–84]; specificity: 82%[80–83]; positive predictive value: 66%[63–68]; negative predictive value: 91%[90–93] and diagnostic odds ratio: 20.4[16.8–24.6]. Of note, risk of hemoglobin &lt;11 g/dL in high and low risk subjects based on this model were 95% and 34%, respectively. Validation of the model in a separate population of patients is currently under way. Discussion: This conditional risk model based on both pretreatment characteristics and first cycle events identified cancer patients receiving chemotherapy at substantial risk for clinically significant anemia. The use of ESAs early in the course of treatment based on individual risk assessment must consider both the potential benefit and risks and careful monitoring is essential.

Validation of a Risk Model for Hospitalized Adult Cancer Patients with Febrile Neutropenia.

Blood ◽  
2004 ◽  
Vol 104 (11) ◽  
pp. 89-89
Author(s):  
Nicole M. Kuderer ◽  
Jeffrey Crawford ◽  
David C. Dale ◽  
Gary H. Lyman ◽  
Keyword(s):  
Cancer Patients ◽  
Risk Score ◽  
Risk Model ◽  
Low Risk ◽  
Inpatient Mortality ◽  
Medical Centers ◽  
Academic Medical ◽  
Independent Population ◽  
Risk Patients ◽  
Very High

Abstract Introduction: Febrile neutropenia (FN) represents the most common dose-limiting toxicity associated with systemic chemotherapy and is associated with considerable morbidity and mortality. The majority of patients with FN are hospitalized for evaluation and administration of empiric broad-spectrum antibiotics. A risk model for mortality in hospitalized adult cancer patients with FN has previously been reported (Kuderer, ASCO 2004). Validation of this risk model in an independent population of hospitalized patients with FN is presented. Methods: Risk model development was based on the records of 40,163 adult non-transplant cancer patients hospitalized with FN at one of 115 academic medical centers reporting to the University HealthSystem Consortium between 1995 and 2000. A risk score for mortality was estimated for each patient based on a weighted summary of all significant variables in a logistic regression model including age, cancer diagnosis, comorbidities (heart, liver, renal, cerebrovascular, pulmonary embolism) and complications (gram+ and gram- sepsis, fungal infection, pneumonia, hypotension, hypovolemia, ICU admission). The risk score was classified into low risk (0–4); intermediate (5–9); high (10–14); and very high (15–22). Results: To validate the FN mortality risk score, it is applied here to an independent population of 16,379 adult cancer patients hospitalized with FN at academic medical centers between 2001–2002. Inpatient mortality was reported in 1,452 (8.9%). Although patients ≥65 years of age represented 30% of the inpatient FN population, they accounted for 43% of deaths. Application of the previously developed risk model to this independent population provided adjusted estimates of relative risk (odds ratios) of: age ≥ 65 (1.5); leukemia (1.3); lung cancer (1.3); heart (1.4); liver (2.1); renal (3.0); cerebrovascular (3.6); pulmonary embolism (2.8); gram+ sepsis (2.1); gram- sepsis (2.4); fungal (1.4); pneumonia(1.8); hypotension (1.9); hypovolemia (1.6) and ICU (3.7) (Global χ2:P<.0001). All significant covariates in the development phase remained significant in the validation study. As shown in the table, the model demonstrated excellent fit (P<.0001) and a high level of discrimination for inpatient mortality (R2 = 0.81; c-statistic = 0.85 [0.84, 0.86]; P<.0001). Risk Score Category Low (0–4) Int (5–9) High (10–14) Very High (15+) Development Phase Patients (%) 74.6 20.6 4.3 0.4 Deaths (%) 3.1 18.9 48.1 64.6 Validation Phase Patients (%) 68.5 23.3 7.0 1.2 Deaths (%) 2.5 14.4 42.3 68.8 Test performance among the 22% of patients with a predicted risk of mortality of ≥10% include: sensitivity: 71%; specificity: 83%; positive predictive value: 29%; negative predictive value: 97%; likelihood ratio positive: 4.2; likelihood ratio negative: 0.3; diagnostic odds ratio: 12.2 [10.8, 13.8]. A low risk subgroup (23%) was also identified with a risk <1%. Similar validation was achieved of previously reported models for length of stay and cost. Conclusions: Previously reported risk models for mortality, length of stay and cost have been validated in a separate population of adult cancer patients hospitalized with FN. These validated models may assist clinicians in identifying both high-risk patients for more aggressive supportive care measures as well as low risk patients as candidates for early discharge. The cost-effectiveness of these models in assisting clinical decision-making is currently under study.

A Risk Model for Chemotherapy-Induced Anemia (CIA) in Cancer Patients.

Blood ◽  
2005 ◽  
Vol 106 (11) ◽  
pp. 754-754
Author(s):  
Gary H. Lyman ◽  
Brandon McMahon ◽  
Nicole M. Kuderer ◽  
Jeffrey Crawford ◽  
Debra Wolff ◽  
...  
Keyword(s):  
Cancer Patients ◽  
Likelihood Ratio ◽  
Test Performance ◽  
Predictive Value ◽  
Risk Model ◽  
Performance Status ◽  
Low Risk ◽  
Performance Characteristics ◽  
Hematological Toxicity ◽  
Predictive Test

Abstract Background: Anemia represents the most common hematological toxicity in cancer patients receiving systemic chemotherapy and is associated with considerable morbidity and cost (Lyman‚ Value in Health 2005). Current ASH/ASCO guidelines call for intervention at a hemoglobin (Hgb) &lt;10 gm/dl. Treatment options include transfusion or administration of an erythropoietic-stimulating protein (ESP). A recent meta-analysis demonstrated the clinical value of early versus late intervention with an ESP (Lyman‚ Cancer‚ 2005 in press). An accurate and valid risk model for CIA is needed to select patients for ESP treatment early in the course of chemotherapy when it can be most effective. Methods: More than 3‚000 patients with cancer of the breast‚ lung‚ colon and ovary or malignant lymphoma initiating a new chemotherapy regimen have been prospectively registered at 115 randomly selected U.S. practice sites. Data on at least one cycle of chemotherapy were available on 2‚842 patients. A logistic regression model for Hgb &lt;10 gm/dl based on pretreatment characteristics was developed and predictive test performance characteristics examined. Results: Over a median of three cycles of chemotherapy, Hgb &lt;10 gm/dl was reported one or more times in 817 (28.7%) patients. Significant independent predictive factors for Hgb &lt;10 gm/dl include: history of peptic ulcer (OR=1.90; P=.015), myocardial infarction (OR=1.94; P=.009), or congestive heart failure (OR=2.13; P=.017), increasing age (OR=1.02; P=.002), female gender (OR=2.40; P&lt;.001), ECOG performance status (OR=1.24; P=.002), Charlson Comorbidity Index (OR=1.06, P=.002), body surface area (OR=3.75, P&lt;.001), low baseline hemoglobin (OR=1.95, P&lt;.001), pretreatment hematocrit (OR=.85, P&lt;.001), and glomerular filtration rate (OR=0.99, P=.027), and regimens containing anthracyclines (OR=3.21, P&lt;.001), cisplatinum (OR=3.86, P&lt;.001) or carboplatinum (OR=2.71, P&lt;.001). Model fit was excellent (P&lt;.001), R2=0.455 and c-statistic = 0.775 [95% CL: .76–.79, P&lt;.0001]. Individual predicted risk of Hgb &lt;10 gm/dl based on the model ranged from 0 to 98% with mean and median probabilities of 0.28 and 0.22, respectively. Based on a risk cutpoint of 20%, 1,541 patients (55%) were classified as high risk and 1,282 as low risk. The average risks of Hgb &lt;10 gm/dl during chemotherapy in high and low risk subjects were 43% and 12%, respectively. Model test performance characteristics [±95% CL] included: sensitivity: 81% [78–84]; specificity: 56% [54–58]; likelihood ratio positive: 1.85 [1.74–1.96]; likelihood ratio negative: 0.34 [0.29–0.39]; positive predictive value: 43% [40–45]; negative predictive value: 88% [86–90] and diagnostic odds ratio: 5.47 [4.50–6.66]. Conclusions: This risk model identified cancer patients initiating chemotherapy who are at risk for clinically significant anemia using common clinical parameters. Validation of the model in a separate population of patients is in progress.

scholarly journals Trastuzumab-induced cardiotoxicity: testing a clinical risk score in a real-world cardio-oncology population

2017 ◽  
Vol 24 (3) ◽  
pp. 176 ◽  
Author(s):  
M. Rushton ◽  
C. Johnson ◽  
S. Dent
Keyword(s):  
Breast Cancer ◽  
High Risk ◽  
Cancer Patients ◽  
Real World ◽  
Predictive Value ◽  
Low Risk ◽  
Clinical Risk Score ◽  
Breast Cancer Patients ◽  
Risk Patients

Background Trastuzumab has improved survival for women with her2-positive breast cancer, but its use is associated with an increased risk of cardiotoxicity. With increased survivorship, the long-term effects of cancer treatment are an important consideration for clinicians and patients. We reviewed the current literature on predicting trastuzumab-related cardiotoxicity and tested a clinical risk score (crs) in a real-world breast cancer population to assess its utility in predicting permanent cardiotoxicity.Methods In this retrospective exploratory cohort study of breast cancer patients referred to a cardio-oncology clinic at a tertiary care centre between October 2008 and August 2014, a crs was calculated for each patient, and a sensitivity analysis was performed.Results Of the 143 patients included in the study, 62 (43%) experienced a cardiac event, and of those 62 patients, 43 (69%) experienced full recovery of cardiac function. In applying the crs, 119 patients (83%) would be considered at low risk, 14 (10%) at moderate risk, and 10 (7%) at high risk to develop heart failure or cardiomyopathy. When applied to the study population, the high-risk cut-off score had a sensitivity of 0.13 [95% confidence interval (ci): 0.08 to 0.20] and a specificity of 0.94 (95% ci: 0.87 to 0.97). The positive predictive value was 0.07 (95% ci: 0.03 to 0.13), and the negative predictive value was 0.93 (95% ci: 0.87 to 0.96).Conclusions The crs demonstrated good specificity and negative predictive value for the development of permanent cardiotoxicity in a real-world population of breast cancer patients, suggesting that intensive cardiac monitoring might not be warranted in low-risk patients, but that high-risk patients might benefit from early referral to cardio-oncology for optimization. Further study using the crs in a larger breast cancer population is warranted to identify patients at low risk of long-term trastuzumab-related cardiotoxicity.

scholarly journals Impact of Different Modules of 21-Gene Assay in Early Breast Cancer Patients

2021 ◽  
Vol 12 ◽  
Author(s):  
Mengdi Chen ◽  
Deyue Liu ◽  
Weilin Chen ◽  
Weiguo Chen ◽  
Kunwei Shen ◽  
...  
Keyword(s):  
Breast Cancer ◽  
High Risk ◽  
Cancer Patients ◽  
Low Risk ◽  
Breast Cancer Patients ◽  
Younger Patients ◽  
Gene Assay ◽  
Risk Patients ◽  
Gene Modules ◽  
The Impact

BackgroundThe 21-gene assay recurrence score (RS) provides additional information on recurrence risk of breast cancer patients and prediction of chemotherapy benefit. Previous studies that examined the contribution of the individual genes and gene modules of RS were conducted mostly in postmenopausal patients. We aimed to evaluate the gene modules of RS in patients of different ages.MethodsA total of 1,078 estrogen receptor (ER)-positive and human epidermal growth factor receptor 2 (HER2)-negative breast cancer patients diagnosed between January 2009 and March 2017 from Shanghai Jiao Tong University Breast Cancer Data Base were included. All patients were divided into three subgroups: Group A, ≤40 years and premenopausal (n = 97); Group B, &gt;40 years and premenopausal (n = 284); Group C, postmenopausal (n = 697). The estrogen, proliferation, invasion, and HER2 module scores from RS were used to characterize the respective molecular features. Spearman correlation and analysis of the variance tests were conducted for RS and its constituent modules.ResultsIn patients &gt;40 years, RS had a strong negative correlation with its estrogen module (ρ = −0.76 and −0.79 in Groups B and C) and a weak positive correlation with its invasion module (ρ = 0.29 and 0.25 in Groups B and C). The proliferation module mostly contributed to the variance in young patients (37.3%) while the ER module contributed most in old patients (54.1% and 53.4% in Groups B and C). In the genetic high-risk (RS &gt;25) group, the proliferation module was the leading driver in all patients (ρ = 0.38, 0.53, and 0.52 in Groups A, B, and C) while the estrogen module had a weaker correlation with RS. The impact of ER module on RS was stronger in clinical low-risk patients while the effect of the proliferation module was stronger in clinical high-risk patients. The association between the RS and estrogen module was weaker among younger patients, especially in genetic low-risk patients.ConclusionsRS was primarily driven by the estrogen module regardless of age, but the proliferation module had a stronger impact on RS in younger patients. The impact of modules varied in patients with different genetic and clinical risks.

scholarly journals A Novel Autophagy-Related lncRNA Gene Signature to Improve the Prognosis of Patients with Melanoma

2020 ◽  
Author(s):  
Yi Ding ◽  
Tian Li ◽  
Min Li ◽  
Tuersong Tayier ◽  
MeiLin Zhang ◽  
...  
Keyword(s):  
Overall Survival ◽  
High Risk ◽  
Risk Model ◽  
Risk Group ◽  
Clinical Information ◽  
Low Risk ◽  
Gene Set Enrichment Analysis ◽  
Positive Regulation ◽  
Risk Patients ◽  
Cox Analysis

Abstract Background: Autophagy and long non-coding RNAs (lncRNAs) have been the focus of research on the pathogenesis of melanoma. However, the autophagy network of lncRNAs in melanoma has not been reported. The purpose of this study was to investigate the lncRNA prognostic markers related to melanoma autophagy and predict the prognosis of patients with melanoma.Methods: We downloaded RNA-sequencing data and clinical information of melanoma from The Cancer Genome Atlas. The co-expression of autophagy-related genes (ARGs) and lncRNAs was analyzed. The risk model of autophagy-related lncRNAs was established by univariate and multivariate COX regression analyses, and the best prognostic index was evaluated combined with clinical data. Finally, gene set enrichment analysis was performed on patients in the high- and low-risk groups.Results: According to the results of the univariate COX analysis, only the overexpression of LINC00520 was associated with poor overall survival, unlike HLA-DQB1-AS1, USP30-AS1, AL645929, AL365361, LINC00324, and AC055822. The results of the multivariate COX analysis showed that the overall survival of patients in the high-risk group was shorter than that recorded in the low-risk group (p<0.001). Moreover, in the receiver operating characteristic curve of the risk model we constructed, the area under the curve (AUC) was 0.734, while the AUC of T and N was 0.707 and 0.658, respectively. The Gene Ontology was mainly enriched with the positive regulation of autophagy and the activation of the immune system. The results of the Kyoto Encyclopedia of Genes and Genomes enrichment were mostly related to autophagy, immunity, and melanin metabolism.Conclusion: The positive regulation of autophagy may slow the transition from low-risk patients to high-risk patients in melanoma. Furthermore, compared with clinical information, the autophagy-related lncRNAs risk model may better predict the prognosis of patients with melanoma and provide new treatment ideas.

Outcome of Severe Vaso-Occlusive Crisis in Sickle Cell Disease Adults Admitted to Referral Centers in Africa and Europe. Introduction of Machine Learning Methods to Improve the Presev Score

Blood ◽  
2020 ◽  
Vol 136 (Supplement 1) ◽  
pp. 14-15
Author(s):  
Yamna Ouchtar ◽  
Christian Kassasseya ◽  
Kene Sekou ◽  
Anne-Laure Pham Hung D'Alexandry D'Orengiani ◽  
Mehdi Kellaf ◽  
...  
Keyword(s):  
Machine Learning ◽  
Sickle Cell Disease ◽  
Decision Tree ◽  
Predictive Value ◽  
Research Funding ◽  
Low Risk ◽  
Cell Disease ◽  
Risk Patients ◽  
Decision Tree Method ◽  
Tree Method

Introduction: Sickle Cell Disease (SCD) is one of the most common genetic disease worldwide. The Acute Chest Syndrome (ACS) is a leading cause of death for SCD patients. The PRESEV1 study was set to produce a predictive score to assess the risk of an ACS development (Bartolucci et al., 2016). PRESEV2 was an international, multicenter prospective confirmatory study to validate the PRESEV score. This study aims at improving these predictions with the addition of a machine learning (ML) method. Patients and methods: Included patients follow PRESEV1 and PRESEV2 studies 'rules. The dataset thus contains 97 patients who developed an ACS episode (18.3%) against 434 patients who did not (81.7%). To compute the PRESEV score, we firstly used the method developed previously with the following variables as input: leukocytes, reticulocytes, hemoglobin levels and cervical spine pain. This method is based on a decision tree with fixed rules and is referred to as the decision tree method throughout this abstract. Secondly we used a ML method using a combined sampling method named SMOTEENN to balance the data and a C-Support Vector Classification (SVC) with fixed parameters to predict the score. This method produces a probability, with a threshold of 0.2, under which the patient is predicted to declare an ACS. We considered the dataset composed of PRESEV1 dataset and 80 percent of PRESEV2 with a randomly choice. The test dataset is thus composed of the remaining 20 percent of PRESEV2. This technique of random choice allowed us to use a 50-cross-validation and compute with Python an average score and a standard deviation (std). In order to allow comparison of the developed score with or without the addition of the ML method, rates were calculated by adding the weight of ACS representation in the dataset. Results: Among all parameters analyzed, the SVC method considered the following variables for calculation of the score: leukocytes, LDH, urea, reticulocytes and hemoglobin levels. A hundred and two adult patients with a severe VOC requiring hospitalization were included. Out of this pool of patients, 26 (25.5%) were predicted with a low risk of developing an ACS episode (SVC method). Sensibility and specificity were of 94.7% and 26.8%, respectfully. The negative predictive value (NPV) was of 95.8% and the positive predictive value (PPV) of 22.4%. Results are resumed in table 1. When compared to the PRESEV score (decision tree method), 44 patients out of 372 were identified with a low risk score (11.8%), Discussion and Conclusion: While the addition of a ML method did not allow the improvement of the sensibility or the NPV of the PRESEV score, it improved both the specificity and the PPV. The addition of artificial intelligence thus provides a better prediction with a higher percentage of "low-risk" patients. As highlighted in the international PRESEV study, this score could represent a useful tool for physicians in hospital settings, with limited beds. While the PRESEV score could allow a better management of "low risk" patients on one side, the identification of "high-risk" patients could also represent a serious advantage to physicians, as it could improve the feasibility of clinical trials for the prevention of this lethal complication in SCD patients. Disclosures Bartolucci: Innovhem: Other; Novartis: Research Funding; Roche: Consultancy; Bluebird: Consultancy; Emmaus: Consultancy; Bluebird: Research Funding; Addmedica: Research Funding; AGIOS: Consultancy; Fabre Foundation: Research Funding; Novartis: Consultancy; ADDMEDICA: Consultancy; HEMANEXT: Consultancy; GBT: Consultancy.

Derivation and Validation of the Surgical Site Infections Risk Model Using Health Administrative Data

2016 ◽  
Vol 37 (4) ◽  
pp. 455-465 ◽  
Author(s):  
Carl van Walraven ◽  
Timothy D. Jackson ◽  
Nick Daneman
Keyword(s):  
Administrative Data ◽  
Risk Model ◽  
Surgical Site Infections ◽  
Low Risk ◽  
Primary Data ◽  
Group Model ◽  
Improvement Program ◽  
Health Administrative Data ◽  
Diagnostic Score ◽  
Risk Patients

OBJECTIVESurgical site infections (SSIs) are common hospital-acquired infections. Tracking SSIs is important to monitor their incidence, and this process requires primary data collection. In this study, we derived and validated a method using health administrative data to predict the probability that a person who had surgery would develop an SSI within 30 days.METHODSAll patients enrolled in the National Surgical Quality Improvement Program (NSQIP) from 2 sites were linked to population-based administrative datasets in Ontario, Canada. We derived a multivariate model, stratified by surgical specialty, to determine the independent association of SSI status with patient and hospitalization covariates as well as physician claim codes. This SSI risk model was validated in 2 cohorts.RESULTSThe derivation cohort included 5,359 patients with a 30-day SSI incidence of 6.0% (n=118). The SSI risk model predicted the probability that a person had an SSI based on 7 covariates: index hospitalization diagnostic score; physician claims score; emergency visit diagnostic score; operation duration; surgical service; and potential SSI codes. More than 90% of patients had predicted SSI risks lower than 10%. In the derivation group, model discrimination and calibration was excellent (C statistic, 0.912; Hosmer-Lemeshow [H-L] statistic, P=.47). In the 2 validation groups, performance decreased slightly (C statistics, 0.853 and 0.812; H-L statistics, 26.4 [P=.0009] and 8.0 [P=.42]), but low-risk patients were accurately identified.CONCLUSIONHealth administrative data can effectively identify postoperative patients with a very low risk of surgical site infection within 30 days of their procedure. Records of higher-risk patients can be reviewed to confirm SSI status.Infect. Control Hosp. Epidemiol. 2016;37(4):455–465

Sign in / Sign up

Export Citation Format

Share Document